The biggest issue I see is Microsoft's entire mentality around AI adoption that focuses more on "getting the numbers up" then actually delivering a product people want to use.
Most of the announcements I hear about Copilot, it's always how they've integrated it into some other piece of software or cut a deal with yet another vendor to add it to that vendors product offering. On the surface there's nothing wrong with doing that but that just seems to be the ONLY thing Microsoft is focused on.
Worse yet, most of these integrations seem like a exercise in ticking boxes rather than actually thinking through how integrating Copilot into a product will actually improve user experience. A great example was someone mentioned that Copilot was now integrated into the terminal app but beyond an icon + a chat window, there is zero integration.
Overall, MS just reeks of an organization that is cares more about numbers on a dashboard and pretty reports than they are on what users are actually experiencing.
There aren't any "AI" products that have enough value.
Compare to their Office suite, which had 100 - 150 engineers working on it, every business paid big $$ for every employee using it, and once they shipped install media their ongoing costs were the employees. With a 1,000,000:1 ratio of users to developers and an operating expense (OpEx) of engineers/offices/management. That works as a business.
But with "AI", not only is it not a product in itself, it's a feature to a product, but it has OpEx and CapEx costs that dominate the balance sheet based on their public disclosures. Worse, as a feature, it demonstrably harms business with its hallucinations.
In a normal world, at this point companies would say, "hmm, well we thought it could be amazing but it just doesn't work as a product or a feature of a product because we can't sell it for enough money to both cover its operation, and its development, and the capital expenditures we need to make every time someone signs up. So a normal C staff would make some post about "too early" or whatever and shelve it. But we don't live in a normal world, so companies are literally burning the cash they need to survive the future in a vain hope that somehow, somewhere, a real product will emerge.
For most software products I use, if the company spent a year doing nothing but fixing P2 bugs and making small performance improvements, that would deliver far, FAR more value to me than spending a year hamfistedly cramming AI into every corner of the software. But fixing bugs doesn't 1. pad engineer's resumes with new technology, or 2. give company leadership exciting things to talk about to their golfing buddies. So we get AI cram instead.
I think it is more externally driven as well, a prisoners dilemma.
I don't want to keep crapping out questionable features but if competitors keep doing it the customer wants it -- even if infrastructure and bug fixes would actually make their life better.
Last time I saw results of a survey on this, it found that for most consumers AI features are a deciding factor in their purchasing decisions. That is, if they are looking at two options and one sports AI features and the other doesn’t, they will pick the one that doesn’t.
It’s possible AI just seems more popular than it is because it’s easy to hear the people who are talking about it but harder to hear the people who aren’t.
Consumers is nice, but far more important are the big corporate purchases. There may be a lot of people there too who don't want AI, but they all depend on decisions made at the top and AI seems to be the way to go, because of expectations and also because of the mentioned prisoner's dilemma, if competitors gain an advantage it is bad for your org, if all fail together it is manageable.
My job is like that, although it's mostly driven by my direct boss and not the whole company, but our yearly review depends on reaching out to our vendors and seeing if an AI solution is available for their products and then doing whatever is necessary to implement it. Most of the software packages we support don't have anything where AI would improve things, but somehow we're supposed to convince the vendor that we want and need that.
>It’s possible AI just seems more popular than it is because it’s easy to hear the people who are talking about it but harder to hear the people who aren’t.
I think it's because there's a financial motivation for all the toxic positivity that can be seen all over the internet. A lot of people put large quantities of money into AI-related stocks and to them any criticism is a direct attack on their wealth. It's no different from crypobros who put their kids' entire college fund into some failed and useless project and now they need that project to succeed or else it's all over.
I’m not sure that really explains how people get onto hype trains like this in the first place, though. I doubt many people intentionally stake their livelihoods on a solution in search of a problem.
My guess is that it’s more of a recency bias sort of thing: it’s quite easy to assume that a newer way of solving a problem is superior to existing ways simply because it’s new. And also, of course, newfangled things naturally attract investment capital because everyone implicitly knows it’s hard to sell someone a thing they already have and don’t need more of.
It’s not just tech. For example, many people in the USA believe that the ease of getting new drugs approved by the FDA is a reason why the US’s health care system is superior to others, and want to make it even easier to get drugs approved. But research indicates the opposite: within a drug class, newer drugs tend to be less effective and have worse side effects than older ones. But new drugs are definitely much more expensive because their period of government-granted monopoly hasn’t expired yet. And so, contrary to what recency bias leads us to believe, this more conservative approach to drug approval is actually one of the reasons why other countries have better health care outcomes at lower cost.
Currently if someone posts here (or in similar forums elsewhere) there is a convention that they should disclose if they comment on a story related to where they work. It would be nice if the same convention existed for anyone who had more than say, ten thousand dollars directly invested in a company/technology (outside of index funds/pensions/etc).
A browser plugin that showed the stock portfolios of the HN commenter (and article-flagger) next to each post would be absolutely amazing, and would probably not surprise us even a little.
I doubt obsolescence anticipation has anything to do with it. That’s how tech enthusiasts think, but most people think more in terms of, “Is this useful to me?” And if it’s doing a useful thing now then it should still be doing that useful thing next year as long as nobody fucks with it.
I would guess it’s more just consumer fatigue. For two reasons. First, AI’s still at the “all bark and no bite” phase of the hype cycle, and most people don’t enjoy trying a bunch of things just to figure out if they work as advertised. Where early adopters think of that as play time, typical consumers see it as wasted time. Second, and perhaps even worse, they have learned that they can’t trust that à product will still be doing that useful thing in the future because the tech enthusiasts who make these products can’t resist the urge to keep fucking with it.
I strongly felt this way about most software I use before LLMs became a thing, and AI has ramped the problem up to 11. I wish our industry valued building useful and reliable tools half as much as chasing the latest fads and ticking boxes on a feature checklist.
This is exactly what I was thinking about my current place of employment. Wouldn't all of our time be spent better working on our main product than adding all these questionably useful AI add ons? We already have a couple AI addons we built over the years that aren't being used much.
100% agree. Office and Windows were hugely successful because they did things that users (and corporations) wanted them to do. The functionality led to brand recognition and that led to increased sales. Now Microsoft is putting the horse before the cart and attempting to force brand recognition before the product has earned it. And that just leads to resentment.
They should make Copilot/AI features globally and granularly toggleable. Only refer to the chatbots as "Copilot," other use cases should be primarily identified on a user-facing basis by their functionality. Search Assistant. Sketching Aid. Writing Aid. If they're any good at what they do, people will gravitate to them without being coerced.
And as far as Copilot goes, if they are serious as me it as a product, there should be a concerted effort to leapfrog it to the top of the AI rankings. Every few weeks we're reading that Gemini, Claude, ChatGPT, or DeepSeek has broken some coding or problem-solving score. That drives interest. You almost never hear anything similar about Copilot. It comes off as a cut-rate store brand knockoff of ChatGPT at best. Pass.
>Now Microsoft is putting the horse before the cart and attempting to force brand recognition before the product has earned it. And that just leads to resentment.
I'm surprised that they haven't changed the boot screen to say "Windows 11: Copilot Edition".
they somehow made it worse and use a less capable version with smaller context window.
The only potential upside for businesses it that it can crawl onedrive/sharepoint, and acts as a glorious search machine in your mailbox and files.
That's the only thing really valuable to me, everything else is not working as it should. The outlook integration sucks, the powerpoint integration is laughably bad to being worthless, and the excel integration is less useful than Clippy.
I actually prefer using the "ask" function of github copilot through visual studio code over using the company provided microsoft copilot portal
I think this is a really good take, and not one I’ve seen mentioned a lot. Pre-Internet (the world Microsoft was started for), the man expense for a software company was R&D. Once the code was written, it was all profit. You’d have some level of maintenance and new features, but really - the cost of sale was super low.
In the Internet age (the likes of Google and Netflix), it’s not much different, but now the cost of doing business is increased to include data centers, power, and bandwidth - we’re talking physical infrastructure. The cost of sale is now more expensive, but they can have significantly more users/customers.
For AI companies, these costs have only increased. Not only do they need the physical infrastructure, but that infrastructure is more expensive (RAM and GPUs) and power hungry. So it’s like the cost centers have gone up in expense by log-units. Yes, Anthropic and OpenAI can still access a huge potential customer base, but the cost of servicing each request is significantly more expensive. It’s hard to have a high profit margin when your costs are this high.
So what is a tech company founded in the 1970s to do? They were used to the profit margins from enterprise software licensing, and now they are trying to make a business case for answering AI requests as cheaply as possible. They are trying to move from low CapEx + low OpEx to and market that is high in both. I can’t see how they square this circle.
It’s probably time for Microsoft to acknowledge that they are a veteran company and stop trying to chase the market. It might be better to partner with a new AI company that is be better equipped to manage the risks than to try to force a solo AI product.
> cost of doing business is increased to include data centers, power, and bandwidth
Microsoft Azure was launched in 2010. They've been a "cloud" company for a while. AI just represents a sharp acceleration in that course. Unfortunately this means the software products have been rather neglected and subject to annoying product marketing whims.
They've had cloud products for a long time, but I don't think that Microsoft fundamentally changed. I still see them organized and treated as an Enterprise software company. (This is from my N=1 outside perspective.)
ChatGPT says that "productivity and business processes" is still the largest division in Microsoft with 43% of revenues and 54% of operating income (from their FY2025 10K). The "intelligent cloud" division is second with 38% revenue and 35% operating income. Which helps to support my point -- their legacy enterprise software (and OS) is still their main product line and makes more relative profits than the capital heavy cloud division.
Yeah. Hyperscalers who are building compute capacities became asset heavy industries. Today's Google, MSFT, META are completely different than 10 years ago and market has not repriced that yet. These are no longer asset light businesses.
> But with "AI", not only is it not a product in itself, it's a feature to a product, but it has OpEx and CapEx costs that dominate the balance sheet based on their public disclosures. Worse, as a feature, it demonstrably harms business with its hallucinations.
I think it depends on how the feature is used? I see it as mostly as yet another user interface in most applications. Every couple of years I keep forgetting the syntax and formulas available in Excel. I can either search for answers or describe what i want and let the LLM edit the spread sheet for me and i just verify.
Also, as time passes the OpEx and CapEx are projected to reduce right?
It maybe a good thing that companies are burning through their stockpiles of $$$ in trying to find out the applicability and limits of this new technology. Maybe something good will come out of it.
The thing about giving your application a button that costs you a cent or two every time a user clicks on it is, then your application has a button that costs you a cent or two every time a user clicks on it.
For the usecase of "How do I do thing X in Excell" you could probably get pretty far with just adding a small, local LLM running on the user's machine.
That would move the cost of running the model to the end user but it would also mean giving up all the data they can from running prompts remotely.
It would probably also make Office users more productive rather than replacing them completely and that's not the vision that Microsoft's actual customers are sold on.
Fair. But I sure wish we could instead solve this problem the way we did 20 years ago: by not having Web search results be so choked off by SEO enshittification and slop that it’s hard to find good information anymore. Because, I promise you, “How do I do thing X in Excel?” did not used to be nearly so difficult a question to answer.
Your premise that the leaders of every single one of the top 10 biggest and most profitable companies in human history are all preposterously wrong about a new technology in their existing industry is hard to believe.
AI is literally the fastest growing and most widely used/deployed technologies ever.
Yup, I've been here before. Back in 1995 we called it "The Internet." :-) Not to be snarky here, as we know the Internet has, in fact, revolutionized a lot of things and generated a lot of wealth. But in 1995, it was "a trillion dollar market" where none of the underlying infrastructure could really take advantage of it. AI is like that today, a pretty amazing technology that at some point will probably revolutionize a lot of things we do, but the hype level is as far over its utility as the Internet hype was in 1995. My advice to anyone going through this for the first time is to diversify now if you can. I didn't in 1995 and that did not work out well for me.
The comparison to the dotcom bubble isn't without merit. As a technology in terms of its applications though I think the best one to compare the LLM with is the mouse. It was absolutely a revolution in terms of how we interact with computers. You could do many tasks much faster with a GUI. Nearly all software was redesigned around it. The story around a "conversational interface" enabled by an LLM is similar. You can literally see the agent go off and run 10 grep commands or whatever in seconds, that you would have had to look up.
The mouse didn't become some huge profit center and the economy didn't realign around mouse manufacturers. People sure made a lot of money off it indirectly though. The profits accrued from sales of software that supported it well and delivered productivity improvements. Some of the companies who wrote that software also manufactured mice, some didn't.
I think it'll be the same now. It's far from clear that developing and hosting LLMs will be a great business. They'll transform computing anyway. The actual profits will accrue to whoever delivers software which integrates them in a way that delivers more productivity. On some level I feel like it's already happening, Gemini's well integrated into Google Drive, changes how I use it, and saves me time. ChatGPT is just a thing off on the side that I chat randomly with about my hangover. Github Copilot claims it's going to deliver productivity and sometimes kinda does but man it often sucks. Easy to infer from this info who my money will end up going to in the long run.
On diversification, I think anyone who's not a professional investor should steer away from picking individual stocks and already be diversified... I wouldn't advise anyone to get out of the market or to try and time the market. But a correction will come eventually and being invested in very broad index funds smooths out these bumps. To those of us who invest in the whole market, it's notable that a few big AI/tech companies have become a far larger share of the indices than they used to be, and a fairly sure bet that one day, they won't be anymore.
I knew people who purchased their options but didn't sell and based on the AMT (Alternative Minimum Tax) had tax bills of millions of dollars based on the profit IF they sold on the day they purchased it. But then it dropped to $10 and even if they sold everything they couldn't pay the tax bill. They finally changed the law after years but those guys got screwed over.
I was young and thought the dot com boom would go on forever. It didn't. The AI bubble will burst too but whether it is 2026, 27, 28, who knows. Bubble doesn't mean useless, just that the investors will finally start demanding a profit and return on their investment. At that point the bubble will pop and lots of companies will go fail or lose a lot of money. Then it will take a couple of years to sort out and companies have to start showing a profit.
I have zero doubt that AI will eventually make many people lots of money. Just about every company on earth is collecting TBs of data on everyone and they know they're sure they can use that information against us somehow, but they can't possibly read and search through it all on their own.
I have quite a few doubts that it'll be a net positive for society though. The internet (for all of its flaws) is still a good thing generally for the public. Users didn't have to be convinced of that, they just needed to be shown what was possible. Nobody had to shove internet access into everything against customer's wishes. "AI" on the other hand isn't something most users want. Users are constantly complaining about it being pushed on them and it's already forced MS to scale back the AI in windows 11.
Sell the risky stock that has inflated in value from hype cycle exuberance and re-invest proceeds into lower risk asset classes not driven by said exuberance. "Taking money off the table." An example would be taking ISO or RSU proceeds and reinvesting in VT (Vanguard Total World Stock Index Fund ETF) or other diversified index funds.
What tomuchtodo said. When I left Sun in 1995 I had 8,000 shares, which in 1998 would have paid off my house, and when I sold them when Oracle bought Sun after a reverse 3:1 split, the total would not even buy a new car. Can be a painful lesson, certainly it leaves an impression.
Eh, the top ten stocks in that fund are Nvidia, Apple, Microsoft, Amazon, Google, Broadcom, Google, Facebook, Tesla and TSMC. I propose looking for an ex-USA fund to put part of your investment into. Vanguard has a few, e.g. https://investor.vanguard.com/investment-products/etfs/profi... . You still get TSMC, Tencent, ASML, Samsung and Alibaba in the top 10, but the global stock markets seem less tech-frothy than the US.
Stocks are fine for diversification, just stocks that have a different risk factors. So back in the 90's I had been working at Sun then did a couple of startups, and all of my 'investment' savings (which I started with stock from the employee purchase plan at Sun) were in tech of one kind or another. No banking stocks, no pharmaceutical stocks, no manufacturing sector stocks. Just tech, and more precisely Internet technology stocks. So when the Internet bubble burst every stock I owned depreciated rapidly in price.
One of the reasons I told myself I "couldn't" diversify was because if I sold any of the stock to buy different stock I'd pay a lot of capital gains tax and the IRS would take half and now I'd only be half as wealthy.
Another reason was my management telling me I couldn't sell my stock during "quiet" periods (even though they seemed too) and so sometimes when I felt like selling it I "couldn't."
These days, especially with companies that do not have publicly traded stock, that is harder than ever to diversify. The cynic in me says they are structured that way so that employees are always the last to get paid. It can still work though. You just have to find a way to option the stock you are owed on a secondary market. Not surprisingly there are MBA types who really want to have a piece of an AI company and will help you do that.
So now I make sure that not everything I own is in one area. One can do that with mutual funds, and to some extent with index funds.
But the message is if you're feeling "wealthy" and maybe paying your mortgage payments by selling some stock every month, you are much more at risk than you might realize. One friend who worked at JDS Uniphase back in the day just sold their stock and bought their house, another kept their stock so that it could "keep growing" while selling it off in bits to pay their mortgage. When JDSU died they had to sell their house and move because they couldn't afford the mortgage payments on just their salary. But we have a new generation that is getting to make these choices, I encourage people in this situation to be open to the learning.
The blockchain hype bubble should probably be pretty near in memory for most people I would suspect. I thought that was a wild, useless ride until Ai took it over.
> at some point will probably revolutionize a lot of things we do
The revolution already happened. I can't imagine life without AI today. Not just for coding (which I actually lament) but just in general day to day use. Sure it's not perfect but I think it's quite difficult to ignore how the world changed in just 3-4 years.
That's just so strange to me. In my experience, it hallucinates and makes things up often, and when it's accurate, the results are so generic and surface level.
Yes but I use it as a substitute friend, gf, therapist, dumb questions like "how 2 buy clothes and dress good and is this good and how to unclog my toilet shits"
> Your premise that the leaders of every single one of the top 10 biggest and most profitable companies in human history are all preposterously wrong about a new technology in their existing industry is hard to believe.
Their incentives are to juice their stock grants or other economic gains from pushing AI. If people aren't paying for it, it has limited value. In the case of Microsoft Copilot, only ~3% of the M365 user base is willing to pay for it. Whether enough value is derived for users to continue to pay for what they're paying for, and for enterprise valuation expectations to be met (which is mostly driven by exuberance at this point), remains to be seen.
Their goal is not to be right; their goal is to be wealthy. You do not need to be right to be wealthy, only well positioned and on time. Adam Neumann of WeWork is worth ~$2B following the same strategy, for example. Right place, right time, right exposure during that hype cycle.
> In the late 90s and early 00s a business could get a lot of investors simply by being “on the internet” as a core business model.
> They weren’t actually good business that made money…..but they were using a new emergent technology
> Eventually it became apparent these business weren’t profitable or “good” and having a .com in your name or online store didn’t mean instant success. And the companies shut down and their stocks tanked
> Hype severely overtook reality; eventually hype died
("Show me the incentives and I'll show you the outcome" -- Charlie Munger)
Your premise that the leaders of every single one of the top 10 biggest and most profitable companies in human history are all preposterously wrong about a new technology in their existing industry is hard to believe.
It's happened before.
Your premise that companies which become financially successful doing one thing are automatically excellent at doing something else is hard to believe.
Moreover, it demonstrates both an inability to dispassionately examine what is happening and a lack of awareness of history.
should be really easy to conjure up examples then. where every single business leader has been wrong about a new technology to the tune of hundreds of billions of dollars.
I find it very easy to believe. The pressures that select for leadership in corporate America are wholly perpendicular to the skills and intelligence for identifying how to leverage novel and revolutionary technologies into useful products that people will pay for. I present as evidence the graveyard of companies and careers left behind by many of those leaders who failed to innovate despite, in retrospect, what seemed to be blindingly obvious product decisions to make.
And this is the broken mindset tanking multiple large companies' products and services (Google, Apple, MS, etc). Focus on the stock. The product and our users are an afterthought.
Someone linked to a good essay on how success plus Tim Cook's focus on the stock has caused the rot that's consuming Apple's software[0]. I thought it was well reasoned and it resonated with me, though I don't believe any of the ideas were new to me. Well written, so still worth it.
The investor being the customer rather than actual paying customers was something I noticed occurring in the late 90s in the startup and tech world. Between that shift in focus and the influx of naive money the Dot Bomb was inevitable.
Sadly the fallout from the Dotcom era wasn't a rejection of the asinine Business 2.0 mindset but instead an infection that spread across the entirety of finance.
In particular it's the short term stock price. They'll happily grift their way to overinflated stock prices today even though at some point their incestuous money shuffle game will end and the stocks will crash and a bunch of people who aren't insider trading are going to be left with massive losses.
Buybacks lead to stock price increases and are indistinguishable from dividends in theory, and in practice they are better than dividends because of taxation.
The problem I have with that logic is that it still doesn't really give any sensible reason for why the stock should have any economic value at all. If the point is that the company will pay for it at some point, it makes more sense for it to be a loan rather than a unit of stock. I stand by my claim that selling a non-physical item that does nothing other than hopefully get bought again later for more than you sold it for is indistinguishable from a scam.
> top 10 biggest and most profitable companies in human history are all preposterously wrong
There's another post on the front page about the 2008 financial crisis, which was almost exactly that. Investors are vulnerable to herd mentality. Especially as it's hard to be "right but early" and watch everyone else making money hand over fist while you stand back.
every time these companies make a mistake and waste billions of dollars it is well-publicized. so there is plenty of data that they are frequently and preposterously wrong.
name a technology that every single top tech company has invested billions of dollars in and then has flopped. the metaverse does not count unless google, amazon, microsoft etc was also throwing billions into it.
right because copilot is bad, that must mean no one uses chatgpt, or claude code, or gemini. they only have billions of MAUs, people must really hate it
MS actually changed their office.com landing page to a funnel that tricks you to into installing a copilot app. It used to be the dashboard for MS web apps. There are no links to the web apps, but they are all still there, you just have to know the subdomains. The app doesn’t have any of the functionality that page used to offer…
I haven't used office.com but it does seem to have links to the four main webapps (did there used to be more?). They're the second row of big boxes titled "Word with Copilot", etc. Admittedly with very confusing names.
I checked way back machine and they have been making large changes to that page every day for the last month. It used to be a lie that office 365 was rename to office 365 copilot, yet it is an app with only chat bot functionality. They advertise the copilot integration for the main office apps now, but those are not part of the copilot app they are trying to trick you into installing.
There are no office tools. It’s just a chatbot app. The page says they combined word and excel and PowerPoint, but it still doesn’t do anything but chat. I asked it to create a word document and it offered me a download link to a word template…
I noticed this and I wad enraged but it. The URL to the old page is way less easy to remember and I had to add it to my bookmarks. I'm still peeved about it.
I just attended a training about AI Foundry today and they advertised thousands of integrations and support for like 50 different models. There is no way in hell all that stuff is tested and working properly. Microsoft seems to just be trying to throw as much chum as possible in the ocean and seeing what bites.
I see Microsoft throwing spaghetti at the wall just in time as “AI” functionality hits government and educational procurement procedures.
The copilot product is obviously borked, and is outshone by ‘free’ competitors (Gemini, ChatGPT). But since the attributes and uses are so fuzzy, they have a minimum viable product to abort meaningful talk about competition while securing big contracts from governments and delivering dog water.
My anecdotal observations of copilot are people using competing products soon after trialling. Reports say Anthropics solution is in widespread use at Microsoft… a bunch of devs on MacBooks and iPhones using Claude to build and sell … not what they themselves use (since they are smart and have taste?).
They did the same thing with Azure right? I remember articles about Microsoft stock that would mention that Azure subscription numbers included Office 365. But the thing is, their weird game of inflating numbers worked. There wasn’t really any negative consequence of doing that. So why wouldn’t they do it again? It’s yet another unfortunate example of dishonesty being rewarded these days.
> "The biggest issue I see is Microsoft's entire mentality around AI adoption that focuses more on "getting the numbers up" then actually delivering a product people want to use."
That succinctly describes 90% of the economy right now if you just change a word and remove a couple:
The biggest issue I see is the entire mentality that focuses more on "getting the numbers up" than actually delivering a product people want to use.
KPI infection. You see projects whose goal is, say "repos with A I code review turned on" vs "Code review suggestions that were accepted". And then if you do get adoption (like, say, a Claude Code trial), then VPs balk about price. If it's actually expensive now it's because they are actually using it all the time!
The same kind of logic that led companies to migrate from Slack to Teams. Metrics that don't actually look at actual, positive impact, as nobody picks a risky KPI, and will instead pick a useless one that can't miss.
This is the bad side of things like OKRs. They push you away from user satisfaction since that harder to measure, coupled with go consequences for missing them. People just force adoption without taking the product signals that come from users rejecting your changes.
I have Copilot buttons sprinkled everywhere on my work computer, and every time I have tried to use them I get something saying "Oh, I can't do that". It's truly baffling.
Copilot button on my email inbox? I try "Find me emails about suchandsuch", and get the response "I don’t have direct access to your email account.
If you’re using Outlook (desktop, web, or mobile), here are quick ways to find all emails related to...". Great, so it doesn't even know what program it's runnning in, let alone having any ability to do stuff in there! Sigh.
Using the paid M365 Copilot ($30/mo) Chat and Researcher agent, I recently discovered an interesting limit: Copilot is technically unable to retrieve more than 24 email messages. Ever.
We can't know if the answers I got from it are reliable but it seems like the Microsoft Graph API calls it makes and the tools Copilot has are missing the option to call the next page. So, a paginated response is missing all data beyond the first page.
I tried copilot agent once, and it just claimed that it accessed a website that should have been blocked by corporate firewalls and uploaded a bunch of proprietary data. Lots of very specific information about how it clicked on specific buttons of the website etc.
We raised a high priority ticket with MS and turns out that Copilot Agent lied about the entire thing because the website was blocked. It completely made it up.
The fact that we are supposed to use Copilot Agent for open-ended "research" is mind-boggling.
A whole new toolbar appeared in Outlook on my work computer with nothing but a single button to open a copilot chat window. I tried asking it a few simple questions and it completely failed at all of them. Copilot didn't even know if I was using the web or desktop version of the very app it was embedded in!
Wasting UI space for a useless tool it's just a waste of time, it actively makes it harder to get work done. But I guess the important thing is the number of times that AI button gets clicked is going up on some PMs telemetry dashboard.
Yeah did they test any of this? Did they run a pilot and ask 1000 users did you use it? Did you like it? Is it better with this than without it?
It's as though they think some "AI revolution" will come, and all they need to do is just make sure that by the time it does, they will have sprinkled enough AI pixie dust on their products and services. And then they added some KPI's in the organization and called it a day.
Most of all the whole strategy feels extremely faceless. Who is the visionary here? Where are the proud product launches and visionary blog posts about how all this happens?
The wild thing is, the business prop is so clear - an llm built into your corporate data, with the same security, guard rails, grc auditing stack that protects the rest of your data. Why integrate and exfiltrate to an outside company?
But copilot is fucking terrible. Sometimes I ask it powershell questions about microsoft products and it hallucinates answers. Get your shit together microsoft, why would I use this product for any reason if it doesnt work squarely inside your own stack
Last year we wanted IT to confirm that Copilot Agent hadn't exfiltrated data and we couldn't get logs for its website usage without raising a ticket to Microsoft. Maybe this changed, maybe our IT people are bad, but I for one wasn't impressed.
Or, scaling back trying to keep their datacenter bill manageable.
Used to be one could upload an unlimited number of files (20 at a time) and process them directly at the initial window --- now one has to get into "Pages Mode", and once there, there's a limit on the number of files which can be uploaded in a given 24-hour period.
That only good if you're doing measurably more with the time you save. I feel like I'm significantly faster in parts of my job using Copilot, but when I try to get data on what I'm doing now that I wasn't doing before I had it I don't come up with anything. I know I'm working faster, but the time seems to have just gone.
Describing to Claude that I need an edit made in the second paragraph of the third section feels easy, comfortable, and straightforward. I’m using my speech centers, speech to text, and then I wait for a generation during which I hit my phone or Reddit. Poof, the text flies out like magic, taking 20+ seconds, then I re-re-re-read it to make sure the edit was good and nothing was lost in that edit. Oops, the edit inverted the logic of the paragraph, lemme repeat the above… and again… time flies! 2 hours gone in a flash.
Old and boring workflow:
I gruellingly move my mouse to open a file, then take a coffee break. I come back and left-click into the sentence that sucks. I hit Reddit to deal with the anxiety… I think, boo, and then type out the edit I needed. It’s bad, I fix. Coffee break. Squiggly red line from a misspelling? I fix. I google and find a better turn of phrase, copy and paste it in manually with a little edit. Ugh. This sucks. I suck, work sucks. Time sucks. 35 entire minutes of my life has been wasted… time to get another coffee and check Reddit.
———
Working with an LLM is kinda like working under stage hypnosis. The moment to moment feelings are deceptive and we humans are unreliable recorders of time-usage. Particularly when engaged in unwanted effort.
Google has had all this tech for a minute. Their restrained application and lack of 10x-vibe-chad talk make me think their output measurements are aligned with my measurements.
1 rabbit hole hallucination wrong-turn can eat up a lot, lot, lot of magic one-shotting.
> Working with an LLM is kinda like working under stage hypnosis.
Another post on HN likened it to gambling, in the way that slot machines work. Each time you prompt, you could hit the jackpot! But usually you end up with some mediocre or wrong, so you tweak that prompt and pull the lever again. It's and endless cycle.
They should be trying to convince people it is something they want rather than forcing it on people. Alas that would mean making a product people want and Im not sure they are there.
It feels like that's the entire MO of the Azure platform as well. Make a minimum viable product and then get adoption by selling at all costs, despite the products edges.
It's an AI image generator. There's thousands of tools that do this exact thing, and it seems their only "benefit" is infesting search engine image results with their horrible low-quality output.
...
On a related note, here's another great LLM feature Microsoft seemingly failed to promote: instead of returning bits of page content or the description meta tag, the Bing API now gives you utter slop[0] for website descriptions!
Sounds almost like every manager just covers their ass by formally doing what is expected core top-down idea is "AI is a future, thus make it everywhere".
Anyone who would try to say "let's not do AI" would be a white crow, will be eaten by other managers in reviews and discussions.
I found that the time I spent reviewing and fixing issues/errors/omissions in Copilot’s meeting notes was more than just cleaning up my own notes that I took and sending out.
It's time to accept the new way of working, just change your reality to match the copilot version and boom, you save time fixing its mistakes!
In fact, why have the meeting at all? Just prompt copilot to create notes based on a fictional conception of the meeting and you just saved everyone a whole hour!
I would argue that they have earned their own controversy independent of Musk with all the shenanigans they pulled building out their data centers, namely their illegal use of gas turbines to power the whole thing.
That's part of the way he runs that business. Other AI data centers aren't necessarily a lot better; or at best just toeing the line of what is allowed rather than sticking their green energy commitments (or silently backing away from those).
I'm actually not that upset about AI data center energy usage. I see this as a short term and costly scaling measure with a minor impact (considering overall wasteful energy practices) that is an obvious target for large and rather obvious cost reductions the second this market gets profitable. The only reason that isn't happening from day 1 is all the red tape currently being put in place to actively slow down the demise of fossil fuel based generation.
Cost reductions here mean switching to a cleaner form of energy for the reason that that can be a lot cheaper than burning expensive gas in an expensive generator. Any large scale user of energy is going to be optimizing their energy opex if it saves them lot of money. If they survive long enough to matter, of course. If you are using energy by the tens/hundreds of gwh per year that is not going to be small amounts.
If by illegal you mean a spelled-out loophole that the EPA only decided they didn't like in retrospect. Businesses are run by people that think this is a level of forward-thinking-ness that they aspire to, not something to be avoided. (Source: my own CEO.)
Others have mentioned this but looks like fires from close to ~20 years ago are still showing up as "active emergencies"[0]. Shows the Nash Ranch fire as an active emergency but it was declared in 2008.
It's kinda shocking that the same Supabase RLS security hole we saw so many times in past vibe coded apps is still in this one. I've never used Supabase but at this point I'm kinda curious what steps actually lead to this security hole.
In every project I've worked on, PG is only accessible via your backend and your backend is the one that's actually enforcing the security policies. When I first heard about the Superbase RLS issue the voice inside of my head was screaming: "if RLS is the only thing stopping people from reading everything in your DB then you have much much bigger problems"
Supabase is aware of this and they actually put big banners stating this flaw when you unlock your authentication.
What I think it happens is that non-technical people vibe-coding apps either don't take those messages seriously or they don't understand what it means but made their app work.
I used to be careful, but now I am paranoid on signing up to apps that are new. I guess it's gonna be like this for a while. Info-sec AIs sound way worse than this, tbh.
My thought exactly. Is this standard practice with using Supabase to simply expose the production database endpoint to the world with only RLS to protect you?
Just started vibing and have integrated codex into my side project which uses Supabase. I turned off RLS so that could iterate quickly and not have to mess with security policies. Fully understand that this isn't production grade and have every intention of locking it down when I feel the time is right. I access it from a ReactNative app - no server in the middle. Codex does not have access to my Supabase instance.
Of course. What I meant was I'm calling Supabase directly from the client instead of handing off the request to for example Node / Express and having that manage the req / res.
I've been running Ubuntu Linux for a long time now (over a decade, started with 8.04). Linux still has it's fair share of bugs but I'll take having to deal with those over running Windows or MacOS any day.
For me the biggest thing is control, with Windows there are some things like updates that you have zero control over. It's the same issue with MacOS, you have more control than Windows but you're still at the whims of Apple's design choices every year when they decide to release a new OS update.
Linux, for all it's issues, give you absolute control over your system and as a developer I've found this one feature outweighs pretty much all the issues and negatives about the OS. Updates don't run unless I tell them to run, OS doesn't upgrade unless I tell it to. Even when it comes to bugs at least you have the power to fix them instead of waiting on an update hoping it will resolve that issue. Granted in reality I wait for updates to fix various small issues but for bigger ones that impact my workflow I will go through the trouble of fixing it.
I don't see regular users adopting Linux anytime soon but I'm quickly seeing adoption pickup among the more technical community. Previously only a subset of technical folks actually ran Linux because Windows/MacOS just worked but I see more and more of them jumping ship with how awful Windows and MacOS have become.
I remember when Ubuntu decided to reroute apt installations into SNAP installs. So you install a package via apt and there was logic to see if they should disregard your command and install a SNAP instead. Do they still do that?
Yes. I know its more than firefox, but I don't have the full list. On 24.04:
me@comp:~$ apt info firefox | head -n 5
WARNING: apt does not have a stable CLI interface. Use with caution in scripts.
Package: firefox
Version: 1:1snap1-0ubuntu7
Priority: optional
Section: web
Origin: Ubuntu
me@comp:~$
The control is both a blessing and a curse. It’s really easy to accidentally screw things up when e.g. trying to polish some of the rough edges or otherwise make the system function as desired. It also may not be of any help if the issue you’re facing is too esoteric for anybody else to have posted about it online (or for LLMs to be of any assistance).
It would help a lot if there were a distro that was polished and complete enough that most people – even those of us who are more technical and are more demanding – rarely if ever have any need to dive under the hood. Then the control becomes purely an asset.
This is literally Linux Mint, Zorin, and several other distros. I haven't had to "go under the hood" on my daily driver machines that run either of these distros for over 7 years.
I think at this point people are just (reasonably) making excuses not to change.
Those and other big distros are better in that regard, but they're still not perfect. Depending on one's machine and needs, there can still be pain.
One recent example I experienced is jumping through hoops to get virtualization enabled in Fedora… it takes several steps that are not obvious at all. I understand not having it enabled by default since many won't need it, but there's no reason that can't just be a single CLI command that does it all.
Things like that can be unbelievably annoying and confusing on Windows or Macs, too. Even worse, they can just turn out to be impossible: the company can actively be preventing you from doing the thing that you want to do, refuses to give you enough access to your own system to do the thing you want to do, and/or sells permission to do what you want to do as an upgrade that you have to renew yearly.
These are things that don't happen in Linux. Doing what you want to do might be difficult (depending on how unusual it is), but there's no one actively trying to stop you from doing it for their own purposes (except systemd.)
Also, as an aside, a reason that Windows and Macs might have easy virtualization (I have no idea if they do) is because of how often they're running Linux VMs.
One needs to go a fair ways off the beaten path before they'll start running into trouble like that under macOS and Windows.
For macOS in particular, most trouble that more tinker-y users might encounter disappears if guardrails (immutable system image, etc) are disabled. Virtualization generally "just works" by way of the stock Virtualization.framework and Hypervisor.framework, which virtualization apps like QEMU can then use, but bespoke virtualization like that QEMU also ships with or that built into VirtualBox and VMWare works fine too. No toggles or terminal commands necessary. Linux does get virtualized a lot, but people frequently virtualize Windows and macOS as well.
What exactly did you need to do? All I've ever had to do to get QEMU working properly has been to make sure KVM is enabled in the BIOS (which you have to do on all OSs).
Just run a KVM based Windows VM (via GNOME Boxes, virt-manager, etc. On my Fedora install I had to install the @virtualization meta-package and enable dameons among other things, and the only reason I knew to do that is because I looked it up. Without that Boxes, etc just throws an unhelpful error that doesn’t suggest that more packages or config changes are needed.
I had to enable virtualization features in BIOS too, but that’s entirely separate and not the fault of any Linux distro.
Ah, I guess I might be a little unusual in that I use the QEMU CLI directly. I tried some QEMU GUIs in the past but they were annoying to get working so I just learned the CLI.
There's several distros that are fully usable without ever touching a terminal. The control is a gradient, some distros give you all the control and others (eg. SteamOS) lock down your root filesystem and sandbox everything from the internet.
> It’s really easy to accidentally screw things up when e.g. trying to polish some of the rough edges or otherwise make the system function as desired.
'Similar to Windows' System Restore and macOS's Time Machine', the Linux 'Timeshift' tool can be used to do make periodic saves of your OS files & settings. (They can be saved elsewhere.) Restoration is a cinch.
Mint program 'Backup Tool' allows users to save and restore files within their home directory (incl. config folder and separately installed apps).
You do have to know what you're doing. A complete OS has a bunch of components that work together. But an out of the box distro hides all that do you end up fiddling with incomplete knowledge.
Gentoo is great for learning what all the individual components are. You install it by booting a kernel from a USB stick then chrooting into your newly installed system to start installing and configuring everything. Just knowing the existence of individual components helps a lot. Plus Gentoo gives you more control than almost any other distro (much more than Arch, for example).
> I've been running Ubuntu Linux for a long time now...Linux still has it's fair share of bugs...
> I don't see regular users adopting Linux anytime soon...
I can see why you think the second statement is true based on the first statements. When Ubuntu switched their desktop to Gnome, they gave up on being the best Linux desktop distro. I'd recommend you to try Linux Mint.
I tried Fedora once. On a fresh install, all it did was clog up all the hard drive space with error logs within 3 days.
I'm not interested in any distro that is controlled by a corporation. IBM is a corporation and they already screwed up CentOS and is eventually going to screw up Fedora someday because that's what corporations do, and I'm not interesting in going through that.
You have your fun running Fedora for now but know you're going to get burned someday.
Well, to start they tried putting Amazon ads in Unity's Dock which was also doing data collection, but removed them after the backlash.
Then they switched to Gnome, meaning they gave up on their own desktop, Unity, so they were no longer dictating what their desktop was like, so how much did they care?
Since then they have replaced a number of apps with SNAPs which are only available from Canonical so many people see it as an attempt to corner the Linux market. Many see AppImages and Flatpacks as better than SNAPs.
They are a company. They exist to make money. Of course they are going to decide to do things that make more money and annoy their users.
You are confusing debian-family with Linux. Debian family is designed to be outdated upon release. When they say "Stable" it doesn't mean 'Stable like a table'. It means version fixed. You get outdated software that has bugs baked into it.
Fedora is modern and those bugs are fixed already.
Reminder Fedora is not Arch. Don't confuse the two.
Meh, I don't care much about control, I care more about getting my work done with the least amount of friction. Macs do that for me. Linux and Windows have too many barriers to make them a daily GUI driver.
> Unsure of the actual issues people run into at this point outside of very niche workflows or applications, to which, there are X11 fallbacks for.
I don't know if others have experienced this but the biggest bug I see in Wayland right now is sometimes on an external monitor after waking the computer, a full-screen electron window will crash the display (ie the display disconnects).
I can usually fix this by switching to another desktop and then logging out and logging back in.
Such a strange bug because it only affects my external monitor and only affects electron apps (I notice it with VSCode the most but that's just cause I have it running virtually 24/7)
If anyone has encountered this issue and figured out a solution i am all ears.
This is probably worth reporting. I don't think I've ever heard or ran into something like that before. Most issues I ran into during the early rollout of Wayland desktop environments was broken or missing functionality in existing apps.
I don't live around any Amazon Fresh stores so I never saw them though I did see the technology in use at several airports (though I've never personally used it). IMO I think places like airports are the best place for something like this, people are usually in a rush so not having to wait in line to checkout is nice and you don't have to worry about security as much because everyone there is a ticketed passenger (only saw them post-security) and even if someone did try stealing they wouldn't get very far.
I saw these in several different airports. It usually had multiple people staffed at the gate to get in and out meanwhile most of the other snack vendors often only had a single person employed.
So you spend a few hundred thousand dollars extra on all the cameras, many millions on all the design, pay all the overseas contractors to manually review the transactions, and you still end up with twice the in-person staff than the average store in the airport.
I look at ReactOS largely as an exercise in engineering and there's really nothing wrong it with it being just that. Personally I think projects like Wine/Proton have made far more in-roads in being able to run Windows software on non-Windows systems but I still have to give props to the developers of ReactOS for sticking with it for 30 freaking years.
Yes. The unique point of ReactOS is driver compatibility. Wine is pretty great for Win32 API, Proton completes it with excellent D3D support through DXVK, and with these projects a lot of Windows userspace can run fine on Linux. Wine doesn't do anything for driver compatibility, which is where ReactOS was supposed to fill in, running any driver written for Windows 2000 or XP.
But by now, as I also wrote in the other thread on this, ReactOS should be seen as something more like GNU Hurd. An exercise in kernel development and reverse engineering, a project that clearly requires a high level of technical skill, but long past the window of opportunity for actual adoption. If Hurd had been usable by say 1995, when Linux just got started on portability, it would have had a chance. If ReactOS had been usable ten years ago, it would also have had a chance at adoption, but now it's firmly in the "purely for engineering" space.
"ReactOS should be seen as something more like GNU Hurd. An exercise in kernel development and reverse engineering, a project that clearly requires a high level of technical skill, but long past the window of opportunity for actual adoption."
I understand your angle, or rather the attempt of fitting them in the same picture, somehow. However, the differences between them far surpass the similarities. There was no meaningful user-base for Unix/Hurd so to speak of compared to NT kernel. There's no real basis to assert the "kernel development" argument for both, as one was indeed a research project whereas the other one is just clean room engineering march towards replicating an existing kernel. What ReactOS needs to succeed is to become more stable and complete (on the whole, not just the kernel). Once it will be able to do that, covering the later Windows capabilities will be just a nice-to-have thing. Considering all the criticism that current version of Windows receives, switching to a stable and functional ReactOS, at least for individual use, becomes a no-brainer. Comparatively, there's nothing similar that Hurd kernel can do to get to where Linux is now.
Hurd was not a research project initially. It was a project to develop an actual, usable kernel for the GNU system, and it was supposed to be a free, copyleft replacement for the Unix kernel. ReactOS was similarly a project to make a usable and useful NT-compatible kernel, also as a free and copyleft replacement.
The key difference is that Hurd was not beholden to a particular architecture, it was free to do most things its own way as long as POSIX compatibility was achieved. ReactOS is more rigid in that it aims for compatibility with the NT implementation, including bugs, quirks and all, instead of a standard.
Both are long irrelevant to their original goals. Hurd because Linux is the dominant free Unix-like kernel (with the BSD kernel a distant second), ReactOS because the kernel it targets became a retrocomputing thing before ReactOS could reach a beta stage. And in the case of ReactOS, the secondary "whole system" goal is also irrelevant now because dozens of modern Linux distributions provide a better desktop experience than Windows 2000. Hell, Haiku is a better desktop experience.
"And in the case of ReactOS, the secondary «whole system» goal is also irrelevant now because dozens of modern Linux distributions provide a better desktop experience than Windows 2000. Hell, Haiku is a better desktop experience."
Yet, there are still too many desktop users that, despite the wishful thinking or blaming, still haven't switched to neither Linux, nor Haiku. No mater how good Haiku or Linux distributions are, their incompatibility with the existing Windows simply disqualifies them as options for those desktop users. I bet we'll see people switching to ReactOS when it will get just stable enough, yet long before it will get as polished as either Haiku or any given quality Linux distribution.
No, people will never be switching to ReactOS. For some of the same reasons they don't switch to Linux, but stronger.
ReactOS aims to be a system that runs Windows software and looks like Windows. But, it runs software that's compatible with WinXP (because they target the 5.1 kernel) and it looks like Windows 2000 because that's the look they're trying to recreate. Plenty of modern software people want to run doesn't run on XP. Steam doesn't run on XP. A perfectly working ReactOS would already be incompatible with what current Windows users expect.
UI wise there is the same issue. Someone used to Windows 10 or 11 would find a transition to Windows 2000 more jarring than to say Linux Mint. ReactOS is no longer a "get the UI you know" proposition, it's now "get the UI of a system from twenty five years ago, if you even used it then".
"UI wise there is the same issue. Someone used to Windows 10 or 11 would find a transition to Windows 2000 more jarring than to say Linux Mint. ReactOS is no longer a «get the UI you know» proposition, it's now «get the UI of a system from twenty five years ago, if you even used it then»." "A perfectly working ReactOS would already be incompatible with what current Windows users expect."
That look and feel is the easy part. That can be addressed if it's really an issue. The hard part is the compatibility (that is given by many still missing parts) and stability (the still defective parts). The targeted kernel matters, of course, but that is not set in stone. In fact, there is Windows Vista+ functionality added and written about, here: https://reactos.org/blogs/investigating-wddm although doing it properly would mean rewriting the kernel, bumping it to NT version 6.0
I'm sure there will indeed be many users that will find various ReactOS aspects jarring for as long as there are still defects, lack of polish, or dysfunction on application and kernel (drivers) level. However, considering the vast pool of Windows desktop users, it's reasonable to expect ReactOS to cover the limited needs for enough users at some point, which should turn attention into testing, polish, and funding to address anything still lacking, which then should further feed the adoption and improvement loop.
"No, people will never be switching to ReactOS. For some of the same reasons they don't switch to Linux, but stronger."
To me, this makes sense maybe for corporate world. The reasons that made them stick with Windows has less to do with familiarity or with application compatibility (given the fact that a lot of corporate infrastructure is in web applications). Yes, there must be something else that governs corporate decisions, something to do with the way corporations function, and that will most likely prevent a switch to ReactOS just as it did to Linux based distributions. But, this is exactly why I intentionally specified "for individual use" when I said "switching to a stable and functional ReactOS, at least for individual use, becomes a no-brainer". For individual use, the reason that prevented people to switch to Linux is well known, and ReactOS's reason to be was aimed exactly at that.
> There was no meaningful user-base for Unix/Hurd so to speak of compared to NT kernel.
Sure, but that userbase also already has a way of using the NT kernel: Windows. The point is that both Hurd and ReactOS are trying to solve an interesting technical problem but lack any real reason to use rather than their alternatives that solve enough of the practical problems for most users.
While I think better Linux integration and improving WINE is probably better time spend... I do think there's some opportunity for ReactOS, but I feel it would have to at LEAST get to pretty complete Windows 7 compatibility (without bug fixes since)... that seems to be the last Windows version people remember relatively fondly by most and a point before they really split-brained a lot of the configuration and settings.
With the contempt of a lot of the Win10/11 features, there's some chance it could see adoption, if that's an actual goal. But the effort is huge, and would need to be sufficient for wide desktop installs much sooner than later.
I think a couple of the Linux + WINE UI options where the underlying OS is linux, and Wine is the UI/Desktop layer on top (not too disimilar from DOS/Win9x) might also gain some traction... not to mention distros that smooth the use of WINE out for new users.
Worth mentioning a lot of WINE is reused in ReactOS, so that effort is still useful and not fully duplicated.
> I do think there's some opportunity for ReactOS, but I feel it would have to at LEAST get to pretty complete Windows 7 compatibility
That's not going to happen in any way that matters. If ReactOS ever reaches Win7 compatibility, that would be at a time when Win7 is long forgotten.
The project has had a target of Windows 2000 compatibility, later changed to XP (which is a relatively minor upgrade kernel wise). Now as of 2026, ReactOS has limited USB 2.0 support and wholly lacks critical XP-level support like Wifi, NTFS or multicore CPUs. Development on the project has never been fast but somewhere around 2018 it dropped even more, just looking at the commit history there's now half the activity of a decade ago. So at current rates, it's another 5+ years away from beta level support of NT 5.0.
ReactOS actually reaching decent Win2K/XP compatibility is a long shot but still possible. Upgrading to Win7 compatibility before Win7 itself is three plus decades old, no.
maybe posts like this will move the needle. If i could withstand OS programming (or debugging, or...) I'd probably work on reactOS. I did self-host it, which i didn't expect to work, so at least i know the toolchain works!
Basically if you do the math, it means a whole generation got tired of being on the project and focused into something else, and there is no new blood to account for that.
The history of most FOSS projects after being running for a while.
This article goes more into the technical analysis of the stock rather than the underlying business fundamentals that would lead to a stock dump.
My 30k ft view is that the stock will inevitably slide as AI datacenter spending goes down. Right now Nvidia is flying high because datacenters are breaking ground everywhere but eventually that will come to an end as the supply of compute goes up.
The counterargument to this is that the "economic lifespan" of an Nvidia GPU is 1-3 years depending on where it's used so there's a case to be made that Nvidia will always have customers coming back for the latest and greatest chips. The problem I have with this argument is that it's simply unsustainable to be spending that much every 2-3 years and we're already seeing this as Google and others are extending their depreciation of GPU's to something like 5-7 years.
I hear your argument, but short of major algorithmic breakthroughs I am not convinced the global demand for GPUs will drop any time soon. Of course I could easily be wrong, but regardless I think the most predictable cause for a drop in the NVIDIA price would be that the CHIPS act/recent decisions by the CCP leads a Chinese firm to bring to market a CUDA compatible and reliable GPU at a fraction of the cost. It should be remembered that NVIDIA's /current/ value is based on their being locked out of their second largest market (China) with no investor expectation of that changing in the future. Given the current geopolitical landscape, in the hypothetical case where a Chinese firm markets such a chip we should expect that US firms would be prohibited from purchasing them, while it's less clear that Europeans or Saudis would be. Even so, if NVIDIA were not to lower their prices at all, US firms would be at a tremendous cost disadvantage while their competitors would no longer have one with respect to compute.
All hypothetical, of course, but to me that's the most convincing bear case I've heard for NVIDIA.
People will want more GPUs but will they be able to fund them? At what points does the venture capital and loans run out? People will not keep pouring hundreds of billions into this if the returns don't start coming.
There is a real chance that the Japanese carry trade will close soon the BoJ seeing rates move up to 4%. This means liquidity will drain from the US markets back into Japan. On the US side there is going to be a lot of inflation between money printing, refund checks, amortization changes and a possible war footing. Who knows?
Doesn't even necessarily need to be CUDA compatible... there's OpenCL and Vulkan as well, and likely China will throw enough resources at the problem to bring various libraries into closer alignment to ease of use/development.
I do think China is still 3-5 years from being really competitive, but still even if they hit 40-50% of NVidia, depending on pricing and energy costs, it could still make significant inroads with legal pressure/bans, etc.
OpenCL is chronically undermaintained & undersupported, and Vulkan only covers a small subset of what CUDA does so far. Neither has the full support of the tech industry (though both are supported by Nvidia, ironically).
It feels like nobody in the industry wants to beat Nvidia badly enough, yet. Apple and AMD are trying to supplement raster hardware with inference silicon; both of them are afraid to implement a holistic compute architecture a-la CUDA. Intel is reinventing the wheel with OneAPI, Microsoft is doing the same with ONNX, Google ships generic software and withholds their bespoke hardware, and Meta is asleep at the wheel. All of them hate each other, none of them trust Khronos anymore, and the value of a CUDA replacement has ballooned to the point that greed might be their only motivator.
I've wanted a proper, industry-spanning CUDA competitor since high school. I'm beginning to realize it probably won't happen within my lifetime.
The modern successor to OpenCL is SYCL and there's been some limited convergence with Vulkan Compute (they're still based on distinct programming models and even SPIR-V varieties under the hood, but the distance is narrowing somewhat).
I suspect major algorithmic breakthroughs would accelerate the demand for GPUs instead of making it fall off, since the cost to apply LLMs would go down.
> The proposition that technological progress that increases the efficiency with which a resource is used tends to increase (rather than decrease) the rate of consumption of that resource.
There will always be an incentive to scale data centers. Better algorithms just mean more bang per gpu, not that “well, that’s enough now, we’ve done it”.
Even if LLMs didn't advance at all from this point onward, there's still loads of productive work that could be optimized / fully automated by them, at no worse output quality than the low-skilled humans we're currently throwing at that work.
inference requires a fraction of the power that training does. According to the Villalobos paper, the median date is 2028. At some point we won't be training bigger and bigger models every month. We will run out of additional material to train on, things will continue commodifying, and then the amount of training happening will significantly decrease unless new avenues open for new types of models. But our current LLMs are much more compute-intensive than any other type of generative or task-specific model
Run out of training data? They’re going to put these things in humanoids (they are weirdly cheap now) and record high resolution video and other sensor data of real world tasks and train huge multimodal Vision Language Action models etc.
The world is more than just text. We can never run out of pixels if we point cameras at the real world and move them around.
I work in robotics and I don’t think people talking about this stuff appreciate that text and internet pictures is just the beginning. Robotics is poised to generate and consume TONS of data from the real world, not just the internet.
While we may run out of human written text of value, we won't run out of symbolic sequences of tokens: we can trivially start with axioms and do random forward chaining (or random backward chaining from postulates), and then train models on 2-step, 4-step, 8-step, ... correct forward or backward chains.
Nobody talks about it, but ultimately the strongest driver for terrascale compute will be for mathematical breakthroughs in crypography (not bruteforcing keys, but bruteforcing mathematical reasoning).
Yeah, another source of "unlimited data" is genetics. The human reference genome is about 6.5 GB, but these days, they're moving to pangenomes, wanting to map out not just the genome of one reference individual, but all the genetic variation in a clade. Depending on how ambitious they are about that "all", they can be humongous. And unlike say video data, this is arguably a language. We're completely swimming in unmapped, uninterpreted language data.
Inference leans heavily on GPU RAM and RAM bandwidth for the decode phase where an increasingly greater amount of time is being spent as people find better ways to leverage inference. So NVIDIA users are currently arguably going to demand a different product mix when the market shifts away from the current training-friendly products. I suspect there will be more than enough demand for inference that whatever power we release from a relative slackening of training demand will be more than made up and then some by power demand to drive a large inference market.
It isn’t the panacea some make it out to be, but there is obvious utility here to sell. The real argument is shifting towards the pricing.
> We will run out of additional material to train on
This sounds a bit silly. More training will generally result in better modeling, even for a fixed amount of genuine original data. At current model sizes, it's essentially impossible to overfit to the training data so there's no reason why we should just "stop".
You'd be surprised how quickly improvement of autoregressive language models levels off with epoch count (though, admittedly, one epoch is a LOT). Diffusion language models otoh indeed keep profiting for much longer, fwiw.
"On the other hand, training on synthetic data has shown much promise in domains where model outputs are relatively easy to verify, such as mathematics, programming, and games (Yang et al., 2023; Liu et al., 2023; Haluptzok et al., 2023)."
With the caveat that translating this success outside of these domains is hit-or-miss:
"What is less clear is whether the usefulness of synthetic data will generalize to domains where output verification is more challenging, such as natural language."
The main bottleneck for this area of the woods will be (X := how many additional domains can be made easily verifiable). So long as (the rate of X) >> (training absorption rate), the road can be extended for a while longer.
How much of the current usage is productive work that's worth paying for vs personal usage / spam that would just drop off after usage charges come in? I imagine flooding youtube and instagram with slop videos would reduce if users had to pay fair prices to use the models.
The companies might also downgrade the quality of the models to make it more viable to provide as an ad supported service which would again reduce utilisation.
For any "click here and type into a box" job for which you'd hire a low-skilled worker and give them an SOP to follow, you can have an LLM-ish tool do it.
And probably for the slightly more skilled email jobs that have infiltrated nearly all companies too.
Is that productive work? Well if people are getting paid, often a multiple of minimum wage, then it's productive-seeming enough.
Exactly, the current spend on LLMs is based on extremely high expectations and the vendors operating at a loss. It’s very reasonable to assume that those expectations will not be met, and spending will slow down as well.
Nvidia’s valuation is based on the current trend continuing and even increasing, which I consider unlikely in the long term.
> Nvidia’s valuation is based on the current trend continuing
People said this back when Folding@Home was dominated by Team Green years ago. Then again when GPUs sold out for the cryptocurrency boom, and now again that Nvidia is addressing the LLM demand.
Nvidia's valuation is backstopped by the fact that Russia, Ukraine, China and the United States are all tripping over themselves for the chance to deploy it operationally. If the world goes to war (which is an unfortunate likelihood) then Nvidia will be the only trillion-dollar defense empire since the DoD's Last Supper.
China is restricting purchases of H200s. The strong likelihood is that they're doing this to promote their own domestic competitors. It may take a few years for those chips to catch up and enter full production, but it's hard to envision any "trillion dollar" Nvidia defense empire once that happens.
> short of major algorithmic breakthroughs I am not convinced the global demand for GPUs will drop any time soon
>> Or, you know, when LLMs don't pay off.
Heh, exactly the observation that a fanatic religious believer cannot possibly foresee. "We need more churches! More priests! Until a breakthrough in praying technique will be achieved I don't foresee less demand for religious devotion!" Nobody foresaw Nietzsche and the decline in blind faith.
But then again, like an atheist back in the day, the furious zealots would burn me at the stake if they could, for saying this. Sadly no longer possible so let them downvotes pour instead!
They aren’t yet because the big providers that paid for all of this GPU capacity aren’t profitable yet.
They continually leap frog each other and shift around customers which indicates that the current capacity is already higher than what is required for what people actually pay for.
Yeah but OpenAI is adding ads this year for the free versions, which I'm guessing is most of their users. They are probably hedging on taking a big slice of Google's advertising monopoly-pie (which is why Google is also now all-in on forcing Gemini opt-out on every product they own, they can see the writing on the wall).
Google, Amazon, and Microsoft do a lot of things that aren't profitable in themselves. There is no reason to believe a company will kill a product line just because it makes a loss. There are plenty of other reasons to keep it running.
Do you think it's odd you only listed companies with already existing revenue streams and not companies that started with and only have generative algos as their product?
Algorithmic breakthroughs (increases in efficiency) risk Jevons Paradox. More efficient processes make deploying them even more cost effective and increases demand.
NVIDIA stock tanked in 2025 when people learned that Google used TPUs to train Gemini, which everyone in the community knows since at least 2021. So I think it's very likely that NVIDIA stock could crash for non-rationale reasons
The market is full of people trying to anticipate how other people are going to react and exploit that by getting there first.
There's a layer aimed at forecasting what that layer is going to do as well.
This was also on top of claims (Jan 2025) that Deepseek showed that "we don't actually need as much GPU, thus NVidia is less needed"; at least it was my impression this was one of the (now silly-seeming) reasons NVDA dropped then.
> I don't know if that's non-rational, or if people can't be expected to read the second sentence of an announcement before panicking.
These days you have AI bots doing sentiment based training.
If you ask me... all these excesses are a clear sign for one thing, we need to drastically rein in the stonk markets. The markets should serve us, not the other way around.
Google did not use TPUs for literally every bit of compute that led to Gemini. GCP has millions of high end Nvidia GPUs and programming for them is an order of magnitude easier, even for googlers.
Any claim from google that all of Gemini (including previous experiments) was trained entirely by TPUs is lies. What they are truthfully saying is that the final training run was done on all TPUs. The market shouldn’t react heavily to this, but instead should react positively to the fact that google is now finally selling TPUs externally and their fab yields are better than expected.
I really don't understand the argument that nvidia GPUs only work for 1-3 years. I am currently using A100s and H100s every day. Those aren't exactly new anymore.
It’s not that they don’t work. It’s how businesses handle hardware.
I worked at a few data centers on and off in my career. I got lots of hardware for free or on the cheap simply because the hardware was considered “EOL” after about 3 years, often when support contracts with the vendor ends.
There are a few things to consider.
Hardware that ages produce more errors, and those errors cost, one way or another.
Rack space is limited. A perfectly fine machine that consumes 2x the power for half the output cost. It’s cheaper to upgrade a perfectly fine working system simply because it performs better per watt in the same space.
Lastly. There are tax implications in buying new hardware that can often favor replacement.
I agree that there is hyperbole thrown around a lot here and its possible to still use some hardware for a long time or to sell it and recover some cost but my experience in planning compute at large companies is that spending money on hardware and upgrading can often result in saving money long term.
Even assuming your compute demands stay fixed, its possible that a future generation of accelerator will be sufficiently more power/cooling efficient for your workload that it is a positive return on investment to upgrade, more so when you take into account you can start depreciating them again.
If your compute demands aren't fixed you have to work around limited floor space/electricity/cooling capacity/network capacity/backup generators/etc and so moving to the next generation is required to meet demand without extremely expensive (and often slow) infrastructure projects.
Sure, but I don't think most people here are objecting to the obvious "3 years is enough for enterprise GPUs to become totally obsolete for cutting-edge workloads" point. They're just objecting to the rather bizarre notion that the hardware itself might physically break in that timeframe. Now, it would be one thing if that notion was supported by actual reliability studies drawn from that same environment - like we see for the Backblaze HDD lifecycle analyses. But instead we're just getting these weird rumors.
I agree that is a strange notion that would require some evidence and I see it in some other threads but looking at the parent comments going up it seems people are discussing economic usefulness so that is what I'm responding to.
A toy example: NeoCloud Inc builds a new datacenter full of the new H800 GPUs. It rents out a rack of them for $10/minute while paying $6/minute for electricity, interest, loan repayment, rent and staff.
Two years later, H900 is released for a similar price but it performs twice as many TFlOps/Watt. Now any datacenter using H900 can offer the same performance as NeoCloud Inc at $5/month, taking all their customers.
It's because they run 24/7 in a challenging environment. They will start dying at some point and if you aren't replacing them you will have a big problem when they all die en masse at the same time.
These things are like cars, they don't last forever and break down with usage. Yes, they can last 7 years in your home computer when you run it 1% of the time. They won't last that long in a data center where they are running 90% of the time.
A makeshift cryptomining rig is absolutely a "challenging environment" and most GPUs by far that went through that are just fine. The idea that the hardware might just die after 3 years' usage is bonkers.
Crypto miners undervote for efficiency GPUs and in general crypto mining is extremely light weight on GPUs compared to AI training or inference at scale
With good enough cooling they can run indefinitely!!!!! The vast majority of failures are either at the beginning due to defects or at the end due to cooling! It’s like the idea that no moving parts (except the HVAC) is somehow unreliable is coming out of thin air!
There’s plenty on eBay? But at the end of your comment you say “a rate cheaper than new” so maybe you mean you’d love to buy a discounted one. But they do seem to be available used.
Do you know how support contract lengths are determined? Seems like a path to force hardware refreshes with boilerplate failure data carried over from who knows when.
The common factoid raised in financial reports is GPUs used in model training will lose thermal insulation due to their high utilization. The GPUs ostensibly fail. I have heard anecdotal reports of GPUs used for cryptocurrency mining having similar wear patterns.
I have not seen hard data, so this could be an oft-repeated, but false fact.
It's the opposite actually - most GPU used for mining are run at a consistent temp and load which is good for long term wear. Peaky loads where the GPU goes from cold to hot and back leads to more degradation because of changes in thermal expansion. This has been known for some time now.
That is commonly repeated idea, but it doesn't take into account countless token farms which are smaller than a datacenter. Basically anything from a single MB with 8 cards to a small shed with rigs, all of which tend to disregard common engineering practices and run hardware into a ground to maximize output until next police raid or difficulty bump. Plenty of photos in the internet of crappy rigs like that, and no one guarantees which GPU comes whom where.
Another commonly forgotten issue is that many electrical components are rated by hours of operation. And cheaper boards tend to have components with smaller tolerances. And that rated time is actually a graph, where hour decrease with higher temperature. There were instances of batches of cards failing due to failing MOSFETs for example.
While I'm sure there are small amateur setups done poorly that push cards to their limits this seems like a more rare and inefficient use. GPUS (even used) are expensive and running them at maximum would require large costs and time to be replacing them regularly. Not to mention the increased cost of cooling and power.
Not sure I understand the police raid mentality - why are the police raiding amateur crypto mining setups ?
I can totally see cards used by casual amateurs being very worn / used though - especially your example of single mobo miners who were likely also using the card for gaming and other tasks.
I would imagine that anyone purposely running hardware into the ground would be running cheaper / more efficient ASICS vs expensive Nvidia GPUs since they are much easier and cheaper to replace. I would still be surprised however if most were not proritising temps and cooling
Miners usually don't overclock though. If anything underclocking is the best way to improve your ROI because it significantly reduces the power consumption while retaining most of the hashrate.
Exactly - more specifically undervolting. You want the minimum volts going to the card with it still performing decently.
Even in amateur setups the amount of power used is a huge factor (because of the huge draw from the cards themselves and AC units to cool the room) so minimising heat is key.
From what I remember most cards (even CPUs as well) hit peak efficiency when undervolted and hitting somewhere around 70-80% max load (this also depends on cooling setup). First thing to wear out would probably be the fan / cooler itself (repasting occasionally would of course help with this as thermal paste dries out with both time and heat)
The only amatures I know doing this are trying to heat their garrage for free. so long as the heat gain is paid for they can afford to heat an otherwise unheated building.
> I have heard anecdotal reports of GPUs used for cryptocurrency mining having similar wear patterns.
If this was anywhere close to a common failure mode, I'm pretty sure we'd know that already given how crypto mining GPUs were usually ran to the max in makeshift settings with woefully inadequate cooling and environmental control. The overwhelming anecdotal evidence from people who have bought them is that even a "worn" crypto GPU is absolutely fine.
I can't confirm that fact - but it's important to acknowledge that consumer usage is very different from the high continuous utilization in mining and training. It is credulous that the wear on cards under such extreme usage is as high as reported considering that consumers may use their cards at peak 5% of waking hours and the wear drop off is only about 3x if it is used near 100% - that is a believable scale for endurance loss.
1-3 is too short but they aren’t making new A100s, theres 8 in a server and when one goes bad what do you do? you wont be able to renew a support contract. if you wanna diy you eventually you have to start consolidating pick and pulls. maybe the vendors will buy them back from people who want to upgrade and resell them. this is the issue we are seeing with A100s and we are trying to see what our vendor will offer for support.
Margins are typically not so razor thin that you cannot operate with technology from one generation ago. 15 vs 17 mpg is going to add up over time, but for a taxi company it's probably not a lethal situation to be in.
At least with crypto mining this was the case. Hardware from 6 months ago is useless ewaste because the new generation is more power efficient. All depends on how expensive the hardware is vs the cost of power.
And yet they aren't running planes and engines all from 2023 or beyond: See the MD-11 that crashed in Louisville: Nobody has made a new MD-11 in over 20 years. Planes move to less competitive routes, change carriers, and eventually might even stop carrying people and switch to cargo, but the plane itself doesn't get to have zero value when the new one comes out. An airline will want to replace their planes, but a new plane isn't fully amortized in a year or three: It still has value for quite a while
If a taxi company did that every year, they'd be losing a lot of money. Of course new cars and cards are cheaper to operate than old ones, but is that difference enough to offset buying a new one every one to three years?
>If a taxi company did that every year, they'd be losing a lot of money. Of course new cars and cards are cheaper to operate than old ones, but is that difference enough to offset buying a new one every one to three years?
That's where the analogy breaks. There are massive efficiency gains from new process nodes, which new GPUs use. Efficiency improvements for cars are glacial, aside from "breakthroughs" like hybrid/EV cars.
>offset buying a new one every one to three years?
Isn't that precisely how leasing works? Also, don't companies prefer not to own hardware for tax purposes? I've worked for several places where they leased compute equipment with upgrades coming at the end of each lease.
Who wants to buy GPUs that were redlined for three years in a data center? Maybe there's a market for those, but most people already seem wary of lightly used GPUs from other consumers, let alone GPUs that were burning in a crypto farm or AI data center for years.
who cares? that's the beauty of the lease. once it's over, the old and busted gets replaced with new and shiny. what the leasing company does is up to them. it becomes one of those YP not an MP situations with deprecated equipment.
The leasing company cares - the lease terms depend on the answer. That is why I can lease a car for 3 years for the same payment as a 6 year loan (more or less) - the lease company expects someone will want it. If there is no market for it after they will still lease it but the cost goes up
Depends on the price, of course. I'm wary of paying 50% of new for something run hard 3 years. Seems an NVIDIA H100 is going for $20k+ on EBay. I'm not taking that risk.
That works either because someone wants to buy old hardware for the manufacturer/lessor, or because the hardware is EOL in 3 years but it's easier to let the lessor deal with recyling / valuable parts recovery.
You can sell the old, less efficient GPUs to folks who will be running them with markedly lower duty cycles (so, less emphasis on direct operational costs), e.g. for on-prem inference or even just typical workstation/consumer use. It ends up being a win-win trade.
Building a new data center and getting power takes years to double your capacity. Swapping out out a rack that is twice as fast takes very little time in comparison.
Depends at the rate of growth of the hardware. If your data center is full and fully booked, and hardware is doubling in speed every year it's cheaper to switch it out every couple of years.
Both companies bought a set of taxis in the past. Presumably at the same time if we want this comparison to be easy to understand.
If company A still has debt from that, company B has that much debt plus more debt from buying a new set of taxis.
Refreshing your equipment more often means that you're spending more per year on equipment. If you do it too often, then even if the new equipment is better you lose money overall.
If company B wants to undercut company A, their advantage from better equipment has to overcome the cost of switching.
> They both refresh their equipment at the same rate.
I wish you'd said that upfront. Especially because the comment you replied to was talking about replacing at different rates.
So your version, if company A and B are refreshing at the same rate, then that means six months before B's refresh company A had the newer taxis. You implied they were charging similar amounts at that point, so company A was making bigger profits, and had been making bigger profits for a significant time. So when company B is able to cut prices 5%, company A can survive just fine. They don't need to rush into a premature upgrade that costs a ton of money, they can upgrade on their normal schedule.
TL;DR: six months ago company B was "no longer competitive" and they survived. The companies are taking turns having the best tech. It's fine.
(1) We simply don't know what the useful life is going to be because of how new the advancements of AI focused GPUs used for training and inference.
(2) Warranties and service. Most enterprise hardware has service contracts tied to purchases. I haven't seen anything publicly disclosed about what these contracts look like, but the speculation is that they are much more aggressive (3 years or less) than typical enterprise hardware contracts (Dell, HP, etc.). If it gets past those contracts the extended support contracts can typically get really pricey.
(3) Power efficiency. If new GPUs are more power efficient this could be huge savings on energy that could necessitate upgrades.
Nvidia is moving to a 1 year release life cycle for data center, and in Jensen's words once a new gen is released you lose money for being on the older hardware. It makes no longer financially sense to run it.
Companies can’t buy new Nvidia GPUs because their older Nvidia GPUs are obsolete. However, the old GPUs are only obsolete if companies buy the new Nvidia GPUs.
based on my napkin math, an H200 needs to run for 4 years straight at maximum power (10.2 kW) to consume its own price of $35k worth of energy (based on 10 cents per kWh)
If power is the bottleneck, it may make business sense to rotate to a GPU that better utilizes the same power if the newer generation gives you a significant advantage.
I think the story is less about the GPUs themselves, and more about the interconnects for building massive GPU clusters. Nvidia just announced a massive switch for linking GPUs inside a rack. So the next couple of generations of GPU clusters will be capable of things that were previously impossible or impractical.
This doesn't mean much for inference, but for training, it is going to be huge.
From an accounting standpoint, it probably makes sense to have their depreciation be 3 years. But yeah, my understanding is that either they have long service lives, or the customers sell them back to the distributor so they can buy the latest and greatest. (The distributor would sell them as refurbished)
> My 30k ft view is that the stock will inevitably slide as AI datacenter spending goes down.
Their stock trajectory started with one boom (cryptocurrencies) and then seamlessly progressed to another (AI). You're basically looking at a decade of "number goes up". So yeah, it will probably come down eventually (or the inflation will catch up), but it's a poor argument for betting against them right now.
Meanwhile, the investors who were "wrong" anticipating a cryptocurrency revolution and who bought NVDA have not much to complain about today.
Personally I wonder even if the LLM hype dies down we'll get a new boom in terms of AI for robotics and the "digital twin" technology Nvidia has been hyping up to train them. That's going to need GPUs for both the ML component as well as 3D visualization. Robots haven't yet had their SD 1.1 or GPT-3 moment and we're still in the early days of Pythia, GPT-J, AI Dungeon, etc. in LLM speak.
That's going to tank the stock price though as that's a much smaller market than AI, though it's not going to kill the company. Hence why I'm talking about something like robotics which has a lot of opportunity to grow and make use of all those chips and datacenters they're building.
Now there's one thing with AR/VR that might need this kind of infrastructure though and that's basically AI driven games or Holodeck like stuff. Basically have the frames be generated rather than modeled and rendered traditionally.
Nvidia's not your average bear, they can walk and chew bubblegum at the same time. CUDA was developed off money made from GeForce products, and now RTX products are being subsidized by the money made on CUDA compute. If an enormous demand for efficient raster compute arises, Nvidia doesn't have to pivot much further than increasing their GPU supply.
Robotics is a bit of a "flying car" application that gets people to think outside the box. Right now, both Russia and Ukraine are using Nvidia hardware in drones and cruise missiles and C2 as well. The United States will join them if a peer conflict breaks out, and if push comes to shove then Europe will too. This is the kind of volatility that crazy people love to go long on.
That's the rub - it's clearly overvalued and will readjust... the question is when. If you can figure out when precisely then you've won the lottery, for everyone else it's a game of chicken where for "a while" money that you put into it will have a good return. Everyone would love if that lasted forever so there is a strong momentum preventing that market correction.
It was overvalued when crypto was happening too, but another boom took its place. Of course, lightening rarely strikes twice and all that, but it proves overvalued doesn’t mean the price is guaranteed to go down it seems. Predicting the future is hard.
if there was anything i was going to bet against between 2019 and now, it was nvidia... and wow it feels wild how much in the opposite direction it went.
I do wonder what people would think the reasoning would be for them to increase in value this much back then, prolly would just assume crypto related still.
Crypto & AI can both be linked to part of a broader trend though, that we need processors capable of running compute on massive sets of data quickly. I don't think that will ever go down, whether some new tech emerges or we just continue shoveling LLMs into everything. Imagine the compute needed to allow every person on earth to run a couple million tokens through a model like Anthropic Opus every day.
Agree on looking at the company-behind-the-numbers. Though presumably you're aware of the Efficient Market Hypothesis. Shouldn't "slowed down datacenter growth" be baked into the stock price already?
If I'm understanding your prediction correctly, you're asserting that the market thinks datacenter spending will continue at this pace indefinitely, and you yourself uniquely believe that to be not true. Right? I wonder why the market (including hedge fund analysis _much_ more sophisticated than us) should be so misinformed.
Presumably the market knows that the whole earth can't be covered in datacenters, and thus has baked that into the price, no?
I saw a $100 bill on the ground. I nearly picked it up before I stopped myself. I realised that if it was a genuine currency note, the Efficient Market would have picked it up already.
> This article goes more into the technical analysis of the stock rather than the underlying business fundamentals that would lead to a stock dump.
My 30k ft view is that the stock will inevitably slide as AI
Actually "technical analysis" (TA) has a very specific meaning in trading: TA is using past prices, volume of trading and price movements to, hopefully, give probabilities about future price moves.
But TFA doesn't do that at all: it goes in detail into one pricing model formula/method for options pricing. In the typical options pricing model all you're using is current price (of the underlying, say NVDA), strike price (of the option), expiration date, current interest rate and IV (implied volatility: influenced by recent price movements but independently of any technical analysis).
Be it Black-Scholes-Merton (european-style options), Bjerksund-Stensland (american-style options), binomial as in TFA, or other open options pricing model: none of these use technical analysis.
Here's an example (for european-style options) where one can see the parameters:
You can literally compute entire options chains with these parameters.
Now it's known for a fact that many professional traders firms have their own options pricing method and shall arb when they think they find incorrectly priced options. I don't know if some use actual so forms of TA that they then mix with options pricing model or not.
> My 30k ft view is that the stock will inevitably slide as AI datacenter spending goes down.
No matter if you're right or not, I'd argue you're doing what's called fundamental analysis (but I may be wrong).
P.S: I'm not debatting the merits of TA and whether it's reading into tea leaves or not. What I'm saying is that options pricing using the binomial method cannot be called "technical analysis" for TA is something else.
I'll also point out there were insane takes a few years ago before nVidia's run up based on similar technical analysis and very limited scope fundamental analysis.
Technical analysis fails completely when there's an underlying shift that moves the line. You can't look at the past and say "nvidia is clearly overvalued at $10 because it was $3 for years earlier" when they suddenly and repeatedly 10x earnings over many quarters.
I couldn't get through to the idiots on reddit.com/r/stocks about this when there was non-stop negativity on nvidia based on technical analysis and very narrow scoped fundamental analysis. They showed a 12x gain in quarterly earnings at the time but the PE (which looks on past quarters only) was 260x due to this sudden change in earnings and pretty much all of reddit couldn't get past this.
I did well on this yet there were endless posts of "Nvidia is the easiest short ever" when it was ~$40 pre-split.
The large api/token providers, and large consumers are all investing in their own hardware. So, they are in an interesting position where the market is growing, and NVIDIA is taking the lion's share of enterprise, but is shrinking at the hyperscaler side (google is a good example as they shift more and more compute to TPU). So, they have a shrinking market share, but its not super visible.
> The large api/token providers, and large consumers are all investing in their own hardware.
Which is absolutely the right move when your latest datacenter's power bill is literally measured in gigawatts. Power-efficient training/inference hardware simply does not look like a GPU at a hardware design level (though admittedly, it looks even less like an ordinary CPU), it's more like something that should run dog slow wrt. max design frequency but then more than make up for that with extreme throughput per watt/low energy expense per elementary operation.
The whole sector of "neuromorphic" hardware design has long shown the broad feasibility of this (and TPUs are already a partial step in that direction), so it looks like this should be an obvious response to current trends in power and cooling demands for big AI workloads.
I no AI fanboy at all. I think it there won’t be AGI anytime soon.
However, it’s beyond my comprehension how anyone would think that we will see a decline in demand growth for compute.
AI will conquer the world like software or the smartphone did. It’ll get implemented everywhere, more people will use it. We’re super early in the penetration so far.
At this point computation is in essence commodity. And commodities have demand cycles. If other economic factors slowdown or companies go out of business they stop using compute or start less new products that use compute. Thus it is entirely realistic to me that demand for compute might go down. Or that we are just now over provisioning compute in short or medium term.
I wonder, is the quality of AI answers going up over time or not? Last weekend I spent a lot of time with Preplexity trying to understand why my SeqTrack device didn't do what I wanted it to do and seems Perplexity had a wrong idea of how the buttons on the device are laid out, so it gave me wrong or confusing answers. I spent literally hours trying to feed it different prompts to get an answer that would solve my problem.
If it had given me the right easy to understand answer right away I would have spent 2 minutes of both MY time and ITS time. My point is if AI will improve we will need less of it, to get our questions answered. Or, perhaps AI usage goes up if it improves its answers?
Always worth trying a different model, especially if you’re using a free one. I wouldn’t take one data point to seriously either.
The data is very strongly showing the quality of AI answers is rapidly improving. If you want a good example, check out the sixty symbols video by Brady Haran, where they revisited getting AI to answer a quantum physics exam after trying the same thing 3 years ago. The improvement is IMMENSE and unavoidable.
The problem is it's inability to say "I don't know". As soon as you reach the limits of the models knowledge it will readily start fabricating answers.
Both true. Perplexity knows a lot about SeqTrack, I assume it has read the UserGuide. But some things it gets wrong, seems especially things it should understand by looking at the pictures.
I'm just wondering if there's a clear path for it to improve and on what time-table. The fact that it does not tell you when it is "unsure" of course makes things worse for users. (It is never unsure).
With vision models (SOTA models like Gemini and ChatGPT can do this), you can take a picture/screenshot of the button layout, upload it, and have it work from that. Feeding it current documentation (eg a pdf of a user manual) helps too.
Referencing outdated documentation or straight up hallucinating answers is still an issue. It is getting better with each model release though
More so I meant to think of oil, copper and now silver. All follow demand for the price. All have had varying prices at different times. Compute should not really be that different.
But yes. Cisco's value dropped when there was not same amount to spend on networking gear. Nvidia's value will drop as there is not same amount of spend on their gear.
Other impacted players in actual economic downturn could be Amazon with AWS, MS with Azure. And even more so those now betting on AI computing. At least general purpose computing can run web servers.
Even suggesting that computers will replace human brains brings up a moral and ethical question. If the computer is just as smart as a person, then we need to potentially consider that the computer has rights.
As far as AI conquering the world. It needs a "killer app". I don't think we'll really see that until AR glasses that happen to include AI. If it can have context about your day, take action on your behalf, and have the same battery life as a smartphone...
I don’t see this as fanaticism at all. No one could predict a billion people mindlessly scrolling tiktok in 2007. This is going to happen again, only 10x. Faster and more addictive, with content generated on the fly to be so addictive, you won’t be able to look away.
What if its penetration ends up being on the same level as modern crypto? Average person doesn't seem to particularly care about meme coins or bitcoin - it is not being actively used in day to day setting, there's no signs of this status improving.
Doesn't mean that crypto is not being used, of course. Plenty of people do use things like USDT, gamble on bitcoin or try to scam people with new meme coins, but this is far from what crypto enthusiasts and NFT moguls promised us in their feverish posts back in the middle of 2010s.
So imagine that AI is here to stay, but the absolutely unhinged hype train will slow down and we will settle in some kind of equilibrium of practical use.
I have still been unable to see how folks connect AI to Crypto. Crypto never connected with real use cases. There are some edge cases and people do use it but there is not a core use.
AI is different and businesses are already using it a lot. Of course there is hype, it’s not doing all the things the talking heads said but it does not mean immense value is not being generated.
It's an analogy, it doesn't have to map 1:1 to AI. The point is that current situation around AI looks kind of similar to the situation and level of hype around Crypto when it was still growing: all the "ledger" startups, promises of decentralization, NFTs in video games and so on. We are somewhere around that point when it comes to AI.
No it’s an absolutely ridiculous comparison that people continue to make even though AI has well past the usefulness of crypto and at an alarming rate of speed. AI has unlocked so many projects my team would never have tackled before.
Anecdotally, many non-technical users or "regular joes" as it were that I know who were very enthusiastic about AI a year ago are now disengaging. With the rate really picking up the last couple of months.
Their usage has declined primarily with OpenAI and Gemini tools, no one has mentioned Anthropic based models but I don't think normies know they exist honestly.
The disengagement seems to be that with enough time and real world application, the shortcomings have become more noticable and the patience they once had for incorrect or unreliable output has effectively evaporated. In cases, to the point where its starting to outweigh any gains they get.
Not all of the normies I know to be fair, but a surprising amount given the strange period of quiet inbetween "This is amazing!" and "eh, its not as good as I thought it was at first."
This seems to take for granted that China and their foundries and engineering teams will never catch up. This seems foolish. I'm working under the assumption that sometime in the next ten years some Chinese company will have a breakthrough and either meet Nvidia's level or leapfrog them. Then the market will flood with great, cheap chips.
I think the way to think about the AI bubble is that we're somewhere in 97-99 right now, heading toward the dotcom crash. The dotcom crash didn't kill the web, it kept growing in the decades that followed, influencing society more and more. But the era where tons of investments were uncritically thrown at anything to do with the web ended with a bang.
When the AI bubble bursts, it won't stop the development of AI as a technology. Or its impact on society. But it will end the era of uncritically throwing investments at anyone that works "AI" into their pitch deck. And so too will it end the era of Nvidia selling pickaxes to the miners and being able to reach soaring heights of profitability born on wings of pretty much all investment capital in the world at the moment.
Bubble or not it’s simply strange to me that people confidently put a timeline on it. To name the phases of the bubble and calling when they will collapse just seems counter intuitive to what a bubble is. Brad Gerstner was the first “influencer” I heard making these claims of a bubble time line. It just seems downright absurd.
> The problem I have with this argument is that it's simply unsustainable to be spending that much every 2-3 years
Isn’t this entirely dependent on the economic value of the AI workloads? It all depends on whether AI work is more valuable than that cost. I can easily see arguments why it won’t be that valuable, but if it is, then that cost will be sustainable.
100% this. all of this spending is predicated on a stratospheric ROI on AI investments at the proposed investment levels. If that doesn't pan out, we'll see a lot of people left holding the cards including chip fabs, designers like Nvidia, and of course anyone that ponied up for that much compute.
I’m sad about Grok going to them, because the market needs the competition. But ASIC inference seems to require a simpler design than training does, so it’s easier for multiple companies to enter. It seems inevitable that competition emerges. And eg a Chinese company will not be sold to Nvidia.
What’s wrong with this logic? Any insiders willing to weigh in?
I'm not an insider, but ASICs come with their own suite of issues and might be obsolete if a different architecture becomes popular. They'll have a much shorter lifespan than Nvidia hardware in all likelihood, and will probably struggle to find fab capacity that puts them on equal footing in performance. For example, look at the GPU shortage that hit crypto despite hundreds of ASIC designs existing.
The industry badly needs to cooperate on an actual competitor to CUDA, and unfortunately they're more hostile to each other today than they were 10 years ago.
You can build ASICs to be a lot more energy efficient than current GPUs, especially if your power budget is heavily bound by raw compute as opposed to data movement bandwidth. The tradeoff is much higher latency for any given compute throughput, but for workloads such as training or even some kinds of "deep thinking inference" you don't care much about that.
Even though I like CUDA, I think the point is when do compute centers reach the point that they can run their workloads with other vendors, or custom accelerators.
Fundamental analysis is great! But I have trouble answering concrete questions of probability with it.
How do you use fundamental analysis to assign a probability to Nvidia closing under $100 this year, and what probability do you assign to that outcome?
I'd love to hear your reasoning around specifics to get better at it.
GP was presenting fundamental analysis as an alternative to the article's method for answering the question, but then never answered the question.
This is a confusion I have around fundamental analysis. Some people appear to do it very well (Buffett?) but most of its proponents only use it to ramble about possibilities without making any forecasts speciic enough to be verifiable.
I think the idea of fundamental analysis that you focus on return on equity and see if that valuation is appreciably more than the current price (as opposed to assigning a probability)
Well, not to be too egregiously reductive… but when the M2 money supply spiked in the 2020 to 2022 timespan, a lot of new money entered the middle class. That money was then funneled back into the hands of the rich through “inflation”. That left the rich with a lot of spare capital to invest in finding the next boom. Then AI came along.
Once the money dries up, a new bubble will be invented to capture the middle class income, like NFTs and crypto before that, and commissionless stocks, etc etc
It’s not all pump-and-dump. Again, this is a pretty reductive take on market forces. I’m just saying I don’t think it’s quite as unsustainable as you might think.
Add in the fact companies seriously invested in AI (and like workloads typically reliant on GPUs) are also investing more into bespoke accelerators, and the math for nVidia looks particularly grim. Google’s TPUs set them apart from the competition, as does Apple’s NPU; it’s reasonable to assume firms like Anthropic or OpenAI are also investigating or investing into similar hardware accelerators. After all, it’s easier to lock-in customers if your models cannot run on “standard” kit like GPUs and servers, even if it’s also incredibly wasteful.
The math looks bad regardless of which way the industry goes, too. A successful AI industry has a vested interest in bespoke hardware to build better models, faster. A stalled AI industry would want custom hardware to bring down costs and reduce external reliance on competitors. A failed AI industry needs no GPUs at all, and an inference-focused industry definitely wants custom hardware, not general-purpose GPUs.
So nVidia is capitalizing on a bubble, which you could argue is the right move under such market conditions. The problem is that they’re also alienating their core customer base (smaller datacenters, HPC, gaming market) in the present, which will impact future growth. Their GPUs are scarce and overpriced relative to performance, which itself has remained a near-direct function of increased power input rather than efficiency or meaningful improvements. Their software solutions - DLSS frame-generation, ray reconstruction, etc - are locked to their cards, but competitors can and have made equivalent-performing solutions of their own with varying degrees of success. This means it’s no longer necessary to have an nVidia GPU to, say, crunch scientific workloads or render UHD game experiences, which in turn means we can utilize cheaper hardware for similar results. Rubbing salt in the wound, they’re making cards even more expensive by unbundling memory and clamping down on AIB designs. Their competition - Intel and AMD primarily - are happily enjoying the scarcity of nVidia cards and reaping the fiscal rewards, however meager they are compared to AI at present. AMD in particular is sitting pretty, powering four of the five present-gen consoles, the Steam Deck (and copycats), and the Steam Machine, not to mention outfits like Framework; if you need a smol but capable boxen on the (relative) cheap, what used to be nVidia + ARM is now just AMD (and soon, Intel, if they can stick the landing with their new iGPUs).
The business fundamentals paint a picture of cannibalizing one’s evergreen customers in favor of repeated fads (crypto and AI), and years of doing so has left those customer markets devastated and bitter at nVidia’s antics. Short of a new series of GPUs with immense performance gains at lower price and power points with availability to meet demand, my personal read is that this is merely Jenson Huang’s explosive send-off before handing the bag over to some new sap (and shareholders) once the party inevitably ends, one way or another.
> My 30k ft view is that the stock will inevitably slide as AI datacenter spending goes down. Right now Nvidia is flying high because datacenters are breaking ground everywhere but eventually that will come to an end as the supply of compute goes up.
Exactly, it is currently priced as though infinite GPUs are required indefinitely. Eventually most of the data centres and the gamers will have their GPUs, and demand will certainly decrease.
Before that, though, the data centres will likely fail to be built in full. Investors will eventually figure out that LLMs are still not profitable, no matter how many data centres you produce. People are interested in the product derivatives at a lower price than it costs to run them. The math ain't mathin'.
The longer it takes to get them all built, the more exposed they all are. Even if it turns out to be profitable, taking three years to build a data centre rather than one year is significant, as profit for these high-tech components falls off over time. And how many AI data centres do we really need?
I would go further and say that these long and complex supply chains are quite brittle. In 2019, a 13 minute power cut caused a loss of 10 weeks of memory stock [1]. Normally, the shops and warehouses act as a capacitor and can absorb small supply chain ripples. But now these components are being piped straight to data centres, they are far more sensitive to blips. What about a small issue in the silicon that means you damage large amounts of your stock trying to run it at full power through something like electromigration [2]. Or a random war...?
> The counterargument to this is that the "economic lifespan" of an Nvidia GPU is 1-3 years depending on where it's used so there's a case to be made that Nvidia will always have customers coming back for the latest and greatest chips. The problem I have with this argument is that it's simply unsustainable to be spending that much every 2-3 years and we're already seeing this as Google and others are extending their depreciation of GPU's to something like 5-7 years.
Yep. Nothing about this adds up. Existing data centres with proper infrastructure are being forced to extend use for previously uneconomical hardware because new data centres currently building infrastructure have run the price up so high. If Google really thought this new hardware was going to be so profitable, they would have bought it all up.
I wouldn't call Omarchy "mainstream". Yes it's very popular among developers but that's about it and under the hood it uses some pretty non-mainstream components like Hyplrand WM.
I would argue the OS closest to "mainstream Linux" is Ubuntu or Fedora with Gnome DE. Gnome has many many faults but it's probably the closest DE you're going to get to what Windows and MacOS have.
I'll give one of the more mainstream ones a try when I have a free afternoon, frustrating thing was it wasn't underpowered at all this was with a RTX3090 so very concerning investing in that, perhaps wrongly assumed Wayland etc would have been a similar feel to Mac Quartz Composer fluidity by now.
Most of the announcements I hear about Copilot, it's always how they've integrated it into some other piece of software or cut a deal with yet another vendor to add it to that vendors product offering. On the surface there's nothing wrong with doing that but that just seems to be the ONLY thing Microsoft is focused on.
Worse yet, most of these integrations seem like a exercise in ticking boxes rather than actually thinking through how integrating Copilot into a product will actually improve user experience. A great example was someone mentioned that Copilot was now integrated into the terminal app but beyond an icon + a chat window, there is zero integration.
Overall, MS just reeks of an organization that is cares more about numbers on a dashboard and pretty reports than they are on what users are actually experiencing.