I'm assuming Google's play here is to bleed its competitors of money and raise prices when they're gone. Building top-tier models is extremely expensive and will probably remain so.

Even companies that do it "on the cheap," like DeepSeek, pay tens of millions to train a single model, and total expenditures for infrastructure and salaries are estimated to surpass $1 billion. This market has an extremely high cost of entry.

So, I guess Google is applying the usual strategy here: undercut competition until it implodes and buy up any promising competitors that arise in the future. Given the current lack of market regulation in the US, this might work.

Yeah, they just have to make it through the hype and innovation cycle.

Too bad they sold Boston Dynamics :)

I feel like they’re trying to increase switching costs. eg was huge reluctance to adopt MCP and each had their own tool framework, until it seemed too big to ignore and everyone was just building MCP tools not OpenAI SDK tools.

You don't have loyalty, but one day there will be no one else to switch to. So, if you're a loyal user or not is a moot point.

History shows it's a self-defeating victory. If one provider were to "win" and stop innovating, they'll become ripe for disruption by the likes of Deepseek, and the second someone like that has a better model, I'll switch.

> If one provider were to "win" and stop innovating, they'll become ripe for disruption by the likes of Deepseek

Yes but that can take decades, till that time Google can keep making money with sub standard products and stop innovating.

Nothing lasts forever, not even empires. This doesn't mean that tech monopoly is any better than any other monopoly. They're all detrimental to society.

Eh, and if you're in the US the 'big guys' will have their favorite paid off politician put in a law that use of Chinese models is illegal or whatever.

Rent seeking behavior is always the end game.

The same was true for Web browsers in 2002, yet MS controlled 95% of the access to the web thanks to that bundling and no other "good enough" competitors until Firefox came along a few years later and took 30% from them giving Google an in to take the whole game with Chrome a few years later.

The strategy worked, Netscape is no more. Eventually Google did the same to Microsoft though. I wonder if any lessons can be taken from the browser wars to how things will play out with AI models.

Remember Google tried to play this trick with ChromeOS?

There is a network effect: more user interaction = more training data. I don't know how important it is, though.

Yep, this is why android phones are now pointing out their gemini features every moment they can. They want to turn their spying device into an AI spying device.