OpenAI and Anthropic have the freedom to do absolutely insane things like negligently hack other companies. It would be stock price suicide if anything even remotely happened with Google.

The answer is far more benign.

They getting a higher ROI renting their TPUs to Anthropic et al instead of performing training and serving their own models. Google cloud has insane backlog, and has rapidly expanded to satisfy it. While those DCs get built, they’re cannibalizing their own products for it.

This makes sense because (1) they are investors in Anthropic, so they still win and (2) they can always catch up on model training later when the profit opportunity shifts, or abandon it if there is no way to recapture that value.

There is no other way to describe the Gemini 3.5 pro delay than as a complete and unmitigated disaster.

It's a huge company.

It's unlikely there is any one person to blame (and entirely possible he has none of it). But things need to change.

> people who truly believe in the economically transformative power of AI

People who truly care about becoming rich*

DeepMind made enormous transformative discoveries, for instance in the world of protein folding. But that will just save human lives, not let CEOs fire their people to grab a larger piece of cake for themselves.

I think something that doesn't get talked about a lot is how bad most large tech companies are at creating new products, in general. Like, if you look at most big tech companies, they have their core offering that got them to be really large and rich, and a few other products that are somewhat successful, and then a really long tail of markets they try to enter and failed at, or projects that were modestly successful but got killed because they weren't game changers (RIP Google Reader). Most of the time when a large company does something new that succeeds, it's via an acquisition of a smaller company (ie, Google with Android or Meta with Instagram and Whatsapp)

It's funny, because I think the company that's going to be best positioned coming out of this bubble is in fact google, because they have the expertise and the capital. But I honestly can't tell you right now what their AI product even is -- I've seen so many things go into the graveyard a few months after its launched that I'm utterly confused what their offering even is at this point.

From what I've heard, DeepMind's pay structure is not tied to Google compensation bands. Which means that top performers and key hires can be and are compensated extremely generously if the need calls for it. Another Alphabet subsidiary, Waymo, was well-known for doing this back in the day.

But yes, even large RSU grants cannot replicate the upside of joining a company that increases in valuation 1000X, but to be fair, that is not a position the vast majority of OpenAI or Anthropic employees find themselves in either.

That was true early but OpenAI/Anthropic have done a ton of hiring over the past couple years at already-huge valuations, as much on the strength of big current base salary + equity, not just future increase speculation.

DeepMind is just a part of Google which does research stuff, and Google is a massive company that does loads of things but mostly sells ads.

OpenAI and Anthropic are in a completely different game, their primary business model is making a product out of advancements that come out of DeepMind and co.

Google was offering researchers the Bell Labs/Xerox PARC model of comfortable budgets and pay but capped upside. And just like with Bell Labs and Xerox PARC the inventions at Google got productized elsewhere by people chasing the uncapped upside. Google would have been happy to keep LLMs in the research lab forever and never productize any of it. ChatGPT forced their hand.

It's not a ML talent problem. You don't need to be a genius deep learning researcher to think "Hey, maybe if we massively throttle and degrade the quality of our model while still charging the same price, that might drive people away" (as happened with Gemini 2.5 Pro, the one model where Google really was SOTA). Google's likely been providing insufficient training compute to DeepMind the same way they've been nickel-and-diming their customers, funneling it all to Search instead because that's where the money comes from.

This makes a lot of sense to me. Probably doesn't explain everything but I could see it explaining a lot. Could also explain openai and anthropic delaying IPO (among other factors)

Is this not an apples-and-pears comparison? DeepMind is a research lab, not an LLM-pilled money furnace.

One theory I've been entertaining is that whenever GPT-3.5 came out a lot of people were talking about the "bitter lesson" and how scale was all we really needed to get to AGI. No need for any fancy tricks, just release a larger model trained on more data, by the time we released a hypothetical "GPT-5 sized" model we'd have AGI.

Anyway, the actual theory is that Google and Meta have fallen behind because they've been playing by this playbook of focusing on scale and training data, whereas OpenAI and Anthropic have done so well because they are likely doing much more interesting things to improve their models over time. It makes sense when you realize that one of Google's key strengths, besides talent, is that they have an incredible amount of data they can use for training due to being both the world's leading search engine as well as having all that video data from YouTube. Scaling the training data makes more sense to them than it does to Anthropic and OpenAI, who are both relatively data-disadvantaged.

You can kind of see this when you look at the Gemini 3 scorecard when it came out (https://blog.google/products-and-platforms/products/gemini/g...) and notice that while it wasn't as good as Claude And GPT at coding, it scored higher on a bunch of other non-coding benchmarks, and I think the reason why is simply because of Google's data advantage.

If true, I feel even more vindicated for believing that the "scale is all we need" narrative was bullshit.

We'll see if anyone gets to cash in on those. All you need is one down round and that gets wiped out. Or if the IPO gets delayed and disappoints then the stock can drop well before the lockouts expire. OpenAI and Anthropic are essentially offering Monopoly money in the hopes that one day you can exchange it for real money.

I think it's just more about market incentive. At it's core, LLMs are bad for google's previous business model, which was to send you to as many sites 'good enough' for what you were looking for and plant Ad land mines along the way, in the search results and in the websites themselves.

The new paradigm is you ask the llm a question, get the answer and cutout the middle man. (yes the answer may or may not be as good as the old google result, but for the sake of the argument lets say it is), Google was in danger of simply getting their arm cut off. so they focused on scaling so they could add LLMs to the search, which they largely have. You can't offer an opus like model on something as big as search (and which is offered for 'free'), so they focused on that model, and the infrastructure to run it, because they cannot afford to lose search.

Meanwhile, they know the power of frontier models, they are working to have the infrastructure to be a huge player in them and I'm sure they will have a frontier capable model, eventually. They are playing a longer game, because they can, and I think it's going to work out very well for them.

I've always thought this as well.