Google's AI division is hemorrhaging talent while the cloud business prints money. The people who built the frontier research are leaving. The people who sell enterprise contracts are staying. This tells you everything about what Google values and nothing you didn't already suspect.
Frontier AI means the cutting-edge stuff. The research that doesn't have a customer yet. The models that might change everything or might be a $300 million electricity bill with no product attached. Google apparently decided that's adorable but Amazon Web Services exists and someone needs to pay the light bill.
So the researchers walk. They go to OpenAI or Anthropic or some startup that lets them pretend compute costs aren't real. Google waves goodbye and promotes another VP of Cloud Sales who's never written a line of code but closes deals with Walmart.
The cloud boom is real. Companies pay actual money for boring infrastructure. They do not pay money for experimental models that might achieve AGI or might just be really expensive autocomplete. Google looked at its options and chose the thing with revenue today over the thing with a Wikipedia page in 2035.
This is called strategy. This is also called giving up.
The retail traders who bought Google stock because they read a blog post about Gemini are thrilled. The line goes up when you sell services instead of burning cash on research nobody understands. They don't know what a transformer is. They don't care that the people who invented it work somewhere else now. They see cloud revenue growth and they do what they always do, which is buy high and act like they meant to.
Google is now an enterprise software company that used to do interesting things. The empire expands. The builders leave. The quarterly earnings call stays on schedule.
Photo by Mitchell Luo on Unsplash
Leave a Comment