GOOGLE’S MOST BRILLIANT MINDS ARE RUNNING AWAY AND TAKING THE FUTURE WITH THEM
Imagine you work at the greatest tech company on the planet. You have access to infinite computing power, a salary that would make a hedge fund manager blush, and colleagues who have literally won Nobel Prizes. And then you decide to quit. Not to retire to a beach somewhere, but to go work for the startup down the street that is actively trying to put your current employer out of business. This is what is happening at Google DeepMind right now, and it is happening at a pace that should genuinely alarm anyone who owns Alphabet stock.
In the last few weeks, Google has lost John Jumper — a man who shared the 2024 Nobel Prize in Chemistry for AlphaFold, one of the most important scientific breakthroughs of the century. He is heading to Anthropic. Before that, Noam Shazeer walked out the door and straight into OpenAI. Shazeer is not just any researcher. He is one of the co-authors of “Attention Is All You Need,” the 2017 paper that gave the world the transformer architecture that every single modern AI system is built on. You know, just the guy who literally invented the thing that made ChatGPT possible. No big deal.
And it does not stop there. Jonas Adler and Alexander Pritzel, two researchers who were considered key architects of Google’s Gemini model, are also reportedly heading to Anthropic. At this point Google is not just losing a few smart people. It is losing the people who built the foundation of modern AI.
The market noticed. Alphabet stock had its worst single day in more than a year, dropping 5% in one session. Investors who were betting on Google winning the AI race are now having a very uncomfortable conversation with themselves.
The reason this is happening is pretty obvious when you think about it. Anthropic and OpenAI are both heading toward IPOs. A mid-level researcher at a publicly traded giant might see their options appreciate modestly over the next decade. The same researcher joining a pre-IPO AI company? That is a lottery ticket worth potentially tens or hundreds of millions of dollars. Google is a massive publicly traded company with decades of bureaucracy and a stock that is already baked into every retirement fund in America. The upside just is not there anymore.
What makes this particularly painful for Google is that they helped create this situation. Google was an early investor in Anthropic. They backed the company financially while it was essentially being staffed by former Google and DeepMind employees. They handed their competitors both capital and talent simultaneously. And now those competitors are pulling off their best remaining people one by one.
The more interesting question is whether Google can actually keep up with two companies that are increasingly staffed by its own former employees, running on its own compute investments, and about to be flush with IPO cash. Right now that is not a question with an obvious answer, and Wall Street is starting to wonder the same thing.
Read the full story at TechCrunch
ANTHROPIC IS NOW OFFICIALLY WORTH MORE THAN OPENAI AND MAKING MORE MONEY TOO
A year ago if you had told someone that a company founded by a bunch of people who quit OpenAI because they were worried about AI safety would end up being more valuable than OpenAI itself, they would have laughed at you. And yet here we are.
Anthropic just crossed a $47 billion annual run rate in revenue. For context, OpenAI’s latest disclosed numbers put its annualized revenue somewhere between $25 and $33 billion. So Anthropic is making nearly double what OpenAI is making, or at least that is what Anthropic is telling people ahead of its IPO filing, and nobody has called them liars yet.
The valuation story is equally surreal. Anthropic raised $65 billion in a Series H funding round at a $965 billion valuation. That puts it above OpenAI in terms of what investors are willing to value the company at. A company that did not exist five years ago is now worth approximately one trillion dollars. The word trillion used to sound impressive. In the AI industry in 2026 it has started to sound like the price of a decent data center.
What drove this? Mostly Claude Code. Anthropic’s AI coding assistant absolutely exploded. Developers fell in love with it, companies started buying enterprise licenses, and the revenue numbers went vertical. Claude Code became the product that made Anthropic a real business, not just a research lab that happened to charge money.
The other driver was the Mythos model. Anthropic’s most powerful model was originally under export controls because the US government decided it was a national security concern, which is a very strange form of endorsement when you think about it. When the controls were lifted in late June and Mythos went into wider release, adoption spiked again.
Meanwhile OpenAI is dealing with a weirder version of the same IPO game. They filed confidentially with the SEC, but reports now suggest they might push the actual IPO to 2027 instead of this September. The markets are confused about the timeline, which is not a great look when you are trying to convince investors you have everything under control.
The rivalry between these two companies is fascinating because they came from the same place. Dario Amodei and his sister Daniela left OpenAI to start Anthropic. They hired a bunch of OpenAI people. Sam Altman spent years treating them like a nuisance. And now Anthropic is worth more and making more money and pulling Nobel laureates away from Google while it prepares to go public. You could not write this story and have people believe it was fiction.
SAM ALTMAN WANTS TO RUN THE WORLD’S ARTIFICIAL INTELLIGENCE AND HE WROTE AN OP-ED ABOUT IT
Sam Altman published an op-ed in the Financial Times recently calling for a new global body to govern artificial intelligence. He wants something like the International Atomic Energy Agency but for AI. A multinational forum with government representatives, technical experts, and oversight mechanisms to keep the AI race from becoming genuinely dangerous.
On paper this sounds reasonable. AI is a global technology with global implications, and the people building it probably should not be the only ones deciding what it is allowed to do. Altman’s proposal includes accepted standards, independent analysis, and a mechanism to hold labs accountable when commercial pressures push them toward cutting corners on safety. These are things that reasonable people across the political spectrum could probably agree with.
And then you read the Fortune piece that ran the same week, and you get the rest of the context. OpenAI is slowly losing ground to both Google and Anthropic. The company that invented the modern era of consumer AI is suddenly in the unusual position of defending its status as the most important player in the game.
OpenAI has lost Noam Shazeer to a talent war it is losing at the top. Its revenue is being outpaced by Anthropic. Google has been releasing model after model trying to claw back ground. And OpenAI’s IPO timeline, once confidently pointed at September, now looks like it might slip to 2027. That is a lot of turbulence for a company trying to tell the world it is the responsible adult in the room.
None of this means Altman’s governance proposal is insincere. He has been talking about AI risk longer than most people have been talking about AI at all. But it is impossible to read the op-ed without noticing that it serves a dual purpose. Positioning yourself as the statesman of the AI age while your competitors are running faster than you is a smart move. It says: even if we are not the fastest, we are the most responsible, and that matters.
The actual substance of the proposal is worth taking seriously regardless of the timing. An IAEA-style body for AI would be genuinely unprecedented, and getting 193 countries to agree on anything related to technology is not something that happens without years of diplomacy. Whether this goes anywhere is very much an open question. But the fact that the CEO of one of the most powerful AI companies on earth is publicly calling for regulation of his own industry is at least interesting, even if you are deeply skeptical of his motives.
The world is going to need some kind of governance structure for this technology eventually. Whether Altman gets to chair the committee that writes the rules is a separate question entirely.
Read the full story at Fortune
EUROPE’S SCRAPPY FRENCH STARTUP IS BUILDING A CLOUD TO CHALLENGE AMAZON AND MICROSOFT
Mistral AI has been a fascinating company to watch. Founded in Paris by researchers who left Meta and DeepMind, they spent most of their first couple of years being described as “Europe’s answer to OpenAI,” which is both high praise and slightly condescending at the same time, like telling someone they make the best pizza outside of Italy.
But Mistral has now stopped being content with being a scrappy model-building startup and has announced a full-scale push to compete with Amazon Web Services, Microsoft Azure, and Google Cloud in the cloud computing market. They are calling it Mistral Compute. It is a four billion euro investment in data centers in France and Sweden, with plans to hit 200 megawatts of capacity by 2027 and a full gigawatt by 2030. For reference, a gigawatt is the kind of number you see attached to national electricity grids and nuclear power stations, not AI startups based in Paris.
Alongside the compute push, they are renaming their AI assistant. Le Chat — which was a perfectly fine name for a French AI product if you enjoy the gentle bilingual pun — is being rebranded as Vibe. Vibe is now a unified agent platform for enterprise productivity, which is Mistral’s way of saying they want to be your company’s AI system for everything, not just a chat window.
The strategic ambition here is genuinely noteworthy. Most European AI companies have been operating in the shadow of American hyperscalers, renting compute from AWS or Azure and accepting that their infrastructure decisions were ultimately made in Seattle or Redmond. Mistral is explicitly rejecting that model. They are building their own hardware infrastructure so that European governments and companies can run sensitive workloads without touching American cloud providers.
This matters for a reason that goes beyond company strategy. The European Union has been deeply uncomfortable with the dependence of its digital economy on US tech infrastructure. The GDPR was partly a reaction to that discomfort. The EU AI Act that governs how models are trained and deployed followed it. The idea of a major European AI company offering not just compliant models but compliant compute, running on hardware physically located in France and Sweden, is exactly what Brussels has been hoping someone would do for years.
Whether Mistral can actually compete with AWS and Azure at scale is a genuine question. Amazon and Microsoft have years of head start, enormous engineering teams, and pricing power that comes from serving millions of customers. Mistral is betting that data sovereignty, regulatory compliance, and European identity are worth paying a premium for. A lot of European enterprises and governments might well agree. If even a fraction of them shift their workloads, Mistral’s numbers start looking very interesting very quickly.