A 30-year Wall Street macro investor who tracks his own AI model usage as a leading indicator says the consumer AI race has already narrowed to two: ChatGPT and Claude. Google’s Gemini, he says, is “near zero” and “not a player anymore.”
Jordi Visser, appearing on the Anthony Pompliano podcast in a episode titled “Why Are Bitcoin & AI Stocks CRASHING?!” (61,000 views in 24 hours), laid out personal usage data that tracks with a broader pattern: Google had the talent, the cash, and the compute — and still lost the consumer AI mindshare.
🔍 THE BOTTOM LINE: When the power users — the people who live in these tools daily — stop using yours, the revenue impact shows up quarters later. Visser’s data is anecdotal. The brain drain is not.
The Usage Data
Visser tracks his own model usage as a proxy for where the market is heading. His numbers, shared on the podcast:
- Claude: ~35% two months ago → ~45% now
- ChatGPT: ~25% two months ago → ~45% now
- Gemini: ~20% two months ago → near zero now
- Others: ~20% → ~10%
“Google’s not a player in this anymore,” Visser said.
This is one man’s usage data, not a market survey. But Visser runs a 100-name AI portfolio and has been in macro investing for three decades. He’s the kind of power user whose behaviour precedes the revenue charts by a quarter or two. When the people who test these tools for a living stop opening yours, that’s a signal.
The Nobel Laureate Who Walked
The same week Visser’s podcast dropped, Business Insider reported that John Jumper — a key member of Google DeepMind’s AlphaFold team for nearly a decade and a Nobel Prize winner — left DeepMind for Anthropic. Reuters confirmed the move on June 19.
This isn’t a mid-level engineer chasing a bigger paycheck. This is one of the scientists behind AlphaFold, the protein-folding breakthrough that won DeepMind a Nobel Prize. He’s going to Anthropic — the company behind Claude.
When the architect of your most celebrated scientific achievement walks to your closest competitor, that’s not a retention problem. That’s a strategic failure.
We covered Jumper’s departure in our export control analysis, but the talent drain narrative is now impossible to ignore. It’s not just Jumper. The G7 AI summit saw Anthropic CEO Dario Amodei sharing the stage with DeepMind CEO Demis Hassabis — but the subtext was clear: Anthropic is absorbing the talent Google can’t keep.
Multiple Compression Before Failure
Visser’s investment thesis is the part that should worry Google shareholders. He’s not saying Google will fail. He’s saying the stock can get crushed without the business breaking.
“A business does not have to fail to destroy shareholder value,” he said on the podcast.
The pattern: Meta in 2022, Microsoft in 2014. The market reprices expectations before the actual business deteriorates. If investors decide Google’s AI position is structurally weaker than they thought — that Gemini is a distant third, not a contender — the multiple compresses. The stock drops 30-40% while the search business still prints money.
Visser wouldn’t short Google. But he’s not long it either.
Why Two Horses, Not Three
The consolidation thesis is straightforward. Consumer AI is a distribution game, and distribution rewards focus. ChatGPT has the brand — it’s the verb. Claude has the power users — developers, researchers, and the kind of people who switch models weekly and notice the difference.
Gemini has neither. It has integration — Gmail, Docs, Search — but integration is a moat only if people use the product. If power users are dropping it to near zero, the integration advantage is a feature nobody opens.
As we noted in our AI adoption reality check, 70% of the US working-age population still isn’t using AI tools at all. The market is early. But among the 30% who are, the split is becoming stark.
❓ FAQ
Is Visser’s usage data scientific? No. It’s one investor’s personal tracking. But he runs an AI-focused portfolio and has been a macro investor for 30+ years. Anecdotal data from sophisticated users often precedes broader trends by a quarter or two.
Did two DeepMind scientists leave for Anthropic? One confirmed departure: John Jumper, Nobel laureate and AlphaFold team member, left DeepMind for Anthropic in June 2026. Reuters and Business Insider both confirmed the move. Visser referenced “two” on the podcast — the second departure hasn’t been independently confirmed.
Does this mean Google is done in AI? No. Google has DeepMind, massive compute, the world’s largest AI research budget, and TPU chips. But having the infrastructure and losing the talent that makes it productive is a recognised pattern in tech. The question isn’t whether Google can build AI — it’s whether anyone uses it.
What about open-source models? Visser grouped all “others” — including open-source — at ~10% and declining. Open-source isn’t winning the consumer race either, though it serves a different market.
🔍 THE BOTTOM LINE: The consumer AI race looks like it’s down to ChatGPT and Claude. Google had every advantage — money, talent, distribution — and fumbled it. A Nobel laureate walking to your competitor is the talent equivalent of multiple compression: the damage shows up in the stock before it shows up in the business. Google isn’t failing. But Google’s AI position might already be priced for a reality it hasn’t admitted yet.
