🤚 The Open-Palm Illumination
In the span of what can only be described as the most fiscally unhinged week in AI history, Anthropic casually collected $65 billion in investment commitments from two companies that are, technically speaking, also its competitors. Google pledged up to $40 billion — $10 billion in cash now, $30 billion more if Anthropic hits certain TPU consumption milestones. Amazon, not to be outperformed in the art of writing checks to frenemies, committed $25 billion tied to AWS compute targets. Combined, that’s $65 billion in a single corporate courtship sprint. The valuation? Currently hovering around $380 billion, with whispers of a $900 billion round within weeks.
But wait — the AI arms race wasn’t content with merely financial spectacle. Moonshot AI dropped Kimi K2.6, a 1-trillion-parameter Mixture-of-Experts model that activates 32 billion parameters per token and can orchestrate up to 300 sub-agents executing 4,000 coordinated steps in a single autonomous run. Meanwhile, OpenAI released GPT-5.5 on April 23rd, its “smartest and most intuitive model yet,” available at $5 per million input tokens and $30 per million output tokens. Fifteen major AI releases in eight weeks. The industry is no longer iterating. It is detonating.
👐 The Two-Handed Reality Check
Let us pause and appreciate the structural absurdity of a company receiving billions from both its cloud providers while simultaneously competing with both of them in the consumer AI market. Google makes its own models — Gemini is right there. Amazon has its own AI ambitions through Bedrock. Yet both are writing Anthropic checks that could fund a small European nation’s annual budget. Why? Because in the AI gold rush, the play isn’t to mine gold yourself — it’s to own the pickaxes, the railroad, and the guy who’s actually finding gold.
The compute deals are the real story. Google is offering 5 gigawatts of TPU capacity over five years. Amazon is bundling its investment with AWS compute commitments. These aren’t philanthropic gestures — they’re infrastructure lock-in agreements dressed in venture capital clothing. Anthropic gets the compute it desperately needs to train frontier models. Google and Amazon get a customer who will spend billions renting their servers. Everyone wins, until someone doesn’t.
And then there’s Kimi K2.6, quietly demonstrating that the frontier model club is no longer American-only. Moonshot AI’s open-weight model can coordinate hundreds of sub-agents autonomously — a capability that makes most enterprise AI deployments look like a calculator with a personality disorder. The modified MIT license is generous enough for startups, restrictive enough for companies with 100 million monthly users. China’s AI ecosystem isn’t just keeping pace. It’s setting pace.
🌿 The Gentle Awakening
Perhaps the most revealing data point of the week came not from a boardroom but from a hospital. A Harvard Medical School study, published April 30th, found that OpenAI’s latest reasoning model matched or outperformed experienced physicians in diagnosing patients using only electronic health records. The AI scored higher than two board-certified doctors who had access to the same information at the time of diagnosis. The doctors, to their credit, had the disadvantage of also needing to deal with insurance forms, sleep deprivation, and the existential weight of being human.
The researchers were quick to note the caveats: the AI worked from text alone, without the benefit of reading body language, listening to breath sounds, or noticing that the patient was, in fact, lying about their alcohol consumption. Still, when an algorithm outperforms a physician on diagnostic accuracy using less information, the conversation shifts from “should AI assist doctors?” to “should doctors assist AI?”
We are not there yet. But the GPS is recalculating.
👑 The Crown Verdict
What Peter Diamandis and his Moonshots panel captured in Episode #252 is the feeling of standing inside a particle accelerator while someone increases the energy. Fifteen major model releases in two months. $65 billion committed to a single company in a single week. A Chinese open-source model coordinating 300 autonomous agents. An AI outdiagnosing Harvard-trained physicians. These aren’t separate stories — they’re symptoms of the same condition: exponential acceleration with no off-ramp in sight.
The AI model race has entered its Formula 1 phase — absurd speeds, astronomical budgets, and the occasional spectacular crash that only makes the audience more enthusiastic. Anthropic’s revenue hit $30 billion annualized in March, up 1,400% year-over-year. OpenAI is pushing GPT-5.5 as its super-app play. Moonshot AI is open-sourcing trillion-parameter models. And somewhere in a hospital, a neural network just made a better diagnosis than the human who spent twelve years in medical school.
The question is no longer whether AI will transform everything. The question is whether we’ll notice before it’s already happened.
Inspired by Anthropic’s $65B Week, The AI Model Race, and ChatGPT Beats Doctors | EP #252 by Peter Diamandis.
Your complacency is showing. Invest wisely.