Anthropic's annualized revenue run rate hit $65 billion at the end of July, leaving OpenAI $25 billion behind. The numbers are staggering—a 622% expansion since late 2025. But geometry remembers what markets forget. This isn't a story of triumph; it's a parable about centralization, dressed in quarterly earnings.
Context
Anthropic, the AI startup behind Claude, is preparing for a public listing as soon as this fall. Its confidential SEC filing and preliminary investor meetings suggest a valuation near $2 trillion. Meanwhile, OpenAI trails at a $40 billion run rate. Both figures come from anonymous sources, not audited books. But the narrative is clear: the AI industry is consolidating into two corporate giants, each chasing infinite growth.
In the crypto world, we've seen this movie before. The ICO boom of 2017 promised decentralization but delivered centralized token distributions. DeFi Summer 2020 promised composability but created liquidity fragmentation across dozens of Layer2s. Now, AI is following the same playbook: massive capital inflows, hyped growth metrics, and a small user base sliced into ever-thinner segments.
Core: The Run Rate Ruse
Annualized run rate is a vanity metric. It extrapolates a single month's revenue into eternity, ignoring seasonality, churn, and competitive dynamics. Anthropic's $65 billion run rate is based on a preliminary Q2 revenue of $11.5 billion—up from $787 million a year ago. That's a 1,361% increase in quarterly revenue. But the May-to-July stretch added only $18 billion in run rate, a 38% gain. The growth is decelerating.
During my years auditing DeFi protocols, I learned that rapid growth often masks structural fragility. In 2022, I analyzed the governance tokens of 12 major DAOs and found centralization flaws in their voting mechanisms. The same pattern applies here: Anthropic's revenue is concentrated in enterprise contracts and API usage. If one large client switches to a rival, the run rate collapses.
DeFi breathes; don't suffocate it. In decentralized finance, liquidity is distributed across thousands of pools, creating resilience. In centralized AI, value is hoarded in corporate treasuries. The $65 billion run rate is a signal of monoculture, not health. It's the same dynamic that makes L2s dangerous: dozens of chains, but the same small user base. Here, dozens of AI models, but the same few dozen customers.
Contrarian: The Efficiency Argument
Some argue that centralized AI is necessary for performance. Training a frontier model requires exascale compute clusters, which only a few companies can afford. Decentralized alternatives like Bittensor or Render Network are orders of magnitude slower and less capable. The market is voting with its wallet: Anthropic and OpenAI are winning because they deliver the best products.
But silence is the loudest warning. The very efficiency of these models creates a dependency risk. If Anthropic's servers go down, or if its API pricing doubles, entire industries grind to a halt. That's not scaling—it's fragility amplified. In crypto, we've seen how a single smart contract bug can drain billions. In AI, a single corporate decision can alter the cognitive infrastructure of the planet.
Prune the dead branches, save the tree. The dead branch here is the assumption that centralized AI is sustainable. The tree is the broader ecosystem of human intelligence. Blockchain offers a counter-narrative: verifiable, permissionless, and composable AI services. Zero-knowledge proofs can verify model outputs without revealing proprietary weights. On-chain governance can distribute control across stakeholders, not shareholders.
Takeaway: The Real Race
Anthropic's IPO is not a milestone—it's a warning. The market is pricing centralized intelligence as a scarce resource, but the true scarcity is trust. As AI-generated content floods the internet, the ability to verify human intent becomes paramount. That's where crypto's "Proof of Human" thesis shines. The race to $2 trillion valuation is a distraction. The real race is for the soul of intelligence itself.
Will we let a handful of corporations define our collective cognition? Or will we build an open, decentralized substrate where intelligence is a public good? The numbers say profit. The geometry says decay. I choose geometry.