We don't talk enough about the quiet war between AI labs. But the numbers hit me like a cold front during Nairobi's dry season. Anthropic's annualized revenue run rate just crossed $65 billion at the end of July, leaving OpenAI $25 billion behind. That's more than the combined market cap of Ethereum, Solana, and Avalanche at today's prices. The bear market didn't kill centralized AI—it supercharged it.
Yet here I am, a decentralized protocol PM in Nairobi, watching this from the sidelines. The irony isn't lost on me. I've spent seven years arguing that trustless systems will eat the world, and the world's most valuable private companies are now built on trust, not code. Anthropic filed a confidential prospectus with the SEC in June, and Bloomberg reports its Wall Street debut could come as early as this fall. The Financial Times expects a $2 trillion valuation.
Let's parse the raw data first. Anthropic's revenue run rate expanded 622% in seven months, from roughly $9 billion at the end of 2025 to $47 billion in May, then to $65 billion by July. The May-to-July stretch alone added $18 billion, a 38% jump. Preliminary second-quarter revenue topped $11.5 billion, against $787 million a year earlier—a 14.6x increase. The company posted positive adjusted operating income for the period. Meanwhile, OpenAI trails at $40 billion run rate, roughly double its end-of-2025 level.
These numbers come from investor updates, not public filings. Neither company confirms the exact calculation methodology. But the direction is clear: centralized AI is capturing unprecedented economic value at a speed that makes even the most aggressive crypto bull runs look pedestrian.
Context: The Centralization Paradox
I've been here before. In 2017, I spent 150 hours tracing the reentrancy vulnerability in The DAO smart contract. That experience taught me that code is law only when the law is auditable and immutable. Today, Anthropic and OpenAI control their models, their training data, and their inference pipelines. They are black boxes with billion-dollar revenue streams. The same investors who pour money into DeFi protocols are also pouring billions into these closed-source AI labs.
This is the centralization paradox of our era. We celebrate permissionless innovation while the most innovative technology on the planet—large language models—remains permissioned. Anthropic's Claude and OpenAI's ChatGPT are accessed through APIs with rate limits, content filters, and centralized servers. If Anthropic's servers go down, so does your AI assistant. If its board decides to change terms, you adapt or leave.
Core: The Technical-Values Analysis
Based on my audit experience, the revenue run rate numbers tell a story about value capture, not value creation. Anthropic is charging for access to intelligence that was trained on public data, much of it scraped without consent. The company's cost structure is dominated by compute—NVIDIA GPUs, cloud infrastructure, electricity. The gross margins are likely high, but the moat is not the model architecture; it's the capital expenditure required to train the next generation.
In crypto, we measure value capture through TVL, fees, and token velocity. Let's run the numbers side by side. Anthropic's $65 billion run rate translates to roughly $5.4 billion per month. The entire DeFi ecosystem, across all chains, generates about $500 million in monthly fees on a good month. That's a 10x difference. The largest DeFi protocol, Lido, has a market cap of about $2 billion. Anthropic is valued at $2 trillion—a 1,000x multiple.
I'm not saying DeFi is failing. I'm saying the market is pricing centralized AI as a utility—like electricity or water—while pricing decentralized finance as a speculative bet. The bear market of 2022-2023 crushed token prices but didn't destroy the underlying technology. Yet the revenue gap between centralized and decentralized systems is widening, not shrinking.
Where the Decentralized AI Thesis Fits
In 2025, I launched a prototype called TruthLayer, a decentralized registry for AI-generated media. The idea was simple: use blockchain to prove whether a piece of content was generated by a human or an AI model. We integrated watermarking algorithms with IPFS storage and attracted 500 beta testers in a month. What I learned from that project is that users don't care about the tech stack. They care about trust. And trust, in the current paradigm, is provided by centralized brands like Anthropic.
But here's the contrarian insight: the revenue run rate explosion is a lagging indicator, not a leading one. Anthropic's Q2 revenue of $11.5 billion is impressive, but it's driven by enterprise contracts and API usage. Those contracts are short-term. The real value in AI will come from models that can be verified, audited, and composed like DeFi legos. That's where blockchain enters the picture.
Contrarian: The Blind Spot in the AI Revenue Story
Counter-intuitive angle: The $65 billion run rate might be a warning sign for the crypto industry, not a validation of centralization. Look at the numbers more carefully. Anthropic's revenue growth is 622% in seven months. That's exponential. But exponential growth in a centralized system creates exponential risk. A single regulatory change, a security breach, or a model collapse could wipe out billions in value overnight. The crypto industry learned this lesson in 2022 with FTX and Terra.
The bear market didn't kill centralized AI—it taught it to raise money at high valuations. But the bull market for centralized AI will end the same way every tech bubble ends: with a crash. The question is whether decentralized alternatives will be ready to absorb the displaced value.
I've seen this pattern before. In 2020, DeFi Summer was dismissed as a passing fad by traditional finance. By 2024, institutions were building on-chain. The same arc will happen with AI. The revenue run rate gap is a mirage. It reflects the current monopoly on compute and data, not the inevitable architecture of AI. The real innovation will come from distributed inference, federated learning, and on-chain model verification.
Takeaway: The Vision Forward
Anthropic's IPO will be a landmark event. It will likely be the largest tech IPO in history, dwarfing even Alibaba's $25 billion debut. But I'm not buying the hype. I'm building the infrastructure for the post-Anthropic world. The bear market of 2022 taught me resilience. The AI boom of 2025 is teaching me patience. The real value of AI won't be captured by a single company. It will be distributed across a network of contributors, validators, and users—all secured by code that can be read by anyone, anywhere.
We don't need another centralized AI. We need a decentralized AI that the bear market can't kill and the bull market can't corrupt. That's the thesis I'm building at TruthLayer and in my work as a protocol PM. The numbers are staggering, but they're not the final word. They're the opening chapter of a longer story.