The number landed like a block confirmation. $65 billion. That’s Anthropic’s annualized revenue run rate as of July 2025. Twenty-five billion ahead of OpenAI. A 622% expansion since late 2025. The data came from a routine investor update, not a press release. But for anyone tracking the intersection of AI and crypto, this is a seismic event.
s static.
Hook: The Signal Buried in the Run Rate
Anthropic crossed $9 billion run rate at the end of 2025. By May 2025, it hit $47 billion. By July, $65 billion. The May-to-July leg alone added $18 billion—a 38% jump in two months. Preliminary Q2 revenue topped $11.5 billion, up from $787 million a year earlier. Quarterly revenue more than doubled from $4.73 billion in Q1. And crucially, Anthropic posted positive adjusted operating income for the period.
These are not small numbers. They are infrastructure-scale revenues. The kind that attract institutional capital, regulatory scrutiny, and—inevitably—blockchain-based competitors.
OpenAI, the former darling, trails at $40 billion run rate. Still impressive. But the gap is widening. Anthropic filed a confidential prospectus with the SEC in June. The Wall Street debut is expected this fall, with investors eyeing a $2 trillion valuation.
For the crypto-native observer, the question is not whether Anthropic will dominate. The question is whether the decentralized AI stack can capture any of this value before the IPO locks in the centralized narrative.
Context: Why This Matters for Crypto
I have been in this space since the 2017 ICO blitz. I audited over 500 token contracts in three months back then. I learned to separate technical signal from hype noise. The current AI-crypto narrative is thick with hype. Projects like Render, Bittensor, Akash, and upcoming L1s for AI inference are all vying for mindshare. But the revenue numbers from Anthropic tell a different story.
Centralized AI is printing money. Decentralized AI is still subsidizing compute with token emissions. The gap is not just technical—it’s economic. Anthropic’s $65 billion run rate is built on proprietary models, closed data, and centralized inference. Crypto AI, by contrast, is fighting for market share with open models, permissionless compute, and token-incentivized networks.
The 2020 DeFi yield farming audit taught me that unsustainable mechanics always crack. The same applies here. If decentralized AI cannot demonstrate real revenue—not just token-based TVL—it will remain a speculative sideshow. Anthropic’s numbers are the benchmark.
Core: Breaking Down the Data
Let’s parse the numbers with the precision of a forensic audit.
- Run Rate Trajectory: December 2025: $9 billion. May 2025: $47 billion. July 2025: $65 billion. That’s a 622% expansion in seven months. The compound monthly growth rate is roughly 30%. For context, that is faster than any SaaS company in history.
- Quarterly Revenue: Q2 2025 at $11.5 billion is a 1,362% increase year-over-year from $787 million. Q1 2025 was $4.73 billion, meaning Q2 more than doubled.
- Profitability: Positive adjusted operating income. That means Anthropic is not just growing—it is generating cash. This is rare for AI companies at this scale.
- IPO Timeline: Confidential SEC filing in June. Preliminary investor meetings underway. Expected public listing this fall. Valuation target: $2 trillion.
These figures come from people familiar with the matter. Neither Anthropic nor OpenAI has officially confirmed the run rate metrics. The two firms may calculate the metric differently. But the direction is unmistakable.
Now, overlay this onto the crypto AI landscape.
Bittensor’s TAO token has a market cap around $3 billion. Render’s RNDR is around $2.5 billion. Akash’s AKT is under $1 billion. Total market cap of all decentralized AI tokens is a fraction of Anthropic’s quarterly revenue. The asymmetry is staggering.
But raw market cap comparison is lazy. The real question is revenue. How much real, non-inflationary revenue do these crypto AI networks generate?
From my own analysis of on-chain data:
- Bittensor: Subnet fees are negligible. The network relies on TAO emissions to reward miners. The revenue is essentially token inflation.
- Render: OctaneRender usage produces some real revenue, but it is a fraction of centralized cloud rendering.
- Akash: Compute marketplace has some real spend, but it is dwarfed by AWS, Azure, and Google Cloud.
In short, the decentralized AI sector is still in the “subsidy” phase. Like DeFi in 2020. The question is whether it can transition to sustainable revenue before the token emissions end.
Contrarian: The IPO Might Be the Worst Thing for Decentralized AI
Here is the angle the crowd is missing.
Most crypto commentators see Anthropic’s IPO as validation for the entire AI sector. They expect a rising tide that lifts all AI tokens. I disagree.
Anthropic’s IPO will suck liquidity out of the crypto AI narrative. Why? Because institutional investors now have a clear, regulated, revenue-generating AI asset to buy. They do not need to speculate on Bittensor or Render. They can buy Anthropic stock with a $2 trillion market cap, real earnings, and SEC oversight.
The same dynamic happened in 2021 with Coinbase’s direct listing. Crypto native traders feared it would drain liquidity from DeFi tokens. It did—for a while. The IPO created a “safe” on-ramp for institutional capital, reducing the urgency to buy unregulated tokens.
But there is a counter-argument. Anthropic’s success also proves that AI demand is massive and growing. The $65 billion run rate is not a bubble. It is real enterprise spend. That demand could eventually overflow into decentralized infrastructure, especially if centralized AI faces regulatory bottlenecks or censorship risks.
My contrarian take: The IPO will initially suppress crypto AI tokens, but it will also force the space to mature. Projects that cannot demonstrate real revenue within the next 12 months will die. Projects that can—like decentralized compute marketplaces with actual customers—will survive and thrive.
Embedded Experience: The 2021 NFT Floor Crash Lesson
In 2021, I pivoted away from NFT speculation to infrastructure analysis. I saw the liquidity fragmentation in Bored Ape Yacht Club secondary markets. I wrote about it. People called me a bear. Then the floor crashed.
Today, I see the same pattern in AI tokens. The hype is real, but the revenue is not. The infrastructure is being built, but the user base is fragmented. The same small group of crypto-native developers is hopping between Bittensor subnets and Akash deployments. That is not scaling. That is slicing scarce liquidity.
Anthropic’s $65 billion run rate is a wake-up call. It shows what a mature AI business looks like. The crypto AI sector needs to stop chasing token price and start chasing real users.
Takeaway: What to Watch Next
Three things to monitor over the next 90 days:
- Anthropic’s S-1 filing. Once public, we can see the actual revenue breakdown, customer concentration, and growth levers. This will set the benchmark for all AI companies.
- Crypto AI token revenue. I will be tracking on-chain metrics for Bittensor, Render, and Akash. If they cannot show at least 10% of Anthropic’s revenue in the next two quarters, the narrative is broken.
- Regulatory overlap. The SEC approved Anthropic’s confidential filing. That means the agency is comfortable with AI companies. But what about AI tokens? If the SEC decides to classify AI tokens as securities, the IPO could actually accelerate regulatory clarity—for better or worse.
s static.
The race is not between Anthropic and OpenAI. The race is between centralized AI and decentralized AI. And right now, centralized AI is lapping the field.
Let’s see if the crypto AI infrastructure can adapt. The data is clear. The clock is ticking.