The Bloomberg terminal blinked. Q2 2025 preliminary revenue for Anthropic: $11.5 billion. A 14x jump from $787 million a year prior. Adjusted operating profit positive. The market cheered. I did not.
Numbers without a verifiable trail are not data; they are marketing. My ten years auditing blockchain protocols have taught me one immutable rule: the ledger does not lie, only the operators do. Anthropic’s revenue surge smells less like a breakthrough and more like a carefully constructed narrative—one that would never pass a basic on-chain proof-of-reserves test.
Context: The AI Hype Cycle Meets the IPO Boom
The AI sector is currently in the euphoria phase of its own hype cycle. According to Bloomberg, IPO financing has reached $256.4 billion so far this year—the highest since 2021, excluding SPACs. That year, 2021, was also the peak of crypto’s last bull run. The parallels are uncomfortable. In 2021, we saw projects with no product, no revenue, and no code raise billions on whitepapers alone. Today, we see AI companies with impressive revenue figures but zero transparency into how those figures are constructed.
Anthropic’s annualized revenue is now quoted at $47 billion, surpassing OpenAI’s claimed $40 billion. Yet the underlying calculation methodology remains proprietary. Neither company publishes a public, auditable proof of their revenue—no on-chain settlement data, no third-party attestation, no verifiable cash flow. In the blockchain world, such opacity would be a red flag triggering immediate withdrawal of liquidity. In the AI world, it triggers a higher valuation.
Core: A Systematic Teardown of the Revenue Claims
Let me apply the same forensic framework I used during the FTX collapse—where I identified a $7.2 billion discrepancy in user asset segregation—to Anthropic’s reported numbers.
First, the growth rate. From $787 million to $11.5 billion in one year implies a compound monthly growth rate of approximately 30%. That is not unheard of in hypergrowth scenarios, but it is statistically anomalous. For context, during the peak of the 2021 crypto bull market, the fastest-growing DeFi protocols (like Uniswap) saw monthly volume growth of 15-20% for a few quarters, then plateaued. Sustained 30% monthly growth over 12 months would require an exponential increase in user adoption that strains credulity.
Second, the revenue composition. Bloomberg reports that Anthropic’s growth is driven by professionals using its software for programming and workflow automation. But what is the unit economics? If each user pays $20 per month, you would need 575 million active users to generate $11.5 billion in a quarter. That is roughly the population of the United States and Canada combined. Alternatively, if the revenue is enterprise-based, with large contracts, then the number of customers is smaller—but the churn risk becomes systemic. Based on my experience auditing the Ethereum 2.0 Merge, where edge cases in the difficulty bomb schedule could have destabilized the entire chain, I recognize that concentrated revenue sources are a single point of failure.
Third, the adjusted operating profit. “Adjusted” is a term that should trigger immediate skepticism. In the crypto world, we have seen countless projects report “adjusted” TVL by excluding certain assets or counting staked tokens multiple times. Anthropic’s “adjusted” profit likely excludes R&D costs, marketing expenses, and capital expenditures. Without a standardized accounting framework—akin to the GAAP for on-chain protocols—these numbers are meaningless for comparison.
I cross-referenced Anthropic’s revenue claims with publicly available data on token sales and API usage. The results are inconclusive. There is no independent source—no blockchain transaction log, no audited financial statement, no regulatory filing—that confirms the $11.5 billion figure. The only evidence is Bloomberg’s report, which itself relies on unnamed sources. In my FTX post-mortem, I highlighted exactly this kind of reliance on unaudited, self-reported metrics as a precursor to collapse.
Contrarian: What the Bulls Got Right
To be fair, I must acknowledge the counterarguments. Enterprise adoption of AI is real. I have seen supply chain managers, legal teams, and even risk consultants like myself use Anthropic’s models to streamline workflows. The utility is undeniable. The company has also secured contracts with government agencies, which provide a degree of revenue stability.
Furthermore, the AI sector’s growth is not purely speculative. Unlike the 2021 crypto bubble, where many projects had no working product, Anthropic has a functional software platform with millions of users. The revenue may be inflated by aggressive sales tactics, but it is not entirely fabricated. The bulls argue that the growth rate reflects genuine market demand, and that the lack of transparency is a temporary artifact of private company reporting.
But here is the blind spot. The same arguments were made about FTX. “The revenue is real, the users are real, the platform works.” Until the day it didn’t. The difference is that FTX had on-chain data that could be audited after the fact. Anthropic has no such trail. The data does not negotiate; it only confirms. And until Anthropic submits to a verifiable, on-chain audit of its revenue, these numbers remain a hypothesis, not a fact.
Takeaway: The Accountability Call
The AI industry is repeating the same mistakes we saw in crypto during the 2021 bull run. Unverified metrics, opaque financial structures, and a collective willingness to trust rather than verify. History is the only reliable audit trail, and history tells us that every hype cycle ends with a crash.
Proof is cheaper than trust, yet still ignored. Anthropic’s $11.5 billion quarter will be cited as a sign of AI’s dominance—until the next correction reveals the fragility behind the numbers. The question is not whether the revenue is real. The question is whether the market will demand a verifiable audit before the music stops.
Silence in the code is a bug waiting to happen. Silence in the financial report is a fraud waiting to be exposed.