Hook
Crypto Briefing, a publication often cited by digital asset enthusiasts, published a headline that claimed Anthropic and OpenAI have achieved a combined revenue of $120 billion, surpassing Starbucks and McDonald's. The data tells a different story. I ran the numbers, and the math doesn't add up. The claim is built on a fundamental confusion between revenue and valuation. Based on my experience auditing ICO tokenomics in 2017, I learned that the most dangerous narratives are those that mix a grain of truth with a lie. This is one of them. Ledgers do not lie, only the narrative does.
Context
The source is Crypto Briefing, a media outlet that primarily covers blockchain and cryptocurrency. Its audience is retail traders and crypto investors, not financial analysts. The article lacks any reference to official filings, audited financial reports, or independent data. In 2020, during DeFi Summer, I analyzed Uniswap V2 liquidity and discovered that many protocols inflated their trading volumes through wash trading. That experience taught me that verification is the only shield against hype. Here, the claimed $120 billion revenue for OpenAI and Anthropic is not just unverified—it contradicts publicly available data. OpenAI's reported annualized revenue as of late 2024 is approximately $3.7 billion (per The Information), while Anthropic's is roughly $1 billion (per source estimates). Combined, they are at $4.7 billion. Starbucks reported $40 billion in fiscal 2024, and McDonald's $25 billion. The real gap is an order of magnitude. The $120 billion figure aligns more closely with their combined valuations: OpenAI was valued at $157 billion in October 2024, and Anthropic at $60 billion, totaling $217 billion. Crypto Briefing likely mistook valuation for revenue—a common but dangerous error.
Core: The On-Chain Evidence Chain
Let me apply the same method I used in 2022 when I modeled contagion risk during the Terra collapse. That crisis was predictable because the math was wrong. Here, the math is also wrong. I will break down the evidence:
- Revenue vs. Valuation: The article's $120B number is statistically impossible as revenue. If OpenAI and Anthropic had $120B in revenue, their sequential growth would imply total AI market size exceeding $1 trillion by 2027. But industry data from Gartner and IDC shows the global AI software market was $62 billion in 2024. The claim would mean these two companies alone command 193% of the market—a logical impossibility. The only number that fits is their combined valuation. This is not a minor error; it is a categorical misrepresentation.
- Comparison to Traditional Giants: Starbucks and McDonald's have decades of stable, audited revenue. Starbucks sells 100 million cups of coffee per day. McDonald's serves 69 million customers. AI companies sell software subscriptions and API calls. Comparing them is like comparing aircraft carriers to speedboats. But even if we accept the cross-industry comparison, the revenue data fails. In 2024, Starbucks' revenue was $40.2 billion, McDonald's $25.8 billion. OpenAI and Anthropic combined at 10% of that. The headline intended shock, not accuracy.
- The Crypto Media Pattern: In 2024, I analyzed the Spot Bitcoin ETF filings from top asset managers. I saw how regulatory data could be twisted to support bullish narratives. Here, Crypto Briefing is using AI hype to attract attention to its crypto-focused audience. The implicit message: AI is so profitable that its associated crypto projects (like decentralized compute networks) must also succeed. This is a classic pump mechanism. During the 2021 NFT frenzy, similar articles claimed artists were making millions, but on-chain data showed wash trading. The pattern repeats: create a false ceiling, then sell the ladder.
- Quantitative Risk Framing:
- Cost of Revenue: For AI companies, inference costs are 30-50% of revenue. If revenue were $120B, inference costs would be $36-60B annually. That would require 2-3 million H100 GPUs at $30,000 each, totaling $60-90B in hardware alone—more than NVIDIA's entire 2024 datacenter revenue of $47.5 billion. This is physically impossible.
- Profit Margins: OpenAI is estimated to lose $5 billion in 2024 (The Information). Anthropic is unprofitable. If revenue were $120B, they would be the most profitable companies in history. The data reveals a high-burn, high-growth phase, not a revenue giant.
- Market Cap vs. Revenue: If we accept the $120B revenue claim, their P/S ratio would be 2x, implying they are undervalued. But with $4.7B real revenue, their P/S is 40x. The market is already pricing in future growth, not current revenue. The article would mislead investors into overvaluing these companies.
- First-Person Technical Experience: I have audited over 50 crypto projects since 2017. The 2017 ICO audit I did uncovered two tokenomics models that guaranteed inflation. The issuers had copied equations from a Medium post without verification. Here, the source likely copied a number from a different context without verification. I have seen this exact behavior in smart contract audits: a variable named 'totalSupply' is sometimes mistakenly assigned to 'balanceOf' a contract. The same cognitive shortcut happens in journalism: 'valuation' becomes 'revenue.'
Contrarian: Correlation Is Not Causation
It is tempting to dismiss the entire article as noise. But the contrarian angle is that the error itself reveals a deeper signal. Crypto media outlets are increasingly desperate for narratives to sustain bull market excitement. In a bull market, as we are now, euphoria masks technical flaws. This article is a perfect example: it uses a data point that sounds impressive but collapses under scrutiny.
However, the truth is not simply that the article is wrong. The truth is that the AI industry is real, and its economic impact is growing. But conflating that with crypto narratives is dangerous. During the 2022 bear market, I saw how projects that linked themselves to Terra's success suffered when Terra collapsed. Now, AI+crypto projects are attracting capital based on the assumption that AI companies are printing money. They are not. The revenue numbers for AI companies are modest compared to traditional enterprises. The hype may lead to a bubble in AI+crypto tokens, which will burst when the gap between narrative and reality becomes undeniable.
Another blind spot: the article ignores the role of infrastructure providers. NVIDIA's datacenter revenue in 2025 exceeded $100 billion—far more than OpenAI and Anthropic combined. The real value capture is in hardware, not applications. Crypto investors who chase AI tokens may miss the fact that the true winners are chipmakers and cloud providers, which are not on-chain. The narrative that crypto will disrupt AI compute is undercut by the sheer scale of centralized infrastructure.
Takeaway: The Next Signal
The next time a crypto media outlet claims that a technology company has surpassed a traditional giant, verify the numbers yourself. Look for audited financials, on-chain data, or official filings. The $120 billion illusion will persist as long as readers accept headlines without scrutiny. In a bull market, survival is the ultimate alpha. Trust the math, ignore the hype. The real question is not whether AI is growing—it is—but whether the crypto market can digest this misinformation without creating new risks. I will be watching for the next fabricated revenue claim, and I suggest you do the same.
Ledgers do not lie, only the narrative does. Trust the math, ignore the hype. Survival is the ultimate alpha in a bear.
