Last week, Crypto Briefing dropped a headline that made my coffee go cold: “Anthropic and OpenAI’s combined ARR tops $115B, closing in on Microsoft.”
I read it three times. Then I laughed. Then I got angry.
Not because I doubt the potential of AI. I’ve spent years watching decentralized protocols struggle for adoption while AI companies burn through capital like it’s Monopoly money. But $115 billion in annual recurring revenue for two companies that, by my math, are still fighting for positive unit economics? That’s not a data point. That’s a fantasy dressed up in a clickbait headline.
Let me be clear: this isn’t an attack on Anthropic or OpenAI. It’s an attack on the culture of unverifiable claims that infects both crypto and AI. And as someone who built a career auditing blockchain whitepapers and then moved into decentralized protocol management, I’ve developed a nose for numbers that don’t pass the smell test. This article is my attempt to deconstruct this particular mirage, and to argue why blockchain’s promise of verifiable truth is the antidote we need.
The Hook: A Number That Breaks the Laws of Physics
The article states that Anthropic and OpenAI’s combined ARR exceeds $115 billion. To put that in perspective: Microsoft’s entire commercial cloud business (Azure + Office 365 + Dynamics 365) generated about $160 billion in revenue in fiscal 2024. OpenAI alone, according to every reputable source from The Information to Bloomberg, had an ARR of roughly $3.7 billion in 2024. Anthropic’s ARR was estimated at around $1 billion. That’s $4.7 billion combined. Not $115 billion.
Even if the article meant $11.5 billion (a common typo when mixing billions and hundreds of billions), that’s still more than double the best available estimates. And the article provides zero sources, zero methodology, zero breakdown.
In my days as a junior copywriter for a Baltic ICO platform in 2017, I audited over 40 whitepapers. I found that 80% of them lacked economic viability. The pattern was always the same: a big number, a vague promise, and no traceable source. This $115B claim triggers the same alarm bells.
But here’s the twist: the article isn’t just wrong. It’s dangerous. Because it fuels a narrative that AI is already a trillion-dollar industry, which in turn justifies sky-high valuations, insane capital burn, and a regulatory race to the bottom. And it does so without any of the transparency that blockchain has fought to establish.
Context: The Decentralization Philosophy Meets Data Integrity
Core to blockchain’s ethos is the idea that trust should be minimized. We don’t believe a bank’s balance sheet because we trust the bank; we believe it because we can verify the transactions on-chain. This philosophy extends to data. When a protocol claims a $100 million TVL, we can check the smart contracts. When a DAO reports revenue, we can audit the treasury.
But the AI industry operates in a black box. OpenAI and Anthropic are private companies. Their financials are disclosed selectively, often through leaked documents or analyst estimates. The $115B figure is a perfect example of a data point that cannot be verified, yet it gets shared, retweeted, and used as a basis for investment decisions.
As a decentralization evangelist, I see this as a fundamental failure of the existing information ecosystem. The same problem that plagued ICOs in 2017—unverifiable claims propped up by hype—is now infecting the AI narrative. And if we, as crypto natives, don’t call it out, we risk losing the moral high ground we’ve fought for.
Core: Deconstructing the $115B Claim Through a Crypto Auditor’s Lens
Let me walk through why this number is so implausible, using the same framework I developed for whitepaper reviews.
1. Revenue per employee
OpenAI has roughly 3,000 employees. Anthropic has about 1,500. Combined, that’s 4,500 people. To generate $115 billion in ARR, each employee would need to produce $25.6 million in revenue. For comparison, Microsoft’s revenue per employee is about $1.1 million. Apple’s is about $2.5 million. The most efficient tech companies (like Meta) top out at $1.6 million per employee. $25.6 million per employee is not just unrealistic—it’s physically impossible without every employee being a hedge fund.
2. API pricing and volume
OpenAI’s API pricing is public. GPT-4 costs about $0.03 per 1K tokens of input and $0.06 per 1K tokens of output. Even if we assume the most generous usage scenarios, the total token volume required to reach $115 billion in revenue would dwarf the entire internet’s data output. For context, all of Google’s search revenue is about $250 billion. OpenAI would need to process 46% of Google’s search volume through its API—at higher prices. That’s absurd.
3. Enterprise adoption is real, but not that real
I’ve spoken with enterprise clients in my role as a protocol PM. The majority of AI spending is still experimental. Companies are buying $1 million to $10 million in API credits, not $100 million. Even the biggest enterprise deals, like Microsoft’s investment in OpenAI, are structured as cloud credits, not straight ARR. The $115B figure would require every Fortune 500 company to spend an average of $230 million annually on AI APIs. That’s more than their entire IT budgets in many cases.
4. The comparison to Microsoft
The article claims the combined ARR is “closing in on Microsoft.” But Microsoft’s AI revenue (including Azure AI, GitHub Copilot, and Microsoft 365 Copilot) is already embedded in its cloud numbers. The comparison is apples to oranges. Microsoft’s “AI” revenue is a fraction of its total cloud revenue, and the $115B figure would put OpenAI+Anthropic at 70% of Microsoft’s total commercial cloud. That’s not “closing in”—that’s a different universe.
Based on my experience auditing ICO whitepapers, I’ve learned to spot the telltale signs of inflated numbers: lack of granularity, absence of sources, and comparisons that feel too good to be true. This article has all three.
Contrarian: But What If the Data Were True?
Let’s play devil’s advocate. Suppose the $115B figure is real—perhaps it includes future commitments, contract values, or some accounting trick that I’m not aware of. What would that mean for the industry?
First, it would mean that AI companies are generating revenue at a pace that surpasses every software company in history. That would justify the massive capex in GPUs and data centers. It would also mean that the AI arms race is heating up, and that traditional cloud providers are at risk of being disrupted.
Second, it would imply that the AI market is already mature, not nascent. That would be a problem for investors looking for growth. If OpenAI and Anthropic are already at $115B combined, where does the next 10x come from? The market would be saturated, and margins would compress.
Third, it would raise serious questions about antitrust. Two companies controlling $115 billion in a market that’s dominated by Microsoft? Regulators would have a field day.
But here’s the rub: even if the data were true, the lack of transparency would still be a problem. Without verifiable on-chain data, we can’t trust the narrative. And that’s the core insight: the problem isn’t just the number—it’s the system that allows numbers to be asserted without proof.
In blockchain, we have a term for this: “trust but verify.” The $115B claim fails the verification test. And until AI companies start publishing their revenue on-chain—or at least via audited, transparent reports—we should treat every headline with skepticism.
Takeaway: The Future of Trust Is Cryptographic
This article isn’t about whether Anthropic or OpenAI will succeed. It’s about the information hygiene of an industry that’s still in its infancy. As a decentralization evangelist, I believe that the same principles that make blockchain valuable—immutability, transparency, verifiability—should apply to all data that drives capital allocation.
Imagine a world where every AI company’s revenue is published as a verifiable claim on-chain, with cryptographic proofs from auditors. Where investors can query the smart contract to see the exact ARR, split by product line, without relying on a single tweet from a crypto media outlet. That’s the vision I’m working toward.
Until then, we need to be vigilant. The $115B mirage is a reminder that hype is a compiler for bad decisions. But debate—rigorous, evidence-based debate—is the compiler for better consensus. Let’s use it.
True ownership begins where the server ends. And true knowledge begins where the data is verifiable.