A single line in a Crypto Briefing article claims Anthropic and OpenAI have a combined ARR of $115B. My first reaction was not surprise, but suspicion. I've seen too many DeFi projects inflate TVL to trust unaudited numbers. The math doesn't add up.
Verification is the only trustless truth.
Let me state the obvious: this data is an outlier. Publicly available estimates from The Information, Bloomberg, and even Microsoft's own filings suggest OpenAI's ARR for 2024 is around $3.7B to $4.5B. Anthropic, with its smaller market share, is likely in the $1B to $1.5B range. Combined, that's roughly $5B to $6B. The Crypto Briefing article claims $115B—a factor of 20x higher.
This is not a rounding error. This is a fundamental break with reality. As a researcher who has spent years auditing smart contracts and financial primitives, I know that when a number deviates that far from consensus, the burden of proof is on the claimant. The article provided zero proof. No source, no methodology, no breakdown. It's a ghost number.
Context: The Genesis of the Anomaly
Crypto Briefing is a media outlet that traditionally covers blockchain and cryptocurrency. Its pivot to AI revenue data is part of a broader trend: merging the hype cycles of AI and crypto. The article's original framing—"Anthropic and OpenAI’s combined ARR tops $115B, closing in on Microsoft"—is designed to shock. It positions two unlisted AI startups as a collective rival to the largest cloud provider. But the comparison is flawed. Microsoft's commercial cloud ARR (Azure, Office 365, Dynamics) is about $160B, and its total revenue exceeds $200B. The $115B claim would mean these two startups are 72% of Microsoft's cloud business—a scenario that defies logic given their headcount, infrastructure, and market penetration.
The article's lack of technical detail is telling. There is no discussion of how ARR is calculated—whether it's annualized recurring revenue from subscriptions, total contract value, or something else. For AI companies, ARR can be inflated by prepaid enterprise contracts, government grants, or even revenue from API credits that are consumed slowly. But even the most generous interpretations cannot bridge the gap to $115B.
Core: Code-Level Verification of the Data
I approached this as I would a smart contract audit: break the claim into its components, test each against known constraints, and identify failure modes.
First, the revenue per employee. OpenAI has roughly 1,500 employees. Anthropic has about 500. Combined, that's 2,000 people. To achieve $115B ARR, each employee would need to generate $57.5M in revenue per year. Compare this to Microsoft, which has 220,000 employees and generates $211B in revenue—about $960K per employee. The AI startups would need to be 60x more productive than the most profitable software company on earth. That's not plausible.
Second, the addressable market. The global AI software market is projected to be around $100B in 2024 (Gartner, IDC estimates). Claiming that two companies capture 115% of the market is nonsense. Even if we consider API services and infrastructure, the total TAM for generative AI is still under $50B. The claim implies a monopoly that doesn't exist.
Third, the growth rate. If OpenAI's ARR was $3.7B in 2024, reaching $115B would require a 30x increase in one year, or a 3000% growth rate. Even OpenAI's actual growth—from around $1.6B in 2023 to $3.7B in 2024—is impressive but not exponential to that degree. The number is detached from any observable trajectory.
I cross-referenced with Microsoft's own AI revenue. Azure AI services grew over 100% in 2024, but the absolute contribution to Microsoft's cloud revenue is still under $5B. The idea that two startups have 20x that is absurd.
Silence in the code speaks louder than hype.
The article provides no data source. In my work auditing DeFi protocols, I treat any unverifiable claim as a vulnerability. Here, the vulnerability is the reader's trust. The publisher may be deliberately inflating numbers to attract attention to their brand, or to create a narrative that AI is booming so that crypto projects can attach themselves to the hype. Either way, the data is not just wrong—it's dangerous if used for investment decisions.
Contrarian: The Blind Spot of Crypto Media's AI Fetish
The counter-intuitive angle is that the article's intent is not to inform, but to signal. Crypto media is desperate for relevance. The crypto market has been in a sideways trend for months, and AI is the only sector generating consistent hype. By publishing a shocking number, Crypto Briefing positions itself as a bridge between AI and crypto, hoping to attract a new audience of AI investors. But the lack of verification is a security flaw.
Consider the incentives. Crypto Briefing is a for-profit media outlet. Its revenue comes from advertising, sponsored content, and possibly token promotions. An article with a $115B headline drives clicks. It doesn't need to be accurate. As a researcher, I see this as a pattern: the same actors who inflated ICO valuations in 2017 are now inflating AI revenue numbers. The underlying mechanism is identical—replace "TVL" with "ARR."
Furthermore, the article's framing of "closing in on Microsoft" is a deliberate misdirection. By combining two competing companies, it creates a false sense of collective strength. In reality, OpenAI and Anthropic are rivals. They compete for the same enterprise customers, the same talent, and the same venture capital. Their combined ARR is irrelevant; what matters is their individual trajectories. And even then, neither is close to a $100B+ revenue run rate.
The blind spot is that many readers will take the headline at face value. They will share it on social media, reinforcing the narrative. The damage is done before verification can occur. This is a classic pump-and-dump of information.
Takeaway: Vulnerability Forecast
I predict that we will see more such data anomalies as AI and crypto converge. The lack of regulatory oversight means anyone can publish unverified numbers. The market will eventually correct when actual earnings reports are released, but by then, the hype cycle may have already moved capital into overvalued assets.
My advice: treat every unaudited ARR number as a potential bug. Require source code—or at least a reputable data provider—before accepting any claim. In the absence of verification, the null hypothesis is that the data is false.
I trust the null set, not the influencer.
This is not a criticism of AI or crypto. It's a call for rigor. The same intellectual honesty that drives zero-knowledge proofs should apply to financial data. If we accept a $115B ARR without proof, we are no better than the investors who trusted unaudited smart contracts and lost everything.
When the hype cycle peaks, will you be the one holding the bag of unverified claims?