The $115B ARR Mirage: What the OpenAI-Anthropic Math Really Tells Us

Two companies now claim a combined annual recurring revenue north of $115 billion. That is the headline. It is also a mathematical impossibility. The number, attributed to OpenAI and Anthropic, paints a picture of two AI labs closing in on Microsoft’s commercial cloud revenue. The reality is far less dramatic—and far more instructive. As someone who has spent years dissecting protocol economics and liquidity flows, I have learned that the most dangerous data is the one that feels inevitable. This figure does not feel inevitable. It feels manufactured. Let me break down why that matters, and what it signals about the AI market structure in 2026. The original report, published by Crypto Briefing, provided no source, no methodology, and no breakdown between the two companies. It simply stated that OpenAI and Anthropic’s combined ARR exceeded $115 billion, approaching Microsoft’s scale. For context, Microsoft’s entire commercial cloud segment—Azure, Office 365, Dynamics—generated roughly $160 billion in annualized revenue in 2024. Are we to believe that two private AI startups, with a combined employee count under 5,000, have achieved nearly 70% of that figure? The public record tells a different story. OpenAI’s annualized revenue was estimated at $3.7 billion in 2024. Anthropic’s was around $1 billion. Together, that is $4.7 billion—about 4% of the reported number. Code is law, but liquidity is truth. In this case, the liquidity of the data itself is suspect. Let me walk through the arithmetic, because it matters. A combined ARR of $115 billion implies a price-to-sales ratio of roughly 1.3x if we accept the implied valuation of $150 billion. That is absurdly low for two companies growing at triple-digit rates. Even established SaaS giants trade at 6-10x revenue. If the true ARR were even one-tenth of the claimed figure, OpenAI and Anthropic would be the most undervalued assets in the history of private markets. They are not. The more plausible explanation: the original report confused total contract value with annual recurring revenue, or simply imported a speculative projection as a fact. This is not a rounding error. It is a category error. Panic sells, logic buys. In data analysis, skepticism is the buy signal. The narrative does serve a purpose, though. It feeds the idea that AI-native companies are displacing traditional incumbents. That framing is emotionally satisfying, but structurally wrong. OpenAI and Anthropic are not a unified bloc. They are fierce competitors, fighting over the same enterprise customers, the same talent pool, and increasingly, the same compute resources. Bundling their revenue into one mega-number obscures the real dynamics: Microsoft captures a significant portion of AI value through its Azure partnership with OpenAI, while Anthropic differentiates on safety and model alignment. The real competitive story is not David versus Goliath. It is Goliath versus a smaller Goliath, with nuanced constraints in between. From my background auditing smart contracts and evaluating liquidity mechanisms, I see a parallel between inflated ARR claims and inflated total value locked figures in DeFi. Both are used to attract capital, to signal momentum, and to justify valuations. Both can be gamed. In 2018, I audited 0x protocol v2 and identified seven critical reentrancy vulnerabilities. The code was elegant. The liquidity was not. That experience taught me a simple rule: when a number is too clean, too round, too convenient, it usually hides the mess. The $115 billion figure is clean, round, and convenient. It aligns perfectly with the AI hype cycle. It aligns perfectly with the crypto media’s desire to connect AI narrative to crypto liquidity. It does not align with the audited realities of enterprise revenue recognition. What does the real data suggest? OpenAI’s ARR is growing, but primarily through enterprise API usage and ChatGPT subscriptions. Anthropic’s growth is smaller, with its Claude models gaining traction among developers who prioritise alignment. Neither company is close to Microsoft’s cloud scale. Microsoft’s AI-related revenue—through Azure OpenAI services, GitHub Copilot, and Microsoft 365 Copilot—is substantial, but it is embedded within a diversified portfolio. The comparison the original article makes is like comparing the revenue of a Formula 1 team to that of a car manufacturer. Yes, both work with engines. No, they are not in the same economic universe. The more instructive question is not whether the ARR figure is real. It is why such a figure would be published. Crypto Briefing’s audience primarily consists of crypto investors, many of whom are looking for the next macro narrative. AI is the largest narrative outside crypto. Fusing the two—by suggesting AI companies are approaching Microsoft—creates a psychological bridge. It implies that AI and crypto are on parallel trajectories of exponential growth. That implication is misleading. The AI market has real fundamentals, but they are not yet at the scale the report suggests. Liquidity dries up when trust breaks. If macro markets ever realise that AI revenue growth is slower than projected, the correction will be sharp. Data speaks louder than sentiment, but sentiment currently speaks loudest. Let me offer a contrarian angle. The original report is probably wrong, but it is wrong in a useful direction. AI infrastructure is genuinely undervalued in relation to the actual compute demands of even the current AI workloads. If OpenAI and Anthropic grow at their true pace—say, doubling annually—their compute needs will outpace supply. That creates a structural opportunity not in AI software companies, but in the physical layer: data centres, power generation, optical networking. The hype around $115 billion in ARR distracts from a more defensible investment thesis: the scarcity of physical resources needed to run the models. I saw the same pattern in DeFi. Everyone chased the highest APY. The real winners were the infrastructure providers who charged fees, not the farmers who took on impermanent loss. For traders, the actionable insight is to ignore the headline and track the signals. OpenAI’s API volume, Anthropic’s enterprise contract wins, and the capacity expansion of tier-1 cloud providers are all measurable. If you want to validate the AI commercialisation story, follow the compute procurement numbers. They are harder to fake than an executive presentation. The lesson from my 2022 experience is relevant here. I faced a $200,000 drawdown when the bear market broke. I did not panic. I deleveraged, converted volatile assets into stablecoins, and bought ETH at $800 during the trough. The same discipline applies to AI narratives. When media reports are extreme, the correct response is to do nothing based on them. Wait for the primary source. Then act. The implications for crypto are indirect but meaningful. If AI revenue growth disappoints, the AI-crypto crossover tokens—decentralised compute networks, data provenance projects—will suffer. If AI revenue beats expectations, the same tokens will likely rally, but not on their own merits. They will ride the AI sentiment wave. As someone who has audited the underlying technology of dozens of DeFi protocols, I can tell you that the correlation between narrative and token price is often stronger than the correlation between fundamentals and token price. That is not an invitation to speculate. It is a warning to prepare. In the long run, the market will correct the record. OpenAI and Anthropic will eventually publish audited financials, or they will remain private and allow the estimate to persist. If the official numbers arrive within the next two quarters, the gap between the $115 billion claim and the reality will create a sentiment shock. Treat that shock as an opportunity, not a threat. Panic sells, logic buys. When everyone realises the emperor has no clothes, the smart money will already be purchasing the assets that benefit from the disappointment—infrastructure, efficiency technologies, and companies that monetise AI without relying on fantasy ARR multiples. Consider this article a checklist. Do not trust single-source narratives, especially when they come from a crypto publication that has no direct reporting relationship with either company. Demand footnotes. Demand segment breakdowns. Demand a timeline. If the information is not there, treat it as noise. I have applied this standard to every protocol audit I have conducted. I have avoided nine-figure losses by dismissing convenient narratives. It is the trader’s version of due diligence, and it has never failed me. Data speaks louder than sentiment, but verified data speaks loudest of all. The next time a headline claims that two private companies have created a combined revenue figure that would place them among the world’s largest technology firms, check the underlying arithmetic. If the revenue per employee defies every known productivity benchmark, it is not revolutionary. It is probably fictional. And in a market that has been humbled repeatedly by unreasonable expectations, the best trade is to stand aside until the truth emerges. The $115 billion ARR story is not an investment thesis. It is a sentiment indicator. And sentiment, as always, is a crude tool. The sharpest traders know that patience is the greatest leverage.