The $115 Billion Mirage: Dissecting the ARR Claim That Defies Mathematics

The data appears on a single line, buried in a crypto news brief. Anthropic and OpenAI's combined annual recurring revenue has surpassed $115 billion, closing in on Microsoft. No source. No breakdown. No methodology. The ledger does not lie, but it forgets. And in this case, the ledger was never consulted. I have spent 27 years in this industry, and I have learned to treat unsourced numbers the way a bank treats unsigned checks. They are not currency. They are promises, and promises require collateral. The collateral here is absent. What remains is a narrative, and narratives are cheap. The question is not whether the story is compelling. The question is whether the numbers survive contact with reality. Let us begin with what is publicly verifiable. OpenAI reported an annualized revenue run rate of approximately $3.7 billion in 2024. Anthropic's figure hovered near $1 billion. The sum is roughly $4.7 billion. That is the baseline. That is the number that emerges from audited financial documents, investor communications, and the reporting of established financial outlets like Bloomberg and The Information. The $115 billion claim is 24 times higher. It is not a rounding error. It is not a methodological discrepancy. It is a different universe of numbers. I have audited tokenomics and smart contract logic since the ICO era of 2017. I have seen what happens when projects inflate their metrics to attract capital. The pattern is always the same: the numbers become aspirational, then they become fictional, then the fiction becomes the foundation for investor decisions. This is not a new phenomenon. It is the oldest trick in the financial playbook, and it is being deployed here with remarkable audacity. The source is Crypto Briefing, a publication that serves the cryptocurrency investment community. The article contains no technical analysis, no customer data, no growth metrics, no retention figures. It is a single claim attached to a comparative headline. The headline is designed to provoke. The question is whether it is designed to inform. I suspect the former, and I have reason to be suspicious. In 2021, I traced the wallet history of an NFT collection and discovered that its deployer was linked to three banned addresses associated with money laundering. The collection's origin story was fabricated. The floor price dropped 40% within a week. The lesson was simple: verify provenance before you endorse value. Let us apply the same standard here. The provenance of this $115 billion figure is unknown. It has no paper trail. It has no audit trail. It has no institutional backer. It is a number floating in the digital ether, and it is being used to construct a narrative about the AI industry's trajectory. That is not journalism. That is marketing. The first dimension of analysis is technical. The original article contains zero technical content. No mention of model architecture, training methodology, or data engineering. This is a significant omission for a publication that claims to be reporting on the AI industry. The technology is the product. The technology is the moat. The technology is the reason customers pay for these services. Without technical context, the ARR figure is meaningless. It is a number without a mechanism. I have spent years dissecting DeFi protocols, and I have learned that the mechanism is everything. A yield figure without a mechanism is a trap. An APY without an understanding of the token emission schedule is a promise that will be broken. The same logic applies here. What is driving this alleged revenue growth? Is it new model capabilities attracting new customers? Is it price reductions creating volume? Is it enterprise contracts that were signed but not yet delivered? The article provides no answers because the article provides no questions. The second dimension is commercialization. The claim of $115 billion in combined ARR implies that OpenAI and Anthropic are generating annualized revenue equivalent to roughly 70% of Microsoft's entire commercial cloud business. Microsoft's commercial cloud revenue was approximately $160 billion in 2024. This includes Azure, Office 365, LinkedIn, and Dynamics 365. It is a massive, diversified portfolio of enterprise software products. The idea that two AI startups, with a combined workforce of approximately 5,000 employees, are generating $115 billion in annualized revenue is not just improbable. It is arithmetically absurd. I analyzed the Terra-Luna collapse in 2022 using reserve audits from 2019 to 2021. The discrepancies in the reported burn rates were consistent. The peg maintenance mechanism was mathematically unstable under stress. The death spiral was predictable. I did not need to speculate. The numbers told the story. The same approach applies here. The $115 billion figure is not supported by any publicly available data. It contradicts the revenue figures that have been reported by credible sources. It is, to use the technical term, a fabrication. There is a possibility, however, that the number is not entirely fabricated. It may be a misunderstanding. The original article may have confused the combined ARR of the two companies with a different metric. For example, OpenAI's projected revenue for 2025 has been estimated at approximately $11.6 billion. Anthropic's projected revenue has been estimated at approximately $2 billion. The sum is approximately $13.6 billion. This is still far below $115 billion, but it is closer. It is possible that the author of the original article saw a projection for OpenAI's 2025 revenue, added a similar figure for Anthropic, and then multiplied by something. The result would be a number that is both plausible and false. This is a common error in financial reporting. It is the difference between ARR and contract value. It is the difference between recognized revenue and committed revenue. It is the difference between what is earned and what is promised. I have seen this confusion in the crypto space repeatedly. Projects announce a partnership, and the market interprets it as revenue. Projects announce a pilot, and the market interprets it as adoption. The gap between announcement and reality is where the noise lives. The third dimension is industry impact. Even if the specific figure is false, the broader trend is real. AI companies are experiencing rapid revenue growth. OpenAI's revenue has been doubling year over year. Anthropic's revenue has been growing even faster from a smaller base. This growth is being driven by genuine enterprise demand. Companies are paying for AI-powered tools because they are seeing measurable returns. This is not a speculative narrative. This is observable behavior. However, the speed of this growth does not match the speed of the narrative. The article implies that AI-native companies are on the verge of displacing traditional software giants. The data suggests otherwise. Microsoft, Google, and Amazon are all generating substantial AI-related revenue from their cloud businesses. Azure AI is growing at over 100% annually. Google Cloud is growing at over 30%. These are established businesses with massive sales forces, existing customer relationships, and proven distribution channels. The AI-native companies are growing faster, but they are growing from a much smaller base. The competitive dynamics are more complex than the article suggests. OpenAI and Anthropic are not allies. They are competitors. They compete for customers, for talent, and for attention. The article's decision to combine their revenue figures is a rhetorical device. It creates the illusion of a unified challenger to Microsoft's dominance. The reality is that OpenAI has a deep partnership with Microsoft. Microsoft is OpenAI's largest investor and its exclusive cloud provider. The relationship is symbiotic. OpenAI provides the models, and Microsoft provides the distribution. They are not adversaries. They are partners. Anthropic, on the other hand, has positioned itself as the safety-focused alternative. The company has attracted investment from Google and Amazon, and it has differentiated itself through its commitment to AI safety. This is a legitimate competitive strategy, but it is not the same as being part of a unified front against Microsoft. The two companies have different business models, different customer bases, and different strategic priorities. Combining their revenue figures obscures more than it reveals. The fourth dimension is ethics and safety. The original article contains no mention of AI safety, alignment, or regulation. This is a significant omission. The AI industry is facing intense scrutiny from regulators, policymakers, and the public. The ethical implications of large language models are a matter of ongoing debate. The article's focus on revenue, to the exclusion of all else, is a symptom of a broader problem in tech journalism. The financialization of the AI narrative has overshadowed the substantive questions about the technology's impact on society. I have seen this pattern before. In the DeFi boom of 2020, the focus was on yield. The mechanisms were secondary. The risks were ignored. The result was predictable. The protocols collapsed, and the investors who had not done their due diligence lost everything. The same dynamic is playing out in the AI space. The focus is on ARR, on valuation, on market share. The questions about safety, about alignment, about the societal impact of these systems, are being pushed to the margins. The fifth dimension is investment and valuation. If the $115 billion figure were accurate, the combined valuation of OpenAI and Anthropic would be in the range of $1.5 trillion, assuming a 10x price-to-sales multiple. This is more than the current valuation of any technology company in the world. It is more than Microsoft, more than Apple, more than NVIDIA. It is a number that defies rational analysis. The actual valuations are more grounded. OpenAI was last valued at approximately $150 billion. Anthropic was valued at approximately $40 billion. The combined valuation is approximately $190 billion. This is substantial, but it is not extraordinary. It is consistent with the revenue figures that have been reported. The price-to-sales multiple is approximately 40x, which is high but not unprecedented for high-growth technology companies. Investors should be wary of data from non-mainstream sources. The crypto media ecosystem has a demonstrated history of exaggerating figures to attract attention and drive investment. This is not a new phenomenon. It is a structural feature of the industry. The incentives are misaligned. The publications that generate the most attention are the ones that tell the most dramatic stories. The stories that are the most dramatic are the ones that are the least accurate. The sixth dimension is infrastructure and compute. The original article contains no information about the compute resources required to support the alleged revenue. This is a significant omission. AI companies are among the largest consumers of computing power in the world. OpenAI and Anthropic spend billions of dollars annually on GPUs, cloud services, and data center capacity. The cost of compute is a major constraint on their growth. If the $115 billion figure were accurate, it would imply an enormous investment in compute infrastructure. It would imply that the two companies are spending tens of billions of dollars annually on hardware and cloud services. It would imply that they are among the largest customers of NVIDIA and the major cloud providers. The market for AI hardware would be even more overheated than it already is. The supply chain would be under even more pressure. The reality is more measured. OpenAI and Anthropic are spending significant amounts on compute, but they are not spending tens of billions of dollars annually. Their compute costs are a substantial portion of their operating expenses, but they are not the dominant factor. The companies are working to improve their cost structures through model optimization, infrastructure improvements, and, in some cases, custom silicon. The trajectory is toward lower costs per unit of compute, not higher. The contrarian angle is worth considering. The bulls on this story would argue that the revenue growth is real, even if the specific figure is inflated. They would argue that the trend is what matters, not the exact number. They would point to the rapid adoption of AI-powered tools by enterprises, the increasing willingness of customers to pay for AI services, and the expanding use cases for large language models. They would argue that the AI industry is on a trajectory that will eventually justify even the most aggressive projections. There is some merit to this argument. The growth of the AI industry has been remarkable. OpenAI's revenue has grown from essentially zero in 2020 to over $3 billion in 2024. Anthropic's revenue has grown from zero in 2021 to over $1 billion in 2024. The growth rates are extraordinary. The question is whether they are sustainable. The question is whether the market can continue to absorb the supply of AI services at current prices. The question is whether the competition will drive prices down to a level that makes the current revenue figures unsustainable. I have seen this pattern before. In the ICO boom of 2017, the growth rates were extraordinary. The projects were raising millions of dollars in hours. The valuations were soaring. The enthusiasm was unbounded. And then the market corrected. The projects that had real technology and real customers survived. The projects that had only narratives and promises collapsed. The same dynamic will play out in the AI industry. The companies with real technology and real customers will thrive. The companies with only narratives will fail. My analysis of the Terra-Luna collapse taught me that the mechanism matters more than the narrative. The peg maintenance mechanism was mathematically unstable. The narrative was that the algorithm would hold the peg. The reality was that the algorithm was a death spiral waiting to happen. The same logic applies to the AI industry. The revenue models are the mechanisms. The growth rates are the narratives. The question is whether the mechanisms can support the narratives. The answer, in this case, is no. The $115 billion figure is not supported by any observable data. It is a narrative without a mechanism. It is a promise without collateral. It is a number that will not survive contact with reality. What should the industry track instead? The short-term signal is whether OpenAI or Anthropic respond to the $115 billion claim. A denial would be notable. A silence would be more notable. The medium-term signal is the actual revenue figures disclosed in fundraising documents or investor communications. The long-term signal is the growth of enterprise AI spending relative to traditional software spending. If AI spending continues to grow at current rates, the AI industry will eventually reach a scale that justifies some of the more aggressive projections. If the growth stalls, the current valuations will look increasingly stretched. I have been writing about this industry for 27 years. I have seen booms and busts. I have seen technologies that changed the world and technologies that disappeared without a trace. The pattern is always the same. The narrative leads, the data follows, and the truth eventually emerges. The $115 billion claim is a narrative. The data will follow. The truth will emerge. The question is whether investors will wait for the truth or act on the narrative. The takeaway is not that AI is a bubble. The takeaway is that specific claims should be treated with skepticism until they are verified. The takeaway is that the ledger does not lie, but it forgets. And in this case, the ledger was never consulted. I would advise readers to focus on the fundamentals. Look at the actual revenue figures. Look at the customer adoption rates. Look at the retention rates. Look at the gross margins. Look at the cash burn. These are the numbers that matter. These are the numbers that will determine which companies survive and which companies fail. The rest is noise. The $115 billion claim is noise. It is a data point without a source, a number without a mechanism, a narrative without a foundation. It should be ignored. What should not be ignored is the underlying trend. The AI industry is growing rapidly. The growth is real. The question is whether the growth can be sustained. That is a question that can only be answered by data, not by narratives. The final thought is this: the next time you see a dramatic claim about AI revenue, ask for the source. Ask for the methodology. Ask for the breakdown. Ask for the audit. If the answer is silence, treat the claim with the skepticism it deserves. The ledger does not lie, but it forgets. And in this case, the ledger was never consulted. The numbers will tell the truth eventually. They always do.