The number appeared without warning, without the customary press release, without the corroborating footnote. Crypto Briefing, a publication whose editorial focus historically orbits digital assets rather than enterprise software, reported that Anthropic and OpenAI have achieved a combined Annual Recurring Revenue of $115 billion. The figure, if accurate, would place these two private companies in rarefied air, alongside the most successful software enterprises in history. But my training, forged in the crucible of cryptographic verification, demands a pause. The claim is not self-authenticating. The source is not a mainstream financial institution. The data, as presented, lacks the granularity required for forensic validation. This is not a dismissal; it is a starting point for a systematic teardown.
The AI industry has entered what market participants call the 'scaling era.' The narrative is seductive: models grow, capabilities expand, and revenue follows in lockstep. The reported $115 billion ARR figure, if taken at face value, would represent a growth trajectory that outpaces the historical benchmarks of the SaaS sector. Salesforce, the archetypal cloud software giant, reported roughly $37.5 billion in revenue for its fiscal year 2024. Microsoft's commercial cloud business, a sprawling portfolio of products, generates over $100 billion annually. The implication of the Crypto Briefing report is that two companies, founded within the last decade, have collectively matched or exceeded the commercial output of these established titans. The claim is not impossible, but it is extraordinary. And extraordinary claims, as the principle of evidentiary standards dictates, require extraordinary proof. The proof, in this case, is conspicuously absent.
My analysis of this report must begin with a fundamental acknowledgment: the data is a single point, sourced from a non-primary outlet, with no official confirmation from either organization. The confidence level for any conclusion drawn from this figure must be graded accordingly. I assign a 'C' rating, indicating that the core data point is plausible but unverified. The subsequent analysis, therefore, is an exercise in scenario modeling based on industry heuristics, not a statement of established fact. The distinction is critical for readers who may be tempted to treat a headline as a balance sheet.
The Commercialization Conundrum: Revenue Quality vs. Revenue Quantity
The reported figure, assuming a split consistent with historical funding rounds and market positioning, would place OpenAI at approximately $80 billion ARR and Anthropic at $35 billion. These are staggering numbers. They suggest that the AI market has transitioned from a phase of technical validation to one of large-scale revenue generation. The monthly run-rate implied by the combined figure approaches $10 billion, a velocity of capital intake that would place these companies among the fastest-growing enterprises in economic history. The word 'accelerates' in the original report is telling; it suggests that the 2026 growth rate exceeds that of 2025, a pattern consistent with the diffusion of a transformative technology from early adopters to the early majority.
However, the report's silence on the composition of this revenue is a significant analytical gap. A forensic approach requires a breakdown. Is the growth driven by API consumption, enterprise subscriptions, or consumer products? The answer determines the quality of the revenue. API revenue, while valuable, is often subject to usage-based volatility. Enterprise subscriptions, particularly multi-year commitments, provide more predictable cash flows. Consumer subscriptions, while sticky, are typically lower in average revenue per user. The absence of this data means we cannot assess the sustainability of the growth. Furthermore, the report does not address gross margins. In the AI industry, the cost of inference—the computational power required to generate responses—is a primary expense. If inference costs are not declining at a rate commensurate with revenue growth, the reported ARR may be accompanied by significant operating losses. The 'growth at all costs' model is viable only if a path to profitability exists. The report provides no evidence of such a path.
There is also the question of customer retention. Net Revenue Retention (NRR) is a key metric for SaaS companies, indicating whether existing customers are expanding their spend. A high NRR, typically above 120%, suggests a product with increasing value. A lower NRR, or a reliance on 'one-time' large deals, would inflate ARR figures without indicating durable demand. The report is silent on this metric. Based on my experience auditing the Compound governance module in 2020, where I quantified how early whale accounts could manipulate parameters, I have learned that the absence of data is often more informative than its presence. The omission of retention metrics in a report touting revenue growth is a red flag, not for fraud, but for a potentially incomplete narrative.
The Duopoly Dynamics: A Market Structure Under Scrutiny
The combined ARR, if accurate, would cement a duopoly in the Western AI market. The concentration of capital, talent, and compute resources within OpenAI and Anthropic would create a formidable barrier to entry. The 'flywheel' effect is well understood: revenue funds better models, better models attract more customers, and more customers generate more revenue. This self-reinforcing cycle makes it increasingly difficult for challengers, including Google's DeepMind, to close the gap. The report's omission of Google is conspicuous. If the $115 billion figure is accurate, and if Google's AI-related revenue is not in the same league, it would represent a significant strategic failure for a company with comparable technical resources. Alternatively, the omission may reflect a data source that is selectively curated, a possibility that cannot be dismissed.
The competitive dynamics between the two leaders are also a point of interest. OpenAI, with its consumer brand recognition through ChatGPT, appears to dominate the general-purpose market. Anthropic, with its focus on safety and enterprise-grade reliability, has carved a niche in regulated industries such as finance, healthcare, and law. This differentiation is rational; it avoids a head-to-head price war and allows each company to extract maximum value from its respective customer base. However, the report does not provide data on customer overlap. Are enterprises adopting a multi-cloud, multi-model strategy, or are they standardizing on a single provider? The answer has profound implications for the long-term pricing power of both companies. If customers are using both platforms, the duopoly is less stable than it appears. If they are exclusive, the moats are deeper.
A critical, and often overlooked, factor is the role of strategic investors. Microsoft is OpenAI's largest backer, and Amazon is Anthropic's. Both cloud providers have a vested interest in the success of their respective AI partners. A portion of the reported ARR may be attributed to 'internal' consumption or preferential pricing arrangements with these investors. This is not necessarily improper, but it does mean that the 'market' revenue may be lower than the headline figure suggests. The report does not address this potential distortion. In my 2024 analysis of Bitcoin ETF custody structures, I found that three major issuers used hybrid custody solutions with inadequate multi-signature thresholds, exposing investors to centralized counterparty risk despite the 'ETF' label. The parallel here is clear: the label of 'ARR' may obscure a more complex reality of related-party transactions and strategic subsidies.
The Valuation Calculus: From Narrative to Numbers
The $115 billion ARR figure, if confirmed, would fundamentally alter the valuation framework for these companies. The standard SaaS valuation metric is the Price-to-Sales (P/S) multiple, often expressed as a multiple of ARR. High-growth companies can command multiples of 10x to 20x ARR. Applying this range to the combined figure yields a combined valuation of $1.15 trillion to $2.3 trillion. This is a staggering range, but it is also a wide one, reflecting the uncertainty in the underlying inputs. The report does not provide the companies' current private market valuations, but based on 2025 funding rounds, OpenAI was valued at approximately $300 billion and Anthropic at $180 billion. These figures imply a P/S multiple of roughly 4x to 8x on the reported ARR. For companies growing at triple-digit rates, this multiple is not unreasonable; it may even be conservative.
The shift in valuation logic is significant. The market has moved from a 'narrative-driven' phase, where valuations were based on potential, to a 'revenue-driven' phase, where valuations are anchored to actual financial performance. This is a sign of market maturation, but it also introduces new risks. If the revenue growth decelerates, the multiples will contract, and the valuations will suffer. The report's use of the word 'accelerates' suggests that growth is not only continuing but increasing. This is a bullish signal, but it is also a fragile one. The sustainability of this acceleration depends on the factors I have already mentioned: the quality of revenue, the cost structure, and the competitive response.
The prospect of an IPO for either company is a logical next step. With ARR at these levels, both companies have the financial profile to be public entities. An IPO would provide liquidity for early investors and employees, and it would subject the companies to the scrutiny of public markets. The report does not mention any IPO plans, but the market will begin to price in this possibility. The entry of these companies into the public markets would be a landmark event, potentially reshaping the composition of major stock indices. The 'Magnificent Seven' of tech stocks may become the 'Magnificent Nine.'
The Infrastructure Imperative: The Hidden Cost of Scale
A $115 billion ARR implies an enormous computational footprint. Based on industry averages, where inference costs represent 20-30% of revenue, the combined annual inference expenditure would be between $23 billion and $34.5 billion. This is a conservative estimate, as it does not include training costs, which are also substantial. This level of spending means that the companies are processing trillions of tokens per day, a scale that requires a global network of data centers and a reliable supply of advanced semiconductors. The supply chain for these chips, dominated by NVIDIA, is a critical vulnerability. Any disruption in supply, whether due to geopolitical tensions or manufacturing constraints, would directly impact the companies' ability to deliver their services and maintain their growth trajectory.
The relationship with cloud providers is a double-edged sword. Microsoft and Amazon provide the necessary compute infrastructure, but they also extract a significant share of the value. The companies are aware of this dependency and are taking steps to mitigate it. OpenAI has partnered with Broadcom to develop custom AI chips, and Anthropic is working with AMD. These efforts are designed to reduce reliance on a single supplier and to improve cost efficiency. The success of these initiatives is a key variable in the long-term profitability equation. If custom chips can deliver comparable performance at a lower cost, gross margins will improve. If they fail, the companies will remain at the mercy of NVIDIA's pricing power.
The energy consumption associated with this scale is another hidden cost. The electricity required to power and cool these data centers is measured in terawatt-hours. This has environmental implications and creates regulatory and ESG compliance pressures. The companies will need to invest in renewable energy sources and carbon offset programs to maintain their social license to operate. These are not trivial expenses; they will be a permanent drag on profitability.
The Contrarian View: What the Bulls Got Right
It would be a disservice to the analysis to ignore the arguments of the bulls. The reported ARR, even if imprecise, points to a fundamental truth: enterprise demand for AI is real and substantial. The transition from pilot projects to production workloads is happening. The 'defensive procurement' hypothesis, which suggests that companies are buying AI out of competitive anxiety rather than clear ROI, is a risk, but it is not the only explanation. It is equally plausible that AI is delivering measurable productivity gains, and that the spending is rational. The report's focus on revenue, rather than on technical benchmarks or philosophical debates about AGI, is a sign of maturity. The market is rewarding companies that can monetize their technology, not just those that can demonstrate its sophistication.
The duopoly structure, while concerning for competition, is also a source of stability. The presence of two strong players, each with a differentiated strategy, reduces the risk of a single point of failure. The competition between OpenAI and Anthropic is driving innovation in safety, reliability, and cost efficiency. This is a positive development for the industry as a whole. The 'race to the bottom' in pricing, which is common in commoditized markets, has not yet materialized. Both companies are maintaining premium pricing, which suggests that their products are perceived as having distinct value.
The Takeaway: A Call for Verification
The $115 billion ARR figure is a data point, not a conclusion. It is a signal that the AI industry has achieved a scale that demands serious attention from investors, policymakers, and competitors. But the signal is noisy. The source is not authoritative, the data is not granular, and the verification is absent. My recommendation is to treat this report as a hypothesis to be tested, not a fact to be acted upon. The onus is on OpenAI and Anthropic to provide transparent financial disclosures. The onus is on mainstream financial media to conduct independent verification. The onus is on investors to demand more than a headline. Trust the code, not the press release. Run the numbers, ignore the hype. The silence from the teams on this specific figure speaks volumes. Until they speak, the prudent course is skepticism. The future of AI is being built on a foundation of data; we must ensure that the foundation is sound.