The number is $190 billion to $200 billion. That is the reported 2028 revenue target for Anthropic, leaked by four anonymous sources to the press. It is a staggering figure. It implies a compound annual growth rate of roughly 60% from a $47 billion annualized run rate in 2025. To put it in perspective: Microsoft's total revenue in 2024 was $245 billion. Anthropic, a company founded in 2021, is projecting to be within 80% of Microsoft's global revenue by year seven. This is not a financial forecast. This is a valuation anchor.
I have spent the better part of a decade auditing smart contracts and analyzing tokenomics. I have seen this pattern before. In 2017, during the ICO craze, projects would project $1 billion in transaction volume to justify a $100 million token sale. In 2021, DeFi protocols would quote 'annualized fee revenue' based on a single week of liquidity mining. The mechanism is the same: publish an extreme, forward-looking number, let the market internalize it as a baseline, then raise capital at a multiple of that baseline. The 2028 revenue target for Anthropic is no different. It is a narrative designed to shift the valuation paradigm from 'what is the company earning now' to 'what will the company earn in a hypothetical future monopoly.'
Context: The Valuation Mechanics of Unprofitable High-Growth Firms
Bankers and investors, per the report, are using enterprise value-to-revenue multiples to price Anthropic. This is standard for high-growth, unprofitable software companies. The twist is the projection horizon. Traditional SaaS valuations use forward revenue for the next twelve months. Extending the projection to three years out—2028 from a 2025 base—is, as the report notes, 'unusual.' It signals that the market is willing to price a story, not a business. The story is: Anthropic will become the default enterprise AI operating system, capturing 25% to 40% of a $500 billion to $800 billion AI software market by 2028.
But here is the problem. The revenue multiple itself is undefined. The report does not state whether it is 10x, 15x, or 20x. At 10x, the implied enterprise value is $1.9 trillion to $2 trillion. At 20x, it is $3.8 trillion to $4 trillion. For context, the entire market capitalization of all cryptocurrencies excluding Bitcoin and Ethereum is roughly $500 billion. Anthropic is being valued at 4 to 8 times that entire market. This is not a valuation. It is a bet on the shape of the future.
Core: Deconstructing the Revenue Target—A Four-Layer Analysis
Let me break this down the way I would audit a smart contract. I will examine four layers: the revenue composition, the cost structure, the competitive moat, and the narrative risk. Each layer reveals a potential flaw.
Layer 1: Revenue Composition. The $190 billion to $200 billion must come from somewhere. The report suggests it will be a mix of API calls, enterprise subscriptions, platform fees from agentic workflows, and potentially compute resale. But the largest component today is API revenue, which is a commodity business. Token prices are falling. The cost per million tokens for GPT-4 class models has dropped from roughly $30 in 2023 to under $5 in 2025. By 2028, it could be $0.50. To maintain $200 billion in revenue, Anthropic would need to serve 400 trillion tokens per year at $0.50 per million tokens. That is 40 times the current total inference volume of the entire industry. This is not impossible, but it requires a demand curve that is exponential, not linear. The assumption is that enterprise agents will drive usage, not human chat. That is a strong assumption. Based on my experience analyzing DeFi protocols, the moment the incentive stops—the moment the hype fades—the users vanish. The same applies to AI agents. If the first wave of enterprise agents fails to deliver measurable ROI, the adoption curve flattens.
Layer 2: Cost Structure. The gross margin on AI inference is not 80% like software. It is 50% to 70% at best, and that is only if compute costs continue to decline. The report implicitly assumes that Anthropic can reduce its per-token cost by a factor of 5 to 10 by 2028. This requires either massive hardware efficiency gains, custom ASICs, or a breakthrough in model architecture. All three are uncertain. The report's own analysis cites a 'high' risk of compute cost and supply bottlenecks. If Anthropic spends $80 billion to $100 billion on inference in 2028, that is a capital expenditure that must be funded by either revenue or dilution. The valuation multiples assume high growth, but they also assume high margins. If margins compress, the multiple contracts. This is exactly what happened to many L2 tokens: the promise of low fees and high throughput was undercut by the reality of data availability costs. The DA layer is overhyped; 99% of rollups don't generate enough data to need dedicated DA. Similarly, 99% of AI companies do not generate enough inference to justify massive compute farms. Anthropic is the exception, but it is also the exception that proves the rule.
Layer 3: Competitive Moat. The report identifies a 'high' probability that OpenAI or Google could leapfrog Anthropic with a generational model in 2026-2027. This is the most critical risk. Anthropic's current advantage is its 'safety-first' branding, which gives it a premium in regulated enterprise sales. But safety is a feature, not a moat. It can be replicated. The report even notes that Anthropic's model capabilities are not 'generationally ahead' of competitors. If Anthropic loses its technical edge, the revenue target collapses. The valuation anchor becomes a weight. In blockchain terms, this is like a DeFi protocol that relies on a single liquidity mining incentive. Stop the incentives, and the TVL vanishes. Code is law, until it isn't. The market is betting that Anthropic's technology will remain best-in-class, but the history of technology is full of market leaders who were displaced within two years.
Layer 4: Narrative Risk. The report itself is the evidence. The article is based on anonymous sources, and it explicitly states that the revenue prediction is being used by investors to justify a valuation. This is a self-referential loop. The narrative creates the valuation, and the valuation justifies the narrative. But narratives are fragile. If any one of the assumptions fails—if the 2028 market size is $300 billion instead of $800 billion, or if Anthropic's revenue growth slows to 40% CAGR—the entire valuation structure unwinds. The report's own analysis gives a 'medium' probability of a 'severe underperformance' scenario where revenue is only $60 billion to $100 billion. A 50% to 70% miss on the revenue target would imply a 50% to 70% collapse in valuation, assuming the same multiple. That is a $1 trillion to $2 trillion loss in value. This is not a risk. It is a certainty of a binary outcome. The market is pricing a success scenario that has a 20% probability, but it discounts the failure scenario at 0%.
Contrarian: The Blind Spot of the 'Platform' Narrative
Every analyst I have read focuses on the revenue target. They ask: 'Can Anthropic achieve $200 billion?' The contrarian question is: 'Why is a $200 billion revenue target even being discussed?' The answer is that it is a tool to raise capital on favorable terms. The blind spot is that the market is treating the prediction as a forecast, when it is actually a negotiation tactic. The report's own analysis hints at this: 'The prediction may be a best-case scenario, not a base case.' The company likely has a base case of $80 billion to $120 billion and a bear case of $40 billion. But the only number that gets leaked is the best case. This is asymmetric information disclosure. It is a classic pattern in crypto: the project announces a $1 billion TVL goal, the token pumps, the founders sell, and the TVL never materializes. The same might happen here. The investors who buy into the $200 billion narrative at a $1 trillion valuation may find themselves holding a bag in 2028 when the actual revenue is $80 billion. The market will then reprice, and the correction will be violent.
There is also a deeper blind spot: the assumption that AI will be centralized. The entire valuation of Anthropic depends on the idea that one or two companies will dominate the AI stack. But the history of the internet argues against this. The web was not controlled by a single portal. It fragmented into millions of sites. The blockchain industry was supposed to be centralized around Bitcoin, but it fractured into thousands of assets. The same pattern could happen in AI. Open-source models, decentralized inference networks, and privacy-preserving compute could erode the moat of any centralized AI company. The report's own analysis of the competitive landscape notes that Meta's open-source Llama models are 'free' and could undermine pricing power. If open-source AI reaches parity with closed-source models by 2028, Anthropic's revenue target becomes a fantasy. The market is not pricing this risk because it is extrapolating the current concentration of AI compute onto a future that may be more distributed.
Takeaway: The Vulnerability Forecast
The $190 billion to $200 billion revenue target is a valuation anchor, not a financial forecast. It is a mechanism to raise capital at a multiple of a hypothetical future. The market is accepting it because the AI narrative is powerful and the fear of missing out is strong. But the structure is fragile. The revenue target assumes continuous model leadership, falling compute costs, and a market that is willing to pay monopoly prices. Each assumption is a potential failure point. The most likely outcome is that Anthropic will achieve a fraction of the target—say $80 billion to $120 billion—and the valuation will reset to a lower multiple. The correction will be painful for late-stage investors, but it will be survivable for the company. The real risk is for the market as a whole: if the narrative collapses, it will take down the entire AI stock bubble with it. The smart money is already hedging. The rest of the market is still buying the anchor. s unintended consequences. The narrative of $200 billion is not just a forecast. It is a trap. The only question is who will be caught in it.
I have seen this before. In 2017, I audited the 0x protocol and found three race conditions in the order matching logic. The market ignored them. The protocol went live, and the front-running bots exploited the flaws. The same thing is happening now. The market is ignoring the flaws in the valuation logic. The race condition is simply the assumption that growth rates can be sustained indefinitely. They cannot. The only way to win is to look at the code—the numbers—and not the narrative. The numbers say $200 billion is possible, but only if everything goes right. Everything never goes right.