The hash does not lie, only the narrative does.
Anthropic, an AI company with no publicly disclosed revenue run rate above $1B, claims it will generate $190-200B by 2028. That is three times AWS's current annual revenue. The number appears in a leaked fundraising document, circulated to Asian capital markets via Jin10 in August 2024. No methodology. No audited baseline. Just a promise.
I trace the blood trail through the blockchain of corporate finance. The pattern is identical to every DeFi whitepaper that promised 10,000% APY. The same structure: an opaque source, a round number that exceeds all historical precedents, and a timing engineered to maximize FOMO. The difference is the asset class — AI, not tokens. The mechanism is the same.
Context
Anthropic was founded in 2021 by former OpenAI employees. Its public narrative centers on "constitutional AI" and safety. It raised over $7B by mid-2024, with backing from Amazon and Google. Its primary product is the Claude series of large language models, sold via API and a consumer chatbot. Estimated annualized revenue in Q2 2024: $5-10B. The leaked forecast targets $190-200B by 2028 — a 20-40x increase in four years.
This is not a projection. It is a signal. Signals have three audiences: investors, enterprise customers, and competitors. The message to each is different, but the mechanism is the same: claim dominance to create dominance.
Core: Systematic Teardown
I dissect the code to find the human error. The forecast fails on five independent dimensions. Each failure is a red flag that any on-chain analyst would flag immediately.
1. Technical Feasibility
$200B in AI revenue implies either massive volume at current prices or a radical increase in value per token. Current Claude pricing is $3/$15 per million tokens. To reach $100B from API alone, Anthropic would need to process ~6.7 quadrillion tokens per year — roughly 200 trillion tokens per day. That is equivalent to generating the entire text content of the internet every 48 hours. No existing or planned infrastructure can support this. Even if inference costs drop 100x, the volume remains absurd.
Alternatively, the revenue must come from high-value autonomous agent services. That requires models that can reliably execute complex workflows across legal, medical, and financial domains. Current models hallucinate. They cannot be trusted with core business decisions. Anthropic's own research on interpretability acknowledges this gap. The forecast assumes a technological leap to near-AGI within four years. That is not a plan. It is a wish.
2. Commercial Scalability
From $10B to $200B in four years requires a compound annual growth rate of 276%. No software company in history has sustained that for more than two years. Even Microsoft during its peak cloud migration grew at 30-40% CAGR. The only precedent is in crypto — where tokens can achieve exponential growth through speculation, not revenue. Anthropic is not a token. It sells services to real businesses with real budgets.
Enterprise sales cycles are 12-18 months. To capture $200B in revenue, Anthropic would need to sign contracts totaling that amount by 2027. The largest enterprise software contract in history is less than $10B. The math does not work without inventing a new category of spending.
3. Competitive Landscape
OpenAI's annualized revenue in mid-2024 was 3-5x Anthropic's. Google has Gemini. Meta has Llama. Microsoft is building its own models. Amazon and Google are both investors in Anthropic — but they also compete. If Anthropic becomes a $200B company, it will be a direct threat to its own cloud partners. The relationship is a double-edged sword. The forecast implicitly assumes Anthropic will dominate, but the data shows a fragmented market with multiple well-funded players.
4. Infrastructure Constraints
$200B in revenue implies processing power equivalent to 500,000+ H100 GPUs running at full capacity. That requires capital expenditure of $50-100B and annual electricity costs of $5-10B. Even if Anthropic signs long-term contracts with Amazon and Google, the sheer scale of hardware procurement will strain global supply chains. There is no evidence that Anthropic has secured this capacity. The forecast does not mention infrastructure.
5. Regulatory and Safety Contradiction
Anthropic's public identity is "safe AI." Its Responsible Scaling Policy (RSP) promises to pause development if models reach certain capability thresholds. A $200B revenue target requires deploying the most capable models at scale, with minimal safety delays. The two goals are fundamentally opposed. Either the RSP is a marketing document, or the revenue target is. You cannot have both.
Contrarian Angle
What the bulls got right: AI is a transformative technology. The total addressable market could be enormous. If autonomous agents replace white-collar labor, the economic value created could exceed $1T annually. Anthropic's safety focus could be a competitive advantage in regulated industries like healthcare and finance. The forecast, while extreme, might be a deliberate stretch target to align internal incentives.
But the bulls ignore the timing. A four-year window to capture 10% of a hypothetical market is not a forecast. It is a narrative designed to justify a $100B+ valuation today. The probability of hitting $200B by 2028 is near zero. The probability of the narrative working is much higher.
Silence is the loudest proof in the ledger.
Anthropic did not release a public forecast. It leaked. The leak is the product. The actual revenue is irrelevant. What matters is that investors, customers, and competitors now act as if the forecast might be true. That action — capital allocation, contract signing, talent acquisition — creates a self-fulfilling loop. The forecast becomes a coordination mechanism.
Consensus is verified, not believed.
I do not believe this forecast. I verify it against known constraints. The verification fails. But the market does not care about verification. It cares about momentum. In that sense, the forecast is perfectly designed for the current cycle.
Takeaway
Anthropic's $200B revenue forecast is a fundraising artifact. It will succeed or fail based on its ability to attract capital, not its accuracy. For investors: track the quarterly revenue run rate, not the 2028 target. For developers: watch for model capability improvements, not press releases. For the rest: recognize the pattern. It is the same pattern that drove every crypto boom. The asset changes. The mechanics do not.
The chain remembers what the mind tries to forget.
Anthropic's true test is not 2028. It is whether it can deliver 5x growth from 2024 to 2025. If it cannot, the narrative collapses. If it can, the narrative survives another round. That is the only signal that matters. Everything else is noise.
