Proof exists; it is merely waiting to be verified.
Hook
A single data point has become the gravitational center of AI valuation: Anthropic's projected 2028 revenue of $1900–$2000 billion. Anonymous sources leaked this internal target to the press, and investors are already using it to justify enterprise-value-to-revenue multiples that would value the company at $2 trillion or more. I have audited enough broken ledgers to recognize the pattern. This is not a financial forecast. It is a narrative anchor—a psychological price floor designed to precondition the market for a massive capital raise. The same technique was used by every crypto project that promised “hypergrowth” before its token collapsed.
Context
Anthropic, the AI safety company behind the Claude model family, has reportedly achieved an annualized revenue run rate of $470 billion by mid-2025. The path from $470 billion to $2000 billion in three years implies a compound annual growth rate of roughly 60%. The article—originally published on a Web3 news outlet—cites two anonymous sources familiar with Anthropic's financial planning and four additional sources involved in fundraising discussions. Bankers and investors are using enterprise-value-to-revenue multiples to price the company, a method typical for high-growth, unprofitable software firms. What makes this unusual is the extension of the projection horizon to 2028—three years beyond the typical forward-looking window. In traditional SaaS, such a move signals desperation for a narrative that justifies an otherwise unjustifiable valuation.
Core: A Systematic Teardown
The 2028 revenue target rests on three implicit assumptions that, when examined forensically, reveal structural fragility.
Assumption 1: Model Capability Dominance Through 2028. Anthropic's Claude models currently rank at or near the top of benchmarks for coding, agent tasks, and long-context reasoning. But the AI landscape is not static. OpenAI and Google are investing at comparable or larger scales. Meta is open-sourcing competitive models for free. If Anthropic loses its technical edge—even for a single generation—its pricing power evaporates. The revenue target implies that Claude 5, 6, or their equivalents will maintain a 12- to 18-month lead over rivals. That is a strong assumption given the history of rapid catch-up in this field.
Assumption 2: Inference Costs Collapse by 80–90%. To serve 2000 billion dollars in revenue, Anthropic must process an astronomical volume of tokens. Assuming a blended gross margin of 60–70%, the cost of inference alone would be $600–$800 billion annually. Today's GPU rental costs make that math impossible without a radical reduction in per-token cost. The company is reportedly exploring custom ASIC chips with partners, but silicon timelines are unforgiving. A single delay in tape-out could push cost reductions beyond the 2028 window. I have seen this movie before: crypto miners promising ASIC efficiency gains that never materialized, burning through capital while the hashrate stayed flat.
Assumption 3: Enterprise Adoption at Unprecedented Scale. Anthropic's revenue today is heavily weighted toward API calls distributed through Amazon Bedrock and Google Vertex AI. To reach $2000 billion, the company needs millions of enterprise customers, each spending hundreds of thousands of dollars annually. That requires moving beyond API consumption into full-platform lock-in—becoming the operating system for corporate AI workflows. No pure-play AI company has ever achieved that scale. Even Salesforce, after 25 years, generates roughly $35 billion in annual revenue. Anthropic expects to generate 50 times that in seven years. The numbers do not survive even a basic sanity check.
My Forensic Experience with Forward Projections
During the FTX collapse, I obtained a leaked internal ledger and spent three weeks reconciling it against on-chain deposits. The discrepancy: $2.4 billion. The root cause was not fraud in the traditional sense—it was a systematic overvaluation of illiquid tokens used as collateral. FTX's leadership had projected future revenues from trading fees that assumed market share growth that was mathematically impossible given the total addressable market. Anthropic's 2028 projection exhibits the same logical flaw: it assumes a market share that, if realized, would make it one of the largest software companies in history, exceeding Microsoft's 2024 revenue of $245 billion. The difference is that FTX's lies were hidden in a private ledger. Anthropic's are being broadcast to the public to set the terms of its next funding round.
The Data Does Not Lie—But the Narrative Does
The article mentions that “bankers and investors use enterprise-value-to-revenue multiples to value Anthropic.” That is true. But it omits the critical detail: the multiple applied to a 2028 revenue estimate is inherently speculative. If the market uses a 10x multiple on $2000 billion, that implies a $20 trillion enterprise value. No private company in history has achieved that. Even if we assume a more conservative 5x multiple, the valuation still reaches $10 trillion. To put that in perspective, the entire global market capitalization of all publicly traded technology companies is roughly $15 trillion. Anthropic is asking the market to believe it will be worth two-thirds of that in three years.
The Algorithm Remembers What the Witness Forgets
I have traced enough blockchain transactions to know that memory is selective. Investors remember the upside scenario and forget the downside. The article itself acknowledges that the 2028 prediction might be an optimistic case, with internal base and pessimistic scenarios likely lower. But those scenarios are not being discussed. The only number circulating is $1900–$2000 billion. That is a classic anchoring bias technique—set a high anchor, and even a discount to that anchor feels reasonable. In crypto, we call this “pump the narrative before the token sale.”
Contrarian: What the Bulls Got Right
To be fair, the bulls have a point. AI is not a speculative technology like most crypto tokens. It is generating real revenue and solving real problems. Anthropic's safety-first positioning gives it a unique moat in regulated industries like healthcare, finance, and law. Enterprise clients are willing to pay a premium for auditable, reliable models. The company's partnership with AWS and Google provides distribution that no other AI startup has. If inference costs drop faster than expected, and if Claude maintains its technical lead, the 2028 target becomes less absurd. The bulls also argue that the market is pricing AI as a new general-purpose technology, not just a software category. That argument has historical precedent: the internet went from zero to trillion-dollar industries in a decade. But the internet had network effects and marginal costs near zero. AI has high marginal costs and no natural monopoly.
Takeaway: Accountability Requires Verification
Ledgers balance, but ethics remain uncalculated. Anthropic's 2028 revenue prediction is a bet on a specific future that may or may not materialize. The market should demand verifiable on-chain metrics—actual API call volumes, customer churn rates, inference cost trends—rather than accepting a forward-looking narrative as gospel. I have seen what happens when investors trust projections without verification. FTX. Terra. Three Arrows. The pattern is always the same: a compelling story, anonymous sources, and a valuation that defies gravity. The only difference this time is that the story is about AI, not crypto. But the math works the same way. The algorithm remembers. The ledger does not lie. It is time to verify.
