Anthropic’s internal revenue projection for 2028 is $1.9-2 trillion. That number is larger than the entire market cap of every crypto asset combined — twice. The market is not irrational; it is inefficiently priced. The alpha isn’t in the silenced code of the prediction itself, but in the structural inefficiencies it reveals about AI infrastructure scalability.
Context: The Prediction and Its Anchors
Anthropic, the AI company behind Claude, has communicated to investors that it expects to generate $1.9-2 trillion in revenue by 2028. This is based on its current annualized revenue run rate of $47 billion (as of early 2025) and a compound annual growth rate of roughly 60%. Bankers and investors are using enterprise value-to-revenue multiples to value the company — a method typically reserved for high-growth software firms without profits. The key anomaly: the projection period extends to 2028, three years beyond the typical 12-month forward window. This is a deliberate anchor, designed to pull valuation frameworks into a “terminal state” narrative.
Core: Dissecting the On-Chain — or Off-Chain — Evidence Chain
From my experience auditing ICO whitepapers in 2017 and later building DeFi arbitrage scripts, I’ve learned that revenue projections must be stress-tested against three constraints: technology, infrastructure, and competition. Anthropic’s prediction fails on none of these outright, but the margin for error is razor-thin.
First, technology. Anthropic must maintain model leadership through 2028. Its Claude series currently ranks top-tier in code and agent tasks (SWE-bench, LMSYS Arena). But the industry moves in generational leaps. One lag in multimodal reasoning or self-improving agents could erase the premium pricing power. The prediction implicitly assumes Anthropic remains at least co-leader with OpenAI and Google. That is not guaranteed.
Second, infrastructure. To support $2 trillion in revenue, assuming a 60-70% gross margin, Anthropic’s annual inference cost would be $600-800 billion. That requires a cluster of millions of H100-equivalent GPUs, with unit inference costs dropping 5-10x from current levels. Anthropic is already working with AWS and Google on custom ASICs and large-scale deployments. But the energy and data center capacity required is national-scale. Any bottleneck — chip supply, power grid, cooling — directly caps the revenue number. In crypto, we see similar constraints with Layer2 rollups: blob data will saturate post-Dencun, and gas fees will double. Scarcity is an algorithm, not a belief system.
Third, competition. The prediction implies Anthropic captures 25-40% of a $5-8 trillion global AI software market in 2028. If the market is only $3 trillion, Anthropic would need a 65% share — unprecedented for any software company. OpenAI, backed by Microsoft, and Google, with its full-stack advantage, are not standing still. The revenue multiple is a bet on oligopoly, not monopoly.
Contrarian: Correlation is Not Causation — The Liquidity Trap
Correlations are the lie; liquidity is the truth. The revenue multiple works only as long as capital flows into AI at its current pace. If interest rates rise or a recession hits, the same multiples contract violently. The 2022 Terra/Luna crash taught me that on-chain liquidity drains before the narrative breaks. For Anthropic, the risk is asymmetric: the prediction is all upside, but the downside is a 50-70% revenue miss. In that scenario, valuation would collapse from $2 trillion to $500 billion overnight. The market is pricing the dream, not the variance.
Moreover, the prediction itself may be a strategic signal for fundraising, not a sober forecast. Four anonymous sources leaked it to the press — likely aligned with Anthropic’s investor relations team. The timing, during a capital raise, is classic anchor bias. I don’t trade narratives; I trade structural inefficiencies. The real inefficiency here is the gap between the projection and the infrastructure reality. The ledger remembers what the marketing forgets.
Takeaway: The Next 36 Months
The alpha isn’t in believing Anthropic’s $2 trillion story. It’s in understanding that the AI compute bottleneck will drive demand for decentralized, permissionless alternatives. Crypto projects like Render, Akash, and new zero-knowledge proof networks for AI verification are positioned to capture the overflow when centralized clusters hit their limits. Due diligence is the only hedge against chaos. Watch for signals: Anthropic’s next funding round, its ASIC tape-out date, and the quarterly revenue growth rate. If the CAGR drops below 40%, the narrative shifts. The market will remember what the hype forgets.
Forward-looking: The blockchain industry’s opportunity is not to replicate AI’s centralized model, but to provide the infrastructure layer that AI’s growth will inevitably require. The next 36 months will reveal whether the prediction is a signal of a new era or a relic of peak narrative.