The numbers are staggering. Anthropic's annualized revenue run rate surged from $14 billion in February to over $47 billion in May—a 236% increase in three months. Its private valuation jumped from $380 billion to $965 billion in the same period. Market chatter now places the IPO valuation north of $2 trillion. But as a protocol developer who has spent years parsing the deterministic core of distributed systems, I see a familiar pattern: the euphoria of a new infrastructure cycle masking the same capital allocation inefficiencies that plague blockchain scaling.
Context: The AI Infrastructure Stack
Anthropic secretly submitted its IPO filing on June 1. The company is not just a model developer; it is a massive consumer of compute, having committed to invest over $100 billion in Amazon Web Services over the next decade. It has secured up to 5GW of new computing power from Amazon, another 5GW of next-generation TPU capacity from Google and Broadcom, and even leases GPU capacity from SpaceX. The May raise of $65 billion was partly earmarked for expanding compute. This is not a software company—it is a hyperscale data center operator with a chatbot frontend.
The AI stack mirrors the blockchain stack: model developers (like Layer 1s), chip manufacturers (like hardware miners), cloud providers (like data availability layers), and software companies (like dApps). The question is not whether AI technology will grow—it will. The question is how the profits will be distributed across this stack. Investors fixated on Anthropic's revenue trajectory are ignoring the capital intensity required to maintain its competitive edge. I have seen this before in the L2 wars: every rollup touted adoption, but the real value accrued to Ethereum's base layer and to the sequencer operators.
Core: Quantitative Economic Preemption—The Cash Conversion Problem
Let me model this. Anthropic's revenue run rate is $47 billion, but its capital expenditure commitments exceed $100 billion over a decade, plus the $65 billion raised in May. Assuming a 10-year amortization, that's roughly $16.5 billion in annual capex. Operating expenses include R&D for cutting-edge models, talent acquisition, and the cost of compute itself (which is a recurring cost, not just capex). Even if Anthropic maintains a 50% gross margin (optimistic for a compute-heavy business), that leaves $23.5 billion to cover capex, SG&A, and R&D. After capex, we have $7 billion. Then R&D—likely another $10 billion annually given the race for frontier models. That yields negative cash flow. The $65 billion raise is essentially covering the gap.
This is not a sustainable business model; it is a venture-subsidized infrastructure build. The same dynamic exists in blockchain: Ethereum's rollups spend heavily on sequencer fees and data availability, but most are unprofitable at the protocol level. Code does not lie, but it often omits context. The context here is that Anthropic's growth is a function of capital injection, not organic demand. The $47 billion run rate includes revenue from services that are themselves subsidized by the cloud partners (like AWS credits).
Contrarian: The Blind Spot—Who Captures the Value?
The conventional narrative is that Anthropic will be the next trillion-dollar tech giant. But my analysis of the economic incentives suggests the opposite: the true value accrues to the infrastructure providers—Amazon, Google, Broadcom, and even SpaceX—not to the model developer. Think of it like the MEV extraction landscape I analyzed in 2025. In Ethereum, 40% of profitable transactions were bot-driven arbitrage, not organic market movement. The block builders and validators captured the value, not the application developers. Similarly, in AI, the chip manufacturers and cloud providers capture the bulk of the economic rent because they control the scarce resource: compute.
Anthropic's agreements with Amazon and Google lock it into long-term commitments that reduce its pricing power. If Anthropic tries to raise model API prices, customers can switch to OpenAI or Google's Gemini. But if Amazon raises its compute prices, Anthropic has no alternative because it has committed to 5GW of capacity. This is a classic vendor lock-in, and it mirrors the relationship between Ethereum rollups and the base layer. The standard is a ceiling, not a foundation.
Takeaway: The Vulnerability Forecast
Based on my experience auditing protocols and analyzing economic models, I project that Anthropic's IPO valuation will face significant compression within 18 months of listing. The market will realize that the company's growth is capital-intensive, its pricing power is constrained by compute dependency, and its competitive moat is eroding as other labs (Meta, Mistral, xAI) also secure capacity. The $2 trillion valuation is a scenario where the market discounts the future too aggressively, similar to the 2021 L1 altcoin craze. Investors should focus on the distribution of profits across the AI stack, not just the revenue of a single player. Parsing the chaos to find the deterministic core means looking at the cash flow statement, not the hype.
