The leaked term sheet hit my Telegram feed at 2:47 AM Jakarta time. Anthropic raising at a $965 billion pre-money in May 2026, with secondary market chatter already pricing the IPO at $2 trillion. The numbers are beautiful. Too beautiful. The kind of beautiful that makes you check the block explorer twice.
Let me be clear: I am not calling this a fraud. I am calling it a faith-based valuation. And faith, in crypto terms, is just liquidity waiting for a mirror.
Context: The $2T Question
Anthropic is the AI darling that built Claude—the model that codes better than it reasons, that aligns better than it competes. The company has been the darling of the safety-first crowd, the Constitutional AI gospel, the MCP protocol that every major player now claims to support. But the market is now pricing it as if it has already won the enterprise AI crown.
$10-12 billion in annualized revenue by end of 2026. That's the whisper number. At $2 trillion valuation, that's a forward price-to-sales ratio of 180x. Compare: Nvidia at 24x. OpenAI—if you believe the $500B valuation with $50B revenue—at 10x. Anthropic's multiple is not just premium. It's a different asset class.
Core: Deconstructing the Narrative
Technology: Modular Innovation, Not Paradigm Shift
Claude's architecture is a deep optimization of the Transformer, not a new computational paradigm. The attention mechanism improvements—long context, sparse attention, thinking mode—are module-level innovations. Impressive, but not the kind of breakthrough that justifies a trillion-dollar monopoly.
Anthropic's real tech edge is in engineering: Constitutional AI alignment, Model Context Protocol, and Claude Code's agentic capabilities. These are product-level differentiators, not architectural moats. MCP has gained adoption, but it's a protocol, not a lock-in. Open source alternatives are already emerging.
Based on my audit experience during the 2020 Uniswap V2 flash loan exposé, I learned that the most dangerous narratives are the ones that conflate first-mover advantage with technology moat. Anthropic's lead in enterprise agentic workflows is real, but it's a speed advantage, not a structural barrier. The switch cost for a developer to move from Claude Code to a competing agent is measured in weeks, not years.
Commercialization: The $2T Math Doesn't Work
$10-12B revenue by 2026. Let's stress-test that. In 2025, Anthropic's annualized revenue was around $5B at best (various reports conflict). To hit $10-12B by end of 2026, the company needs to double revenue in two years. That's a 100% CAGR—achievable in a hypergrowth market, but the multiples are the issue.
180x P/S at the high end of revenue. What does that imply? The market is assigning a premium for AGI option value. But AGI is not a binary option; it's a spectrum. And the path to AGI is littered with inference costs. Anthropic's gross margin is likely below 50% given its reliance on cloud partners (Microsoft Azure, AWS, Google Cloud) and massive compute capex. At $10-12B revenue with <50% gross margin, the company is still deeply unprofitable. A deeply unprofitable company at 180x P/S is a historical anomaly.
Competition: The Three-Front War
Anthropic is fighting on three fronts:
- OpenAI: Superior consumer reach, stronger multimodal (Sora, real-time voice), larger revenue base. OpenAI's valuation is lower, but its revenue is higher. If OpenAI goes public at $3-5T with $50-100B revenue, Anthropic's P/S premium becomes indefensible.
- Google: Gemini's integration with search, Android, and cloud gives it a distribution moat that Anthropic cannot match. Google can afford to subsidize AI to win enterprise mindshare.
- Open-source: DeepSeek, Llama, Qwen are closing the gap on coding benchmarks at a fraction of the cost. The price-performance ratio of open-source models is improving faster than proprietary models can keep up. Anthropic's API pricing is near parity with OpenAI, but open-source models are already 10x cheaper for inference.
Anthropic's lead in enterprise agentic workflows is real, but it's a lead, not a lock. The moment a Fortune 500 company can deploy a comparable open-source agent with a week of fine-tuning, the switching cost evaporates.
Industry Impact: The Signal That Changes Everything
If Anthropic IPOs at $2T, it rewrites the rulebook for enterprise software valuation. The market will be forced to price AI companies on a different metric: not revenue, but token consumption rate. A new paradigm emerges: valuation = (daily inference tokens) × (price per token) × (growth multiple). This is not the SaaS model. It's a utility model.
Chaos is just data we haven't indexed. The IPO will create a new asset class: pure-play AI model companies. The ripple effect will hit traditional SaaS (Salesforce, Snowflake) as investors reprice them for obsolescence. It will boost GPU infrastructure plays (Nvidia, AMD, Marvell). It will also accelerate the "AI agent replaces software subscription" narrative, which is already spooking legacy software vendors.
But the hidden signal is the one no one is talking about: the IPO is a liquidity event for early investors and employees, not a capital raise. The company doesn't need $2T to grow; it needs to cash out. That's a red flag. In crypto, we call that a token unlock event. And we all know what happens to prices after the unlock.
Contrarian: The Real Value Is Not in the Model
The market is pricing Anthropic for AGI. But the real value of the company lies in something more mundane: the enterprise data flywheel. Every Claude Code session, every enterprise contract, every API call generates data that improves the model's performance on specific workflows. This is the switching cost that matters: not the model itself, but the accumulated knowledge of the client's codebase, processes, and domain.
Influence flows where attention bleeds. Anthropic's attention is on enterprise, but its bleeding edge is in consumer. The company has no ChatGPT equivalent, no viral consumer product. Its enterprise focus is a strength, but it also means the valuation is entirely dependent on a single revenue stream. If enterprise adoption slows—due to economic downturn, regulation, or competition—the entire valuation story collapses.
Another blind spot: the "alignment tax." Anthropic's commitment to safety—Responsible Scaling Policy, Constitutional AI—may slow down model releases. If Claude 5 or 6 is delayed due to safety reviews, while OpenAI or Google ships faster, the market will penalize Anthropic not for being safe, but for being slow. The market rewards speed, not caution.
Takeaway: The Next Watch
The IPO filing will reveal the true numbers. I will be watching three things: (1) revenue composition—how much is API vs. enterprise vs. subscription; (2) gross margin trajectory—are they burning cash to buy growth?; (3) the lock-up period—when do insiders dump? The answer to these questions will determine whether $2T is a floor or a ceiling.
Launch day is a promise; the code is the betrayal. Anthropic's code is solid. But the market's promise of $2T may be built on sand. In a sideways market, capital flows to narratives that survive stress-testing. This one hasn't been stress-tested yet.
Eyes on the block.