‘A pressure cooker ready to explode’: The wild secondaries scramble for Anthropic shares

Anthropic is reportedly raising up to $50 billion at a roughly $900 billion valuation, up from a prior $380 billion mark, amid extraordinary investor demand for secondary shares. The article says the company could be on pace for about $45 billion in annualized revenue run rate, though that figure is explicitly described as a snapshot rather than realized revenue. The main signal is frothy AI private-market demand and elevated positioning ahead of a widely expected 2026 IPO window for Anthropic and OpenAI.

Analysis

The key signal is not just exuberance in one private asset, but the emergence of a repeatable liquidity regime where AI winners can be financed at near-public-market scale before public-market scrutiny arrives. That creates a powerful reflexive loop: fresh capital validates product-market superiority, which pulls in more customer spend, which then justifies still higher private marks. The second-order winner is the broader AI supply chain—compute, networking, and data-center infrastructure—because every incremental dollar of private demand ultimately leaks into capex demand for the picks-and-shovels stack.

This also tells us something important about positioning: when secondary demand becomes price-insensitive, the marginal buyer is no longer underwriting fundamentals but scarcity. That tends to compress future IPO upside, because the public listing becomes a liquidity event for insiders rather than a discovery event for new investors. In the next 3-6 months, the most likely spillover is multiple expansion across the “AI toll roads” rather than the flagship model names themselves, since public investors will look for cleaner ways to express the trade without paying private-market premiums.

The contrarian risk is that these prints become a local top in sentiment if the market starts to question whether annualized revenue can justify the valuation velocity. If growth decelerates even modestly over the next 1-2 quarters, secondary buyers sitting on thin paper gains could flip from scarcity-chasing to distribution mode very quickly. That would hit late-stage private AI comps first, then high-beta public AI software names that have been trading as proxy exposures.

The other non-obvious risk is policy and process: the more chaotic the allocation process, the more likely regulators and boards become uncomfortable with opaque secondaries and side-letter economics. That would not kill demand, but it could reduce transaction velocity and widen bid-ask spreads, forcing a reset in implied liquidity and compressing near-term markups.

Key Decisions for Investors

• Go long NVDA on any 3-5% pullback over the next 2-4 weeks; the cleanest second-order beneficiary is incremental AI capex, with asymmetric upside if private-market enthusiasm converts into another wave of infrastructure orders.
• Add to ANET and VRT on weakness; these names monetize the same demand cycle with lower valuation risk than the frontier model layer, offering a better risk/reward if AI private marks stall.
• Buy a basket of high-beta AI software names vs short a broad software index over the next 1-3 months; the trade expresses continued scarcity-driven multiple expansion while limiting exposure to non-AI enterprise deceleration.
• For event risk, consider long-dated call spreads on NVDA or ANET into the next earnings cycle; upside is driven by capex acceleration, while downside is capped if the private-market froth proves temporary.
• Avoid chasing late-stage private AI analogs without liquidity; if public-market proxies start underperforming while private prints stay elevated, that is often the first sign that secondary demand is peaking.