To get into Anthropic right now, you need at least $25 million. For OpenAI, it’s closer to $1 million.

In the pre-IPO secondary market, demand for Anthropic shares is reported to far exceed supply by roughly 3x–5x, with about $1.5B of interest since Q2 2026 and valuation cited around ~$1.4T (+400% YoY), while OpenAI demand is said to have only partially recovered. Anthropic’s $65B fundraise is framed as driving a “demand explosion,” alongside tighter cap-table access that makes <$25M tickets difficult and typically requires ~$50M+ for access at $1.3–1.4T valuations. One broker view calls OpenAI “underpriced” and Anthropic “overpriced,” suggesting an opportunity for investors able to access both names.

Analysis

Secondary-market clearing prices here are a liquidity signal, not a clean read on franchise value. The bigger mechanism is access scarcity: when cap-table control tightens, reported premiums can outrun fundamentals and then gap down quickly if a tender widens or insiders sell more stock. That makes the current spread more useful as a measure of who is allowed to buy risk than as evidence that one model company is structurally superior.

For public-market investors, the actionable takeaway is that capital is still being allocated to the AI layer, but the economic capture likely sits upstream in compute, networking, and cloud distribution rather than in the private marks themselves. If these companies keep raising at escalating valuations, the immediate beneficiary is the infrastructure stack; if the next financing or rumored IPO reveals slower revenue conversion or heavier burn, the secondary premium can mean-revert over days to weeks. The consensus may be overreading scarcity as moat.

The contrarian risk is that OpenAI’s broader distribution ultimately matters more than Anthropic’s tighter float, which would make today’s relative pricing look backward-looking once monetization data becomes visible. Over 6-18 months, the real falsifier is any sign that model quality converges faster than expected or that hyperscaler capex decelerates, which would compress the whole AI complex’s valuation ladder rather than just the private names.

Key Decisions for Investors

• No direct position in private secondary names; treat this as a watch item until there is a new primary round, tender, or IPO filing. If broader supply comes to market and the spread persists, the scarcity-premium thesis weakens.
• Long NVDA / AVGO on a 1-3 month horizon into continued AI-funding headlines; risk/reward favors the infrastructure layer because every dollar of frontier-model capital formation ultimately pulls through to compute and networking demand.
• Pair trade: long MSFT / short AI (C3.ai) as a quality-vs-hype expression for the next 1-3 months. The trade works if investors keep paying for platform optionality and punish names with weaker monetization visibility.
• Set an alert on hyperscaler capex commentary and any OpenAI/Anthropic financing prints. A meaningful downshift in capex or a widening secondary supply over the next quarter would be the first sign to fade the AI premium.