Openai flops on secondary market while anthropic bags $2b in waiting lists
Secondary brokers can’t give OpenAI stock away. Literally. “We can’t find a single institutional buyer,” says Ken Smythe, whose shop Next Round has moved $2.5B of pre-IPO paper. Meanwhile, rival Anthropic has $2B of pent-up demand from the same client base. The scent of a regime change is filling server rooms from Menlo Park to Midtown.
Price tags tell the story
OpenAI still trades at roughly double Anthropic’s internal valuation, but that gap is shrinking fast. Investors are betting Anthropic closes it before either company rings the opening bell, a move both are eyeing for this year. Goldman Sachs is so desperate to offload OpenAI slices that it waives success fees for private-bank clients, a perk it refuses to extend for Anthropic, where it charges the standard 15-20% carry. Translation: the bank needs coaxing to move OpenAI; it can afford to charge full freight for Anthropic.
The math is brutal. If you buy OpenAI today you’re paying peak hype prices for a firm burning billions on model training and consumer freebies. Anthropic sells mainly to Fortune 500 accounts, carries fatter margins, and projects a steeper revenue ramp. “Risk-reward is simply better,” snaps Adam Crawley, co-founder of Augment, one of the busiest secondary platforms. His dashboard shows record inbound requests for Anthropic allocations while OpenAI listings gather digital dust.
Primary cash vs secondary cold shoulder
OpenAI’s $12.2B primary round last week—led by SoftBank, Microsoft, and a parade of sovereign funds—looks heroic on paper. In practice, it may have sealed the secondary stall. Early employees and VCs who got in at sub-billion valuations now want liquidity, but new money won’t bite at the $157B tag. Instead they’re wiring checks to special-purpose vehicles tracking Anthropic, hoping to front-run an IPO pop that secondary desks pencil in for October.
Dragos Rasmussen, a former Pentagon cyber officer turned hedge-fund partner, spent last week urging limited partners to ignore the DoD’s supply-chain warning against Anthropic. He frames the Pentagon memo as bureaucratic turf warfare, not a technical red flag. His pitch deck: “If Washington flinches, Wall Street feasts.” Two family offices already allocated $200M each.
The customer card
ChatGPT’s 400M weekly users don’t pay the bills. OpenAI’s enterprise ARR is still sub-$1B and churn is creeping up. Anthropic’s Claude is embedded inside Amazon Bedrock and inside Google Cloud’s enterprise suite, locking in multi-year contracts worth nine figures apiece. One leaked slide from a recent board deck shows Anthropic’s gross margin at 72%, 18 points above OpenAI’s last disclosed figure. Investors read that as a proxy for pricing power.
Both firms forbid secondary trades without consent, yet shares circulate through offshore SPVs and forward contracts. Brokers now price Anthropic at a 35% premium to the last primary round; OpenAI changes hands at a 12% discount. The spread widened after Anthropic’s second source-code leak this month—buyers shrugged, arguing any model can be jailbroken, but not every board can convert fear into revenue.
Bottom line
OpenAI invented the generative gold rush, yet today it is the one stuck holding the bag. Anthropic, the once-dismissed safety-first shop, is flipping the script: higher margins, tighter governance, and a secondary market that treats its paper like bottled lightning. When the opening bell rings, the company that worried less about headlines and more about invoices may sprint out of the gate while the poster child watches from behind.