Analysis
Public markets aren't rejecting IPOs wholesale this year -- they're rejecting a specific kind of story. Lycia Therapeutics set terms for a Nasdaq listing at up to a $937.7 million valuation on the strength of a Nobel laureate-founded drug pipeline with real clinical milestones, and biotech peers like Iambic Therapeutics popped 18.19% on their trading debut, part of a run Value Add Pulse has tracked all month. Meanwhile, Firmus couldn't find buyers for its GPU-cloud IPO even after bankers floated cutting its ask from above $30 billion toward a steep discount.
The difference isn't deal size -- it's what investors are being asked to underwrite. A biotech IPO or a modestly sized SPAC asks investors to price a specific, knowable catalyst: a trial result, a target sector, a defined trust account. A mega-valuation AI infrastructure IPO asks investors to price years of continued capital expenditure against a growth curve that depends on demand nobody can fully verify yet. Public markets in 2026 are telling us, through repeated withdrawals, that they'll pay for the first kind of story at nearly any multiple and balk at the second kind even at a discount.
“Compare that to Lycia's tight $16-18 per share range, set with far more conviction because the underlying clinical data gives bankers a concrete basis for the number.”
That has a direct implication for the pipeline of AI infrastructure and neocloud companies sitting on huge private valuations. If Firmus -- backed by Nvidia itself, with real revenue and real customers -- can't clear the IPO bar at a discount, the bar for any comparable company is higher than most private marks assume. Founders and GPs holding paper gains on infrastructure unicorns should treat the next IPO attempt in this category as the real price discovery event, not the private round that set the mark.
The pattern extends beyond infrastructure. Nscale's pending NYSE listing has seen its reported valuation range swing between $25 billion and $35 billion in the weeks before its roadshow -- the kind of pricing uncertainty that reflects bankers testing investor appetite in real time rather than a settled market view. Compare that to Lycia's tight $16-18 per share range, set with far more conviction because the underlying clinical data gives bankers a concrete basis for the number.
Room for disagreement: the IPO window could reopen fast for infrastructure names the moment one mega-deal successfully prices. Investor appetite for an entire category often turns on a single proof point rather than shifting gradually, and Firmus's failure doesn't guarantee Nscale's does too -- a clean Nscale pricing in the next few weeks would undercut this thesis quickly. There's also a timing argument: AI infrastructure demand could simply outrun today's skepticism if capacity shortages keep intensifying, making this year's withdrawals look like bad timing rather than a structural repricing.