Analysis
I don't think the 2026 IPO market is rewarding what most founders think it's rewarding.
Look at the actual receptions: Cerebras priced at $185, opened at $350, and closed its first day at $95 billion in market value (TechCrunch) -- a chip company with $510 million in revenue and $88 million in net income.
Oura is filing at a $16 billion-plus target on $1.2 billion in nine-month revenue and real profitability (Bloomberg).
“Room for disagreement: this could just be 2026's specific cohort, not a durable structural shift.”
SpaceX raised $75 billion at $1.77 trillion. Every one of this year's marquee listings has a physical or infrastructure component -- chips, satellites, hardware sensors -- not a pure-software multiple.
That's not an accident, and it's not just AI hype. It's investors pricing scarcity they can underwrite with their eyes open: there are a finite number of companies that can manufacture wafer-scale chips, launch reusable rockets, or ship a validated health-sensing ring at scale, and public markets are paying a premium for businesses where the moat is physical and hard to replicate, not just a better interface on top of a foundation-model API. Compare that to the software IPO candidates still sitting in registration -- Databricks, Stripe, Canva -- all with real revenue and real growth, but none of which have priced in 2026, and none getting the "biggest IPO ever" headlines Cerebras and SpaceX got.
The thesis founders need to internalize
If you're building software with a defensible moat that's genuinely hard to replicate -- proprietary data, deep workflow lock-in, network effects -- this doesn't change your thesis. But if your pitch is "we built a better interface on top of GPT-6 or Gemini," the public market is telling you, loudly, that it doesn't currently reward that category at a premium multiple, no matter how fast your ARR grows. The AI application layer is getting commoditized in real time -- see Microsoft cutting transcription prices 72% this same week -- while the infrastructure layer underneath it is getting rarer and more expensive to replicate.
Room for disagreement: this could just be 2026's specific cohort, not a durable structural shift. Wonderful just doubled its valuation to $5 billion on pure software orchestration with no hardware component at all, and Gimlet Labs hit $3 billion on inference software, not chips. Private markets are still paying software multiples happily -- it's specifically the IPO window, where public investors demand near-term profitability or physical scarcity, that's currently favoring infrastructure. That's a liquidity-event problem, not necessarily a valuation problem, and it may resolve itself the moment one software-only company IPOs well and resets the pattern.
