Analysis
2026 has been framed as the biggest IPO year in a decade, and by the topline numbers it is: US IPOs have raised roughly $251 billion across 86 deals through late July, comfortably surpassing 2025's full-year total of about $47.4 billion. But that headline obscures a market that is really splitting into two very different regimes, and this week's news cycle put both on display simultaneously.
On one side sits AI-infrastructure-scale issuance, led overwhelmingly by SpaceX's record $1.75 trillion pricing, which alone accounted for roughly a third of every dollar raised in US IPOs this year. Anthropic and OpenAI are both reportedly working through confidential S-1 processes at similarly outsized prospective valuations. These are the names driving the eye-popping topline dollar figure, and they're also the names most exposed to exactly the kind of sentiment swing this week's chip-stock selloff represents -- a market questioning whether AI infrastructure capex is outrunning actual demand.
On the other side sits everything else: Jersey Mike's pricing this week at up to a $7.9 billion valuation for a 70-year-old, 3,300-location sandwich chain with a knowable, understandable margin profile; a steady background rate of smaller S-1 filers across biotech, land, power systems and crypto; and growth-but-not-AI-infrastructure names like fintech data network Plaid, which reached an $8 billion valuation and has held early talks with banks about a potential listing, and wearables maker Oura, which confidentially filed at an $11 billion mark built on genuine consumer revenue growth rather than infrastructure capex.
“Anthropic and OpenAI are both reportedly working through confidential S-1 processes at similarly outsized prospective valuations.”
The distinction that matters for anyone pricing risk right now is capital intensity. AI-infrastructure names are being underwritten on the promise that massive, ongoing capex converts into proportional future returns -- exactly the promise this week's chip-stock selloff is calling into question. Jersey Mike's, Plaid and Oura are underwritten on more traditional growth-and-margin stories that don't depend on that same capex-to-return conversion holding up. In a week when SK Hynix posted record profit and still got punished 10% for merely growing instead of re-accelerating, the market's preference for the second group over the first is becoming harder to ignore.
For founders and GPs thinking about exit timing over the next 12 months, the practical read is: don't assume the 2026 mega-IPO window's multiples apply to your company if your growth story isn't genuinely AI-infrastructure-scale. A clean, profitable, understandable business -- the Jersey Mike's model -- may find a friendlier public-market reception right now than a capital-intensive AI story competing for the same investor dollars as SpaceX, Anthropic and the chip names currently under pressure.
What to watch: whether Plaid and Oura's IPO processes, if they proceed, get priced on AI-infrastructure-style multiples or on more traditional growth-company benchmarks, and whether the current chip-sector selloff bleeds into softer reception for the AI-infrastructure mega-IPOs still in the pipeline behind SpaceX.
