AI IPOs 2026: Wall Street finally gets a real bill
Iambic has priced a $150 million offering, while Anthropic’s compute commitments and OpenAI’s missing date…
The short version
- Iambic’s priced $150 million offering is the first confirmed transaction among the closely watched AI IPO candidates of 2026.
- Anthropic lacks a public filing, while reported infrastructure commitments exceed $500 billion and could outlast customer demand.
- Investors should prioritize filings, audited economics, payment schedules, and clinical evidence over rumored dates, run rates, and preclinical charts.
Wall Street finally has an AI IPO to buy, and it’s a cancer-drug startup most people hadn’t heard of a month ago. Iambic Therapeutics priced its offering on October 14, 2026, expecting $150 million in gross proceeds before fees and expenses. Finally, a receipt. Among the closely watched AI IPOs of 2026, Iambic has crossed the pricing line. Anthropic still owes investors a public filing. OpenAI has no confirmed IPO date. One biotech reached the cashier while the frontier labs browse with extremely expensive taste.
Their risks differ wildly. Iambic can design a gorgeous molecule that face-plants in patients. Anthropic can reserve years of compute, then discover customers treat model providers like streaming subscriptions. As an Italian, I respect theatrical branding, but eventually the bill lands.
Iambic finally puts a price on the pitch
Iambic offered about 9.4 million shares at $16 each, expecting $150 million in gross proceeds before underwriting discounts, commissions and other expenses. That was the midpoint of its proposed $15-to-$17 range. The banks found enough demand without retreating to the cheap seats, though the first weeks of trading will test its durability.
I use a strict confirmation ladder because three reposts and a podcast appearance can turn IPO rumors into “basically happening.” Investors first need inspectable public documents. Then come regulatory review, pricing and share allocation. Trading can begin before closing, once remaining conditions are satisfied and money changes hands. A confidential filing ranks much lower because ordinary investors cannot read it. A rumored target month barely ranks at all. Iambic now has a priced transaction and dollar amount, putting it ahead of its louder cousins.
The headline demand needs seasoning. ARK Investment Management and Duquesne Family Office indicated they might buy up to $60 million, 40% of the base deal, leaving 60% for others. Those indications are nonbinding. Allocations can shrink or vanish, so two famous buyers do not prove broad public-market appetite.
Convertible-note holders get a sweeter deal. At closing, their notes convert at the lower of the valuation-cap price or 85% of the IPO offer price, a 15% discount to new buyers. I, too, would order tiramisù if someone else paid full menu price.
Gross proceeds are clear. Net proceeds after underwriting discounts, commissions and expenses are not. Anyone assigning the entire headline amount to drug development is spending money Iambic has not confirmed it will receive.
The software stops where human biology begins
Iambic’s pitch gets interesting beyond “molecular superintelligence,” which sounds like a model named itself after two Negronis. The company says lead candidate IAM1363 went from program launch to human testing in about two years. That measures speed to Phase 1, not approval, efficacy or commercial success. Several brutally expensive steps remain.
Here’s the machinery. Without a high-resolution inhibitor-bound experimental structure for KIF18A, Iambic used NeuralPLexer to generate structural insights guiding IAM217’s design. Enchant and other prediction tools helped optimize potency and oral bioavailability while checking drug-interaction liabilities and central-nervous-system penetration. Iambic paired those predictions with automated chemistry and biology experiments at scale. Lab evidence fed another design cycle, eliminating weak compounds and refining promising ones. After choosing a candidate, Iambic prepared an investigational new drug submission. Human testing requires FDA clearance. This workflow may compress discovery; clinical trials still decide whether the medicine is tolerable and helps anyone.
PotionUI described model-specific controls this way:
Forms, not wiring, each model exposes only the controls it actually understands.
IAM1363 has produced an early human signal. In an exploratory, open-label data set, Iambic reported a 23% confirmed objective response rate, without a randomized comparator showing how patients might have fared under another treatment.
That meant 12 responses among 53 evaluable patients with targetable HER2 alterations and measurable systemic disease. All received daily doses of at least 960 mg. The result deserves attention, but the trial design sharply limits my conclusions.
IAM217 is earlier. Iambic reported about 90% regression of established intracranial tumors versus pretreatment burden in a brain-metastasis model. No patients were involved.
Pete Olson, Iambic’s chief scientific officer, described the ambition carefully:
We believe IAM217 has the potential to be a differentiated inhibitor for cancers that rely on KIF18A to proliferate.
“Potential” is doing honest work. Iambic still needs FDA clearance to start the proposed trial, and nobody knows whether the brain-tumor result will translate into human safety or efficacy. Biology has spent decades humiliating gorgeous preclinical charts. AI has not negotiated an exemption.
Public investors will probably value Iambic mainly as a biotech. Its platform may improve candidate selection and shorten discovery, but clinical evidence controls cash flows. Repeatable human data from IAM1363 would boost the software’s credibility. If the pipeline disappoints, nobody will frame the NeuralPLexer screenshots.
Anthropic has booked the banquet hall
The Anthropic IPO has a stranger problem. Its product has customers, but its infrastructure commitments may outlive their enthusiasm by years.
Reported figures put Anthropic’s annualized revenue run rate, based on July revenue, at $65 billion, versus $4.6 billion of reported revenue in 2025. That growth grabbed me. I was more impressed than I should have been, annoying for someone who spends an indecent amount of time correcting other founders’ denominator crimes.
A run rate annualizes one recent period. Audited annual revenue may differ, and the figure says nothing about cash surviving model-serving costs. Without a public filing, nobody outside Anthropic has its audited GAAP revenue, gross margin or cash flow. We also lack complete customer-concentration data and contractual payment schedules. Any precise valuation feels like ordering wine without seeing the bottle.
Reported infrastructure commitments exceed $500 billion over the next decade, without a published annual payment schedule. Leaked filing information describes multiyear take-or-pay contracts. These can secure scarce compute before Anthropic needs it, meeting demand without waiting years for new data centers. Higher utilization spreads committed costs across more paid usage. Trouble starts if customer spending slows faster than obligations. Buyers can reduce API usage or switch models while Anthropic keeps paying for reserved capacity. Newer chips could also make today’s infrastructure less attractive before the contracts expire.
There is a serious bull case. Anthropic says Claude Haiku 5.5 costs about 75% less to run on average than Haiku 4.5, though workload and context length affect the reduction. Lower serving costs can support cheaper usage and better unit economics. If demand keeps compounding, securing infrastructure early could look brilliant.
The Mesmer Tools benchmark author offered this caution:
I wouldn’t call it lossless.
PitchBook offers the cleaner skeptical case. A reported $2 trillion valuation would sit against margins, customer spending and contracted-capacity economics public investors cannot inspect. Anthropic needs sustained margin improvement and enough demand to use its reservations. If pricing weakens, or efficiency reduces required compute, capacity security becomes a spectacularly elaborate financial trap.
From an Anthropic IPO filing, I want audited revenue, gross margin, cash flow and the infrastructure deals’ payment schedule. Until then, the valuation debate is financial cosplay with nicer spreadsheets.
When is the OpenAI IPO? Microsoft and Oracle can’t tell us
Anyone asking when the OpenAI IPO will happen gets my boring answer: available sources confirm no date. Unlike Iambic’s priced offering and disclosed share terms, OpenAI has no public filing or documented trading date for a listing calendar.
Anthropic’s preparations create no deadline for its rival. Companies time public offerings around their own capital needs and disclosure obligations. An Anthropic filing would give bankers a valuation reference while revealing investors’ questions about compute commitments. OpenAI can watch without following it through the door.
I rank dates by evidence. A company’s investor-relations announcement comes first because it states the event directly. A public securities filing follows, providing company-specific accounting or contractual details. Partner earnings add context, though Microsoft’s cloud business includes far more than one AI lab. Supplier results are broader still. Third-party calendars often extrapolate from historical patterns. Each step adds assumptions, turning guessed dates into fake deadlines before breakfast.
NVIDIA’s calendar dispute shows why. AltIndex, on a page updated October 8, 2026, lists the next earnings report for Tuesday, November 17, 2026, and says the exact time will be announced closer to the event.
Finance Calendar instead expects Wednesday, November 18, 2026, after market close, based on NVIDIA’s usual reporting pattern. It acknowledges NVIDIA had not formally confirmed the date. Both calendars estimate. NVIDIA investor relations gets the final word.
The same rule applies to the MSFT earnings date. Supplied sources do not establish Microsoft’s next report date, so I would verify it through Microsoft investor relations. Its earnings may reveal cloud demand, but not Anthropic’s margins or contractual payment schedule.
The ORCL earnings date also needs direct confirmation from Oracle. Available evidence does not establish Oracle as a direct Anthropic infrastructure partner, so using its earnings as an Anthropic proxy takes several imaginative leaps. My nonna used fewer substitutions when the pantry was empty.
When Anthropic’s filing arrives, I’ll search “take-or-pay” before reading one benchmark chart. Here’s my dated prediction, available for future humiliation: this cycle’s best-performing AI listing will be the company whose customer commitments last at least as long as its infrastructure bills.
Revenue growth gets the television graphic. The footnote decides who keeps the money.
Frequently asked questions
Which AI companies have confirmed IPOs in 2026?
Iambic Therapeutics is the confirmed priced deal among the closely watched 2026 candidates. It offered about 9.4 million shares at $16, expecting $150 million in gross proceeds. Anthropic still lacks a public filing, and OpenAI has no confirmed IPO date.
Has Anthropic filed for an IPO?
An Anthropic IPO has no inspectable public filing in the available evidence. Investors therefore lack audited GAAP revenue, gross margin, cash flow, customer concentration, contractual payment schedules, and full details of reported infrastructure commitments exceeding $500 billion over the next decade.
When is the OpenAI IPO?
OpenAI has no confirmed IPO date or documented public trading date. Anthropic’s preparations do not create a deadline for OpenAI because each company times a public offering around its own capital needs and disclosure obligations. A confirmed date requires direct company information or a public securities filing.
Sources
- Iambic Announces Pricing of Initial Public Offering
- Brookmont Pre-IPO AI Interval Fund, Form N-2 Registration Statement
- US Capital Markets Watch Q3 2026
- Moonshot AI Said to Eye Early 2027 IPO After Value Hits $50 Billion
- Failed AI IPO puts fear before FOMO
- The drug disconnect between Iambic and Isomorphic