By any headline measure 2026 has been the best year for US listings since 2021. Renaissance Capital counts $160.6 billion raised through mid-August, against 2021’s full-year record of $175 billion, with four months still to run. Two hundred and thirty-two companies have gone public, about 4% more than at the same point last year.

Then look at where the money came from. SpaceX raised $86.2 billion in June including the overallotment. One listing, more than half the year’s proceeds.

Strip it out and the picture is respectable rather than historic — and it becomes a question about what happens next rather than a celebration of what already did. Because the companies that would broaden the number are the ones now deciding whether to go at all.

The year has been genuinely strong at the top end. SpaceX in June at roughly $1.8 trillion. SK Hynix in July, raising $26.5 billion in the largest foreign listing ever done in the US. Cerebras in May, up 108% on its first day. Firefly Aerospace, Fervo Energy and BitGo alongside them, and earlier in the year Klarna, Chime, Figma, StubHub and Hinge Health.

Twelve companies raised more than $1 billion in the first half, against four in the same period last year. EY counted 62 US IPOs above $50 million through June, roughly double last year’s 34.

That is a real market, not a statistical illusion. But it is a market whose weight sits in a handful of names, and whose most-watched participants have not arrived.

The largest listing still in front of the market is Anthropic’s. It filed confidentially on June 1 at a private valuation approaching $965 billion, with Goldman Sachs, JPMorgan and Morgan Stanley leading and Nasdaq as the venue. The timeline has slipped twice. Reuters reported on September 5 that the public S-1 has moved to late September and the roadshow to mid-October at the earliest, which would put pricing within days of the November midterms.

OpenAI filed a week later, on June 8, and then immediately declined to commit. The company said it had not decided on timing and that it might be a while. Reporting since has consistently put it in 2027.

Anduril, valued around $61 billion, has still not filed at all. Databricks, Shield AI and Quantinuum remain in the same position — frequently named, not on file.

What has filed and is moving: Standard Nuclear and Syntiant in July, Panera Brands confidentially, Jersey Mike’s and Tailored Brands more recently. JPMorgan’s Keith Canton has said he expects a dozen jumbo IPOs in the second half, meaning raises of $1 billion or more.

So the pipeline has not emptied. It has split. A group of mid-sized issuers is proceeding, and the handful of companies large enough to move the annual total are making individual decisions about a window most of them are choosing not to use.

Worth noticing what has actually been listing, because the composition says more than the total does.

The year’s proceeds have come overwhelmingly from three places. AI and the infrastructure that runs it — Cerebras in chips, SK Hynix in memory, with EY putting the strongest momentum in semiconductors, power and data center capacity. Aerospace and defense, where SpaceX and Firefly both priced. And energy, where Fervo listed a geothermal business into a market that has spent three years indifferent to the sector. Fintech and digital assets contributed at the edges, with Klarna, Chime and BitGo.

That is a narrow set of stories, and it is the same set the private markets have been funding for four years. The IPO window did not open for everyone. It opened for companies whose businesses investors had already decided they wanted exposure to.

The pipeline ahead looks like more of it. Anthropic, OpenAI and Databricks in AI. Anduril and Shield AI in defense. Quantinuum in quantum computing, Standard Nuclear in reactors, Syntiant in edge chips. If those listings happen, the concentration gets worse before it gets better.

But the more interesting filings of the last two months are not any of those. Panera Brands has filed confidentially. So have Jersey Mike’s and Tailored Brands — a sandwich chain and a menswear retailer, both private-equity owned, neither of which has anything to do with artificial intelligence.

Those are the companies that tell you whether the window is genuinely open. A market that will price a trillion-dollar AI lab is not necessarily a market that will price a restaurant franchise at a sensible multiple. If the consumer names get away in the fourth quarter, 2026 stops being a story about four companies and becomes a story about a functioning IPO market. If they pull, it does not.

Three things are pushing timelines, and they compound.

The first is SpaceX itself. It priced at $135, ran to $225 within days, then gave back roughly a third. For a company whose float was under 5%, that was arguably a supply artifact rather than a verdict. But it is the most visible AI-adjacent listing in history and it traded badly in its first six weeks, and every board considering a Q4 listing has now watched it happen.

The second is the election. Midterm years have a well-documented shape: elevated volatility and muted returns before the vote, then a rally once the outcome is known. Capital Group’s analysis of more than ninety years of data finds the pattern holds consistently, and the S&P has risen in the twelve months following every midterm since 1950. From the trough of a midterm-year correction, the index has averaged roughly 31% over the subsequent year across sixteen cycles.

If that pattern holds, the best window is not October. It is December, or the first half of 2027.

The third is the macro backdrop, which has been unusually noisy. The Iran conflict has run since March, when the S&P fell 9% peak to trough before recovering. The ten-year sits near 4.7%, a level that has historically preceded volatility. Fed futures now put a September hike above 85%, after a month in which that number swung from 40% to 69% and back before Friday’s inflation print pushed it higher still. Oil has been above $100.

Morgan Stanley’s research finds equity markets have grown less sensitive to political headlines over time, and that bond markets have not. For a company planning a listing, the bond market is the one that sets the discount rate.

Which leaves an unusual situation. The most valuable private company in the world is preparing to price into the least favorable part of the calendar, and its closest competitor has stepped back to the more favorable part.

Both cannot be optimizing for the same thing.

The company going early is buying the right to set the comparable — to be the reference price against which every subsequent AI listing is measured, rather than the one measured against someone else. The company waiting is buying market conditions.

That trade-off is not unique to trillion-dollar AI labs. Any CFO who has taken a company to market has faced a version of it: the window you want is rarely the window that is open, and the cost of waiting is usually paid to whoever did not wait.

Our take: the useful conclusion from a record year is not that the window is open. It is that “the window” was never a single thing. It has been wide open for large, mature, well-understood businesses and considerably narrower for everyone else — 232 listings against a proceeds total that one company supplied half of. If you are planning a raise in the next eighteen months, the question is not whether the market is receptive. It is whether it is receptive to you, and the answer sits in your comparables rather than in the headline number.

Ava is the first AI BDR to run outbound end to end, and you decide whether she runs autonomously or on copilot.

She finds leads or ingests accounts from your CRM, enriches them, sends personalized emails on behalf of your reps, follows up, handles replies and books meetings. Website visitor de-anonymization, intent signals, and a parallel dialer come built in.

A small team can manage her centrally for thousands of reps who never log in. Everything syncs two-way with Salesforce and HubSpot.

Ava runs outbound for companies like DoorDash and Grammarly, and one customer deploys her across 1,000+ reps. Ava is SOC 2 Type II audited, SSO and GDPR ready. Ava is how revenue teams grow pipeline without growing headcount.

Book a demo

Six weeks ago we asked which of Anthropic, OpenAI or Anduril would list first. Of the 154 of you who voted, 35% picked OpenAI, 31% Anduril, 22% Anthropic, and 12% said none of them before 2027.

On present evidence the least popular answer is winning. Anthropic has a filing, underwriters, a venue and a slipping but live timeline. OpenAI has a filing and has said it may be a while. Anduril has neither.

Worth noting what that turns on. It was not a judgment about the companies. It was a judgment about which one would decide to move first — and the deciding factor appears to have been a competitor’s retreat rather than anything either company controlled.

Four things every CFO needs to know this week.

· The Fed meets today, and the market has stopped debating. August CPI held at 3.4% annually with prices up 0.4% on the month. Core inflation actually eased to 2.4%, the lowest reading since March 2021. But gasoline rose 3.9% in August and accounted for more than a third of the monthly increase, and futures now price a quarter-point hike at almost 93% — up from 71% the day before the print, and from 40% six weeks ago. Nationwide’s Kathy Bostjancic put the logic plainly: Warsh signaled rates could stay on hold only if disinflation continued, and August did not deliver it. The decision comes tomorrow, and at these odds the news will be in the statement rather than the number.

· Your freight costs are the story inside the story. The same CPI report put diesel up 52% year over year and fuel oil up 52%, against gasoline’s 27.4%. That is a direct hit to trucking and delivery costs, and it flows into the landed cost of everything you move. If your 2027 plan assumes freight normalizes, check what it is actually built on — the Iran conflict that started this is not resolved, and energy is now doing the work that tariffs did last year.

· Two-thirds of finance leaders expect AI to cost more next year, and are spending anyway. In a Deloitte survey reported this month, nearly two-thirds of respondents expect AI-related costs and operational complexity to rise in the coming year. Most are prioritizing the investment regardless. That is not irrational — but it is a budget line growing faster than the returns are being measured, and the measurement is the part most finance functions have not built yet.

· The pharma working capital number worth borrowing. Hackett Group research found the pharmaceutical sector’s cash conversion cycle climbed to nearly 186 days in 2025, as companies increased investment in biologics and GLP-1 manufacturing. The number matters beyond pharma: it is what happens when a sector commits to capital-intensive production faster than its receivables and inventory cycles adjust. Any finance function funding a large capacity build should know what its own cycle did last year, and whether anyone noticed.

1. Which half of the market are you in? The 2026 numbers describe two different markets. Large, mature businesses with comprehensible economics have been able to raise almost anything. Everyone else has faced a normal, selective market. Be honest about which description fits, because the headline proceeds figure describes the first and most companies are the second.

2. What are your comparables actually worth? If your benchmark names listed this year with floats under 10%, their prices reflect scarcity as much as fundamentals — a point we made in August and which the SpaceX recovery has not settled. A valuation built on distorted comps is distorted.

3. Have you priced the cost of waiting? Every timeline that slips has a cost, and it is rarely modelled. Another two quarters of private funding, another year of option grants at a stale strike, another cycle of investors asking when. The company that waits for better conditions sometimes finds the conditions improved and its position did not.

4. What is your bridge if the window closes? Anthropic’s most important milestone before its roadshow is reportedly a $15 billion credit facility. That is the thing that lets a company walk away from a soft market rather than price into one. Whatever the equivalent is at your scale — a revolver, a bridge, an extended runway — it determines whether you have a choice or a deadline.

5. Are you first or second? Being the first comparable in a category is worth something real, and so is watching someone else discover the price. Neither is obviously right. But the decision should be made deliberately rather than by default, and most companies default to going when they are ready rather than when the sequencing favors them.

SPONSOR THIS NEWSLETTER

Your name, in front of the CFOs and senior finance leaders who read The CFO Desk every Tuesday. The primary slot is available for Q4. Write to [email protected].

A record year in which one company supplied half the proceeds is two stories wearing one number. For the companies at the top, the market has been as good as it has ever been. For everyone else it has been ordinary, and ordinary is not what the headlines have described.

The next nine weeks will settle which story 2026 turns out to be about.

One question before you go: if you were taking a company public, would you price before the midterms or wait?

One click. And if you have a view on why, reply — we read everything.

The Editors hold no position in any company mentioned.

— The CFO Desk
[email protected] · thecfodesk.com
CAPITAL · FINANCE · LEADERSHIP