The multiple is not a rounding error. It is the explicit market price for the belief that trapped-ion quantum computing will become commercially meaningful before the end of the decade, and that Quantinuum — specifically, not its better-funded competitors — will be the company that captures the value when it does. Both propositions deserve scrutiny before the roadshow closes.

What the S-1 Actually Discloses

The prospectus, filed with the SEC on May 8, is unusually candid about where the company stands commercially. Quantinuum’s customers are engaging with its systems primarily through “exploratory, research-driven or pilot programs, rather than long-term production deployments,” the filing states. The company has accumulated a deficit of approximately $1.5 billion since inception and has invested roughly $2 billion in research and development across its predecessor organizations over the past decade. The $79.3 million in bookings disclosed for 2025 represents signed customer agreements that may convert into future revenue, not recognized revenue — and the gap between bookings and recognized revenue in Q1 2026 is stark.

Honeywell will retain roughly 49% of the votes after the offering. Founding shareholders Honeywell and Cambridge Quantum Holdings together will hold approximately 82% of equity post-IPO. That concentration means the public float is relatively thin, and it means existing HON shareholders do not automatically receive QNT shares — direct quantum exposure requires participating in the IPO or buying on the secondary market after listing.

“We believe that we are executing a roadmap to the first commercial-scale, fully fault-tolerant quantum computer before the end of this decade, the Apollo system.”
— Rajeeb Hazra, President and CEO, Quantinuum, Letter to Investors, S-1 Prospectus, May 2026

The technical pitch rests on accuracy rather than qubit count. Quantinuum’s ion-trap architecture — which uses charged atoms held in electromagnetic fields as qubits rather than the superconducting circuits used by most rivals — delivers what the company claims is industry-leading gate fidelity: 99.921% on two-qubit operations for the Helios system. The argument is that fewer, higher-quality qubits running more reliable operations can outperform architectures with larger raw qubit counts. Whether that holds at the scale required for commercially useful applications is the question the Apollo roadmap is supposed to answer by 2029.

The Peer Comparison Nobody Wants to Sit With

Quantinuum is pricing into a public quantum sector that has rebounded sharply from its March lows. IonQ is up roughly 132% since the end of March. D-Wave is up approximately 110%. Rigetti and Quantum Computing Inc. are both up more than 85% over the same period. The rally has made the sector look healthy in isolation. The revenue comparison makes it look considerably less straightforward.

IonQ reported $130 million in full-year 2025 revenue — more than four times Quantinuum’s figure — and guided to between $260 million and $270 million for 2026 after posting $64.7 million in Q1 alone. IonQ’s implied price-to-sales multiple, at roughly 77 times trailing revenue, is aggressive by any conventional measure. Against Quantinuum at 411 times, it looks almost conservative. The investor who chooses Quantinuum over IonQ is not just making a bet on quantum computing — they are making a more specific bet that Quantinuum’s architecture and software stack will prove more durable than IonQ’s commercial momentum, despite IonQ having more than four times the current revenue base and a clear near-term growth trajectory. That is a defensible position, but it requires a framework that goes well beyond the S-1 financials.

The structural difference is that Quantinuum took the traditional IPO route rather than the SPAC path that brought most public quantum names to market. J.P. Morgan and Morgan Stanley are joint lead bookrunners. That combination of institutional underwriting and Honeywell’s balance sheet backing gives the deal a different profile from the earlier generation of quantum listings. It also means the institutional allocation process will be a genuine signal: if the book fills well at the $45 to $50 range, it indicates that sophisticated long-only capital is willing to hold a 400x revenue multiple on a research-stage asset. That has implications beyond Quantinuum.

The Government Backstop and What It Actually Covers

The timing of this IPO is not coincidental. On May 21, four days before Quantinuum set its pricing terms, the Trump administration announced more than $2 billion in Commerce Department funding for a group of U.S. quantum computing firms. IBM received the anchor award at $1 billion to build Anderon, a domestic quantum wafer foundry. Quantinuum is set to receive up to $100 million, structured through a non-binding letter of intent under the CHIPS Act, to be disbursed in tranches tied to specific technical milestones: developing low-loss integrated photonics, prototyping control chips for cryogenic operation, and packaging optical components for trapped-ion systems.

The government backstop matters, but the S-1/A is explicit that it is not yet finalized. In exchange for the funding, Quantinuum will issue equity securities to the Department of Commerce on the award date — dilutive to public shareholders, though the precise stake size has not been publicly disclosed. The funding is milestone-gated, not guaranteed, and the letter of intent is non-binding. Investors who are pricing the government relationship as a firm commitment rather than a conditional one are reading a different document than the one filed with the SEC.

What Pricing Day Will Actually Test

Quantinuum is not the only large, technically ambitious IPO landing in the first half of June. SpaceX is targeting June 12. The book-build for a $75 billion raise at a $1.75 trillion valuation is running simultaneously with Quantinuum’s $1.05 billion roadshow. The capital allocation question — whether institutional investors have the appetite and the mandate to participate in both — is real, though Quantinuum’s much smaller raise means it is unlikely to be directly crowded out by SpaceX demand.

The more relevant test is what first-day trading communicates about sector pricing. Quantinuum will likely be the first quantum computing company to price via a traditional IPO with full institutional bookrunner involvement. How the stock opens relative to the $45 to $50 range, and whether it holds above issue price in the first week of trading, will set the reference point for how public markets are currently willing to value the quantum computing category. That has direct implications for IonQ’s multiple, D-Wave’s recovery, and — less obviously — for any private quantum company that has been using the public sector’s rebound to support its own valuation narrative in secondary markets.

The BCG Quantum Forecast cited in Quantinuum’s own prospectus projects $5 to $10 billion in end-user value from quantum computing by 2030, scaling to up to $850 billion by 2040. The IPO is priced on the assumption that investors believe the 2040 number, are willing to pay for it in 2026, and have concluded that Quantinuum — over IonQ, over IBM’s quantum division, over Google’s Willow program — will be the platform that captures a meaningful share of it. That is the bet on the table. The roadshow will show how many institutional investors are prepared to take it at 411 times trailing revenue.

What to Watch Next

• Quantinuum’s first-day trading relative to the $45 to $50 pricing range. An opening above issue price with institutional-driven volume confirms that long-only capital is willing to hold 400x-plus revenue multiples in the quantum sector. A first-day dip or flat open signals that even with Honeywell backing and government endorsement, the market found the multiple too rich — which would immediately pressure IonQ and D-Wave valuations and complicate any private quantum company using the sector rally in secondary pricing.
  
• The CHIPS Act award definitive agreement. The non-binding LOI for $100 million remains unsigned. Watch for Commerce Department confirmation of a final award date; the equity dilution terms — currently undisclosed — will become calculable for public shareholders only after the definitive documents are filed. The milestone structure (photonics fabrication, ASIC prototyping) will also reveal Quantinuum’s technical execution timeline in more granular terms than the S-1 roadmap provides.
  
• IonQ Q2 2026 results, expected early August. IonQ guided to $65 to $68 million in Q2 revenue after posting $64.7 million in Q1. If IonQ delivers again at or above the midpoint of guidance, the company will have logged more revenue in the first half of 2026 than Quantinuum generated in all of 2025 — a comparison that will become harder to ignore as analysts recalibrate relative valuations across the public quantum sector post-IPO.
  
• Quantinuum’s Sol system disclosure timeline. The prospectus targets Sol, the generation after Helios, for 2027. Any technical delay announcement, partner-access preview, or early performance specification will serve as the first real-world signal on whether the 2027 milestone is tracking — and whether the Apollo 2029 target remains credible.
  
• SpaceX first-day performance on June 12 and its downstream effect on IPO risk appetite. If SpaceX prices at or above $1.75 trillion and opens strong, it validates the broader thesis that public markets in 2026 will absorb large, pre-profitability technology listings at premium valuations. That would benefit Quantinuum’s secondary-market trading and extend the window for OpenAI’s September roadshow. A SpaceX stumble compresses all three timelines simultaneously.

That story is well-covered. The one running alongside it is not.

According to Dealroom data cited by CNBC in late April, venture capitalists have already funneled $18.8 billion in 2026 into AI startups founded since the start of 2025. That figure is tracking ahead of the $27.9 billion raised by similarly new companies across all of 2025. A parallel funding market is running — smaller in absolute terms, structurally different in almost every other way, and increasingly relevant to investors trying to find returns that don’t require getting into a $1 trillion IPO at the ground floor.

What the Frontier Labs Left on the Table

The concentration of capital at the top of the AI stack has a predictable side effect. When the largest labs compete to release the best frontier model on the shortest cycle, large areas of applied research get deprioritized — not because they lack value, but because they don’t move benchmark scores or win press cycles. That creates an opening.

“When you’re in a race, you narrow focus. That creates a vacuum. Entire areas of research, like new architectures, agents, interpretability and vertical models, are being deprioritised, not because they don’t matter, but because they don’t win the immediate race.”
— Elise Stern, Managing Director, Eurazeo, speaking to CNBC, April 28, 2026

Eurazeo backed AMI Labs, the Paris-based startup founded by Yann LeCun after he left his role as Meta’s chief AI scientist. AMI Labs raised $1.03 billion in a seed round — from Bezos Expeditions, Nvidia, and Samsung — focused on AI systems that understand the physical world, maintain long-term memory, and handle complex reasoning in real-world settings. That is precisely the territory the frontier labs have had least incentive to develop aggressively while the benchmark arms race for text and multimodal generation remains the primary commercial and reputational priority.

AMI Labs is not alone. Former Google DeepMind reinforcement learning researcher David Silver announced a $1.1 billion seed round for Ineffable Intelligence in late April. Another ex-DeepMind researcher, Tim Rocktäschel, is reported to be raising up to $1 billion for Recursive Superintelligence, which focuses on continuous learning architectures. Recursive Intelligence — co-founded by former Anthropic and Google DeepMind researchers Anna Goldie and Azalia Mirhoseini — raised $335 million across two rounds after launching in September 2025, focused on AI tools for chip design. These are seed-stage investments. Most of these companies have no product and no revenue. Investors are paying for the talent, the insight into what the frontier labs are not doing, and the architectural bets that may not produce returns for years.

The More Investable Side of This Wave

The billion-dollar seed rounds capture attention, but they are the thinnest part of the $18.8 billion figure to underwrite. The more instructive pattern — and the more relevant one for sophisticated investors evaluating private market exposure — is what is happening in the application layer, where companies founded since 2025 are raising at Series A and B on the strength of real enterprise revenue rather than researcher pedigree.

The vertical AI thesis has moved from pitch deck to commercial reality in several regulated sectors. Chapter, an AI platform for Medicare advisory, raised $100 million in April 2026 to scale into one of the most structurally complex and high-willingness-to-pay environments in U.S. healthcare. AcuityMD, a commercial intelligence platform for medical device manufacturers, raised $80 million in a Series C led by StepStone Group — notable because the lead is a large institutional alternative asset manager rather than a traditional venture firm, signaling that private markets capital is following AI revenue into medtech sales infrastructure. In fintech, the pattern is consistent: Performativ raised €5.5 million for an AI-native operating system for wealth management; Marloo raised $10 million for financial adviser workflow tools. These are not demo-stage companies. They are solving compliance-heavy operational problems that enterprise buyers have budget to fix.

The common thread is workflow ownership. Vertical AI companies that sit inside a daily operational workflow — not alongside it as an assistant, but embedded in the process itself — generate the retention and expansion economics that justify enterprise software multiples. A Medicare advisory platform that improves by learning from every case it handles is harder to displace than a chatbot wrapper. A commercial intelligence platform that integrates into a sales team’s daily motion builds switching costs over time. The investors writing the larger checks in the second wave are distinguishing between these two categories sharply.

The Dependency the Second Wave Cannot Ignore

The structural question that sits beneath all of this is how closely the second wave’s returns are correlated with the first wave’s success. The application-layer companies building on top of frontier models are, to varying degrees, dependent on the pricing, availability, and capability trajectory of the underlying infrastructure they use. An agentic healthcare platform built on a foundation model faces a different risk profile in 2028 if that model’s provider has repriced API access post-IPO, or if a better-performing open-source alternative has compressed the incumbent’s moat.

That dependency runs both ways. The $300 billion Q1 investment wave was premised partly on enterprise adoption accelerating as fast as the private market assumes. The vertical AI companies now raising at Series A and B on real revenue are the clearest available evidence for whether that assumption is correct. If Chapter’s Medicare platform scales, if AcuityMD’s medtech customers renew and expand, if Performativ’s wealth management operating system crosses into institutional distribution — those outcomes provide the revenue substance that the frontier lab valuations require to be justified from below. The second wave and the first wave need each other, even when they are funded by different parts of the market on different timelines.

What Sophisticated Investors Are Actually Pricing

The $18.8 billion flowing into post-2025 AI startups is not homogeneous. The billion-dollar seed rounds for researcher-founded labs are effectively long-dated bets on architectural approaches that may or may not produce commercial products in the next three to five years. They are venture capital in its purest form — high conviction, long horizon, binary outcomes. The Series A and B rounds for vertical AI companies with paying enterprise customers are something closer to growth equity underwriting: the product works, the market is real, the question is execution at scale.

Investors who conflate these two categories — treating all $18.8 billion as a uniform signal about AI startup health — will draw the wrong conclusions in both directions. The researcher-pedigree seed rounds tell you something about where the frontier labs are not competing. The vertical AI revenue rounds tell you something about where enterprise AI spending is actually landing. Both signals matter. They require different analytical frameworks, different hold periods, and different risk tolerances. The fact that they are both being described as the “second wave” of AI funding is a convenience, not a thesis.

What to Watch Next

• First revenue disclosures from the billion-dollar seed labs. AMI Labs, Ineffable Intelligence, and Recursive Superintelligence have no disclosed revenue. When any of them announces a first commercial product or a named enterprise customer, it will be the first data point separating architectural ambition from commercial execution.
  
• Series B conversion rates for 2025-founded vertical AI companies. The cohort that raised Series A rounds in 2025 on early enterprise traction is now entering its Series B window. How many convert — and at what valuation step-up — will establish whether the vertical AI revenue thesis is compounding or plateauing after early adopter saturation.
  
• Foundation model API pricing post-IPO. OpenAI’s S-1 filing process will eventually disclose its API cost structure. Any post-IPO repricing of inference access changes the margin math for every application-layer company built on top of it and may accelerate the shift toward open-source model deployment among enterprise buyers.
  
• Enterprise AI renewal data in Q2 and Q3 earnings. The hyperscalers’ Q1 results confirmed that AI infrastructure spend is accelerating. The more important signal — whether enterprise software buyers are renewing, expanding, and deepening their vertical AI deployments — will start surfacing in midmarket SaaS earnings through the summer.
  
• Yann LeCun’s first AMI Labs product announcement. LeCun’s specific focus on grounding, causality, and physical-world reasoning represents a direct architectural critique of the current transformer-dominant paradigm. The first public demonstration of whether that critique produces a commercially usable system will be one of the more consequential AI product announcements of the year.

That sequencing matters. Because what the market priced in the first session and what the announcement actually contains are not the same thing.

Letters of Intent Are Not Contracts

The Department of Commerce was careful about its language. The agency announced the signing of nine letters of intent — a standard pre-award step that initiates a due diligence and negotiation process before any funds change hands. LOIs are not disbursement orders. They are the beginning of a process that includes compliance review, final term negotiation, and, in some cases, Congressional notification requirements. The CHIPS Act semiconductor awards that preceded these quantum grants — including the Intel deal announced in August 2025 — went through months of negotiation between LOI signing and final agreement execution. There is no disclosed timeline for when these quantum LOIs convert to binding agreements.

That is not a reason to dismiss the announcement. The policy signal is real: the U.S. government has formally designated fault-tolerant quantum computing as a strategic infrastructure priority, extended the CHIPS Act industrial policy model to an entirely new sector, and named nine specific technology approaches worth backing. That has lasting implications for the companies involved and for the broader competitive dynamic with China. But it is a different thing from nine companies receiving $2 billion in cash.

“With today’s CHIPS Research and Development investments in quantum computing, the Trump administration is leading the world into a new era of American innovation.”
— Howard Lutnick, Secretary of Commerce, Department of Commerce press release, May 21, 2026

The Equity Stake Provision Is the Part Worth Understanding

Every one of the nine awards includes a government equity stake as a condition of funding. This is the same structure applied to the Intel CHIPS award last year — a model the administration has now explicitly extended beyond semiconductors into quantum hardware. For sophisticated investors, the equity provision changes the ownership math in ways that weren’t priced into the session-day spike.

A government equity position means dilution for existing shareholders, the creation of a new class of stakeholder with no profit motive and potentially different priorities, and the introduction of ongoing disclosure and compliance obligations tied to the award terms. The specific stake sizes and governance rights have not been disclosed for any of the nine recipients. Until the LOIs convert to final agreements — and the terms of those agreements are made public — the equity provision is an open variable sitting inside a valuation that moved 15–17% in a single session on incomplete information.

The Portfolio Deliberately Avoids Picking a Winner

The nine recipients span nearly every viable approach to quantum computing that exists today. IBM and Rigetti pursue superconducting qubits. D-Wave operates quantum annealing systems — a fundamentally different architecture that is the only approach currently running enterprise applications in production at meaningful scale. Quantinuum works with trapped-ion technology. Atom Computing and Infleqtion develop neutral-atom systems. PsiQuantum builds photonic quantum processors. Diraq is developing silicon-spin qubits using standard CMOS fabrication lines, which, if it works at scale, could be the most manufacturable approach of any on the list.

IBM anchors the portfolio with a $1 billion commitment to build Anderon, a new standalone subsidiary described as America’s first pure-play quantum wafer foundry. GlobalFoundries receives $375 million to establish a multi-modality quantum foundry covering superconducting, trapped-ion, photonic, topological, and silicon-spin architectures. The foundry layer is the strategic bet beneath all the others: without domestic manufacturing infrastructure for quantum-grade superconducting wafers, every other approach on this list eventually hits a supply chain dependency. The government is trying to solve the whole stack, not just fund the most visible names.

D-Wave Is the Outlier in This Group — and That Is Worth Noting

Most of the seven quantum computing recipients are research-stage companies with limited or no commercial revenue. Rigetti reported $7.1 million in full-year 2025 revenue against a GAAP net loss of $216 million. IonQ, the largest of the publicly traded names by market capitalization, generated $110 million in 2025 revenue — meaningful, but still heavily weighted toward government contracts and research institutions rather than enterprise commercial deployments at scale.

D-Wave stands apart. The company reported $24.6 million in 2025 revenue, up 179% year-over-year, with more than 135 paying customers including over two dozen Forbes Global 2000 enterprises using its annealing systems in production. That is not a lab experiment. D-Wave’s annealing architecture is architecturally distinct from the gate-model systems most of its co-recipients are building toward, and the $100 million grant it received is the same size as Rigetti, Quantinuum, and Infleqtion — companies with fundamentally different revenue profiles. The government’s equal-weight treatment of companies at very different stages of commercial maturity is a deliberate hedge, not a performance ranking. Investors who interpreted the uniform move in quantum stocks as a uniform validation of all nine companies were reading the announcement incorrectly.

The China Context Is the Real Driver

The announcement’s geopolitical framing was not incidental. Commerce Secretary Lutnick’s statement emphasized American leadership and domestic industry explicitly, consistent with the administration’s broader posture on strategic technology competition. China has made quantum computing a national priority under its 14th and 15th Five-Year Plans, with state investment in quantum research estimated to exceed U.S. levels on a sustained basis since 2021. The CHIPS Act quantum awards are designed to do what the semiconductor awards were designed to do: establish domestic manufacturing infrastructure for a technology the government has decided it cannot afford to import.

For private market investors, the quantum grant structure also raises a question that extends beyond this announcement. If the CHIPS Act model — government grant plus equity stake, conditioned on domestic production and supply chain requirements — becomes the standard path for quantum hardware companies to reach commercial scale, it changes the return profile for early venture investors in these companies. A government co-investor with a minority equity stake and compliance strings attached is a different thing from a clean cap table. That conversation has barely started.

What to Watch Next

• LOI conversion timelines for each of the nine recipients. The CHIPS Act semiconductor awards took months to move from LOI to final agreement. Watch for any Commerce Department disclosure of a review schedule — that will set the timeline for when these incentives become binding.
  
• Government equity stake terms when disclosed. The specific stake sizes, governance rights, and exit provisions have not been published. When they are, the dilution math for existing shareholders in Rigetti, D-Wave, and IonQ becomes calculable for the first time.
  
• IBM Anderon capitalization structure. IBM said it expects outside investors to join as Anderon scales beyond the initial $2 billion commitment. A foundry that manufactures quantum-grade superconducting wafers for multiple architecture types — and that carries government backing — could attract strategic investment from every company on the recipient list.
  
• D-Wave Q1 2026 results and bookings trajectory. The company entered 2026 with over $32.8 million in post-year-end bookings already closed. First-quarter results will show whether its commercial momentum is accelerating ahead of, or independent from, the government grant runway.
  
• China’s response. Watch for any accelerated procurement announcements from Beijing targeting quantum hardware companies in Europe, Australia, or Canada — particularly PsiQuantum, which is Australian-founded — as the U.S. moves to lock in domestic supply chain relationships.

The 10-year Treasury yield touched 4.60% on May 18, its highest level in 15 months. The 30-year bond crossed 5.2% the same week, a threshold last seen before the 2007 financial crisis. Neither number is a crisis signal on its own. Together, in the context of three pre-profitability growth companies preparing to ask public investors to accept revenue multiples above 75x, they represent a meaningful repricing of the environment those roadshows must navigate.

Two Reports in One Week Closed the Door on Rate Cuts

The bond market move was not a slow drift. It was triggered by back-to-back inflation data that arrived the week of May 12 and left rate-cut expectations in pieces. April CPI came in at 3.8% year-over-year, the highest reading since May 2023, driven primarily by a 28.4% surge in gasoline prices and 17.9% energy costs broadly, both flowing from the disruption to Strait of Hormuz shipping that began in late February. Core CPI, which excludes food and energy, rose 2.8% annually, still well above the Fed’s 2% target. The following day, April PPI came in at 6% year-over-year, the fastest pace since 2022.

The CME FedWatch tool, which tracks fed funds futures, shifted sharply. By midweek, traders were pricing a 25% probability of a rate hike by year-end, up from roughly 21% the prior Monday. That is not a majority view, but it is a meaningful one. A market that entered 2026 expecting two or three cuts this year is now debating whether the next move is tighter. That is the rate context in which SpaceX is preparing a June roadshow, and in which OpenAI is contemplating a September one.

“Today’s inflation report is certainly another nail in the coffin of the idea Fed officials have to welcome the new Fed Chair with an interest rate cut this year.”<
— Chris Rupkey, Chief Economist, FWDBONDS, May 12, 2026

Warsh Inherits a Fed That Cannot Give Him What He Wants

The timing could not be more complicated. Kevin Warsh was confirmed as Federal Reserve Chair by the Senate on May 13 in a 54-45 vote, the most contested confirmation in the institution’s history, and was sworn in on May 15 as Jerome Powell’s term expired. Warsh had been nominated in part on the argument that AI-driven productivity gains would allow the Fed to ease without reigniting inflation. The data released the week of his confirmation made that argument harder to sustain immediately. The Fed funds rate has been held steady at 3.5% to 3.75% since December, and Warsh’s first FOMC meeting as chair is scheduled for June 16-17, coinciding almost exactly with SpaceX’s anticipated roadshow open.

The bond market’s message to the incoming chair has been blunt. Ed Yardeni of Yardeni Research, cited in CNBC reporting the week of Warsh’s confirmation, noted that the 2-year Treasury yield sitting above the federal funds rate is a signal that the bond market believes the current policy rate is not high enough to contain inflation. That configuration rarely resolves without either the Fed tightening or inflation breaking on its own. Neither outcome is clearly in sight before the SpaceX roadshow begins.

What 4.5% Actually Does to a 95x Revenue Multiple

The mechanics are not complicated, but they are worth stating precisely. A discounted cash flow model for a pre-profitability company is extremely sensitive to the risk-free rate. When the 10-year sits at 4.0%, a company with plausible long-run margins and strong revenue growth can support a very high current multiple because the terminal value, discounted back, is large relative to near-term cash flows. When the 10-year sits at 4.6%, that same terminal value shrinks. The math does not change the business; it changes what the business is worth today.

SpaceX’s S-1 implies a consolidated revenue multiple of roughly 95 to 107 times its 2025 revenue of $18.7 billion at the reported $1.75 to $2 trillion valuation range, according to analysis published alongside the filing. OpenAI’s reported target of above $1 trillion implies a multiple above 75 times estimated 2025 full-year revenue. These are not multiples that compress gracefully when the discount rate moves 40 to 60 basis points. The Starlink segment of SpaceX generates real operating income and partially anchors the valuation, but the AI and enterprise applications segments that carry most of the implied value are loss-making and speculative by any conventional measure. That is the portion of the valuation most exposed to a rate move of the magnitude the market has seen since February.

This Has Happened Before — Just Not at This Scale

The geopolitical origin of the current rate spike is a useful point of reference. When the Iran conflict began in late February, markets initially read it as an IPO window-closing event, and they were right. The IPO market effectively froze through March, with only four U.S. companies pricing above $1 billion in Q1. What has happened since is more complicated: markets recovered in April and May, the SpaceX S-1 dropped on May 20, and the surface-level reading is that the window reopened. But the rate environment that accompanied the recovery is meaningfully tighter than what existed before the conflict, because the energy price shock translated directly into the CPI and PPI readings that triggered the May bond selloff.

The 2022 parallel that rate strategists keep reaching for is instructive but imperfect. In 2022, the Nasdaq fell 33% peak to trough as the 10-year moved from 1.5% to 4.3%. The starting multiple in tech was higher than today, and the rate move was vertical. The 2026 move has been a slower grind, from roughly 4.2% in February to 4.6% in May, and tech earnings have been genuinely strong throughout. Nvidia reported $81.6 billion in Q1 FY2027 revenue on May 20 with 85% year-over-year growth. The underlying businesses are not in the same distress as 2022 growth stocks. But the multiple compression logic is the same, and it applies with particular force to companies that are still building the earnings that will eventually justify the valuation investors are asked to pay on day one.

The Roadshow Calendar Is Now a Rate Calendar

SpaceX’s planned June roadshow open falls between Warsh’s first FOMC meeting on June 16-17 and whatever market signal that meeting produces. If Warsh signals a hawkish tilt, or simply removes the easing bias the FOMC has carried since December, the 10-year could test 4.7% or higher during the active book-build period. That is not a scenario that kills the SpaceX IPO — the institutional demand for a company of this visibility is different from the demand that supports a mid-cap growth listing — but it is a scenario that affects the clearing price and the retail allocation math. The S-1 explicitly lists Schwab, Fidelity, and Robinhood in the selling group, targeting retail investors at the same price as institutions. That retail participation makes the rate sensitivity more direct: individual investors buying into a June IPO at a 100x-revenue multiple are doing so with a 4.6% risk-free alternative sitting one click away in a money market fund.

OpenAI’s situation is more exposed. A September listing target gives the company four additional months of inflation data, two more FOMC meetings, and presumably the early market signal from SpaceX’s aftermarket trading. If SpaceX prices cleanly and holds, it opens the window for OpenAI to follow. If SpaceX prices and struggles — particularly in the AI and enterprise segments where the multiples are least defensible on current financials — OpenAI’s bankers at Goldman Sachs and Morgan Stanley will have a harder conversation with the company about whether September is realistic. CFO Sarah Friar has already expressed reservations about the readiness of the business for public markets. A rate environment that compresses terminal values would add external evidence to that internal caution.

Anthropic’s October target sits furthest out, which is either a structural advantage or a prolonged exposure to a deteriorating rate environment, depending on which way the next four months resolve. The company’s reported $900 billion valuation and above-70% gross margin profile give it a different argument than either SpaceX or OpenAI. But it is still pricing into a market where the risk-free rate may be 50 basis points higher than the market expected in January, and where the Fed’s posture under a new chair remains genuinely uncertain.

The IPO wave is real. The businesses behind it are larger, better capitalized, and more operationally mature than anything that came before them at this scale. None of that changes the fact that the rate environment they are walking into is the most complicated since the 2022 growth selloff, and that the specific mechanism driving rates — an energy-driven inflation spike from a geopolitical conflict with no clear resolution timeline — is one the Federal Reserve has limited ability to directly control. Warsh’s first months in the chair will do as much to determine the final pricing on these deals as any roadshow deck.

What to Watch Next

• The June 16-17 FOMC meeting under Warsh. His first statement as chair will either confirm or retire the easing bias the committee has carried since December. Any language suggesting rate hikes are back on the table would immediately reprice the SpaceX book-build environment and could force a delay or valuation adjustment before pricing.
  
• May CPI, due mid-June. If the May report shows April’s energy-driven spike beginning to moderate, it gives the roadshow a cleaner backdrop. If energy costs remain elevated above $4.50 at the pump and the CPI print comes in above 3.5%, the institutional allocation for SpaceX will tighten and the retail participation story becomes harder to tell.
  
• SpaceX first-day trading and aftermarket performance through July. The pricing and early trading will be the first real-world test of whether institutional investors will absorb a 95x-revenue multiple on a net-loss company in a 4.6% rate environment. That answer sets the risk appetite for OpenAI’s September roadshow more directly than any other variable.
  
• OpenAI’s internal revenue trajectory through Q2. The company has missed internal revenue and user growth targets at points in 2026. If Q2 data, which will be disclosed in the confidential S-1 review process, shows reacceleration, it gives the company and its bankers the evidence needed to hold the $1 trillion target. If growth is flat or slowing, the September window compresses toward Q4 regardless of rate conditions.
  
• The 30-year yield relative to 5.2%. The 30-year touched 5.2% during the May bond rout, its highest since 2007. Sustained trading above that level would signal that the term premium is repricing structurally, not just in response to a geopolitical spike. That would be the most consequential rate signal for anyone modeling the terminal values that underpin the three mega-IPO valuations.

Cerebras Systems raised $5.55 billion on Nasdaq on May 14, pricing 30 million Class A shares at $185 each — more than double the original marketed range of $115 to $125 — after an order book that closed roughly 20 times oversubscribed. The stock opened at $350, hit an intraday high of $386.34 that briefly triggered a trading halt, and closed the first session up 68% at $311.07. By the following day it had given back more than 10%. It is the largest U.S. tech IPO since Uber in 2019. It is also one of the most concentrated customer bases ever presented to public market investors in a semiconductor offering.

What the S-1 Actually Says

Cerebras filed its S-1 registration statement on April 17, 2026, and amended terms on May 4. The prospectus discloses $510 million in 2025 revenue, up 76% from $290 million in 2024. Revenue has grown from $25 million in 2022 — a compelling trajectory by any measure.

The profitability picture, however, requires a close reading. Cerebras reported net income of $87.9 million in 2025 on a GAAP basis, swinging from a $484.8 million net loss in 2024. The swing is real, but it was driven primarily by a $363.3 million non-cash gain tied to extinguishing a forward-contract liability connected to G42’s original investment. Strip that out and the company posted an operating loss of approximately $146 million. Gross margins improved materially — from roughly 12% in 2022 to the low 40s by 2025 — and the direction of travel is genuine. But the company that went public on May 14 is not yet operationally profitable, and investors who read only the headline net income figure bought a different story than the one in the footnotes.

The customer concentration is worse than the summary disclosure suggests. G42, the Abu Dhabi AI conglomerate that once accounted for 85% of Cerebras revenue in 2024, has been reduced to 24% of 2025 revenue — a figure the roadshow presented as evidence of diversification. But the Mohamed bin Zayed University of Artificial Intelligence, which the prospectus identifies explicitly as a G42 related party, accounted for 62% of 2025 revenue. Together the two UAE-linked entities represented 86% of revenue. MBZUAI alone accounted for 77.9% of accounts receivable as of December 31, 2025. The S-1 states directly: “The loss of, any substantial reduction in sales to, or the default on payments by, any of our significant customers would harm our business, financial condition, results of operations, and prospects.”

What the roadshow called customer diversification was, in substance, a revenue reallocation between two connected entities in the same country. Public investors absorbed that risk at a valuation that briefly touched $100 billion on a fully diluted basis.

“There’s some whales out there, there’s some really big customers. That is one of the characteristics of this market.”<
— Andrew Feldman, CEO, Cerebras Systems, CNBC interview, May 14, 2026

The OpenAI Relationship Is the Bull Case — and the Complication

The deal that changed the Cerebras story enough to bring the IPO back from a 2024 withdrawal is the $20 billion Master Relationship Agreement with OpenAI, covering 750 megawatts of AI inference compute capacity with an option to expand to 2 gigawatts by 2030. That agreement, signed in January 2026, is the single most important commercial development in Cerebras’s history. It converts the company from a sovereign AI supplier for the UAE into something that at least resembles a hyperscaler-grade infrastructure business.

The structure of the arrangement, though, carries its own complexity. OpenAI advanced Cerebras a $1 billion working capital loan in January at 6% annual interest, secured by warrants to purchase up to 33.4 million Cerebras shares at a near-zero exercise price. Amazon Web Services holds similar warrants as part of a March 2026 term sheet that would make AWS the first hyperscaler to deploy Cerebras hardware in its own data centers. At the $350 opening price, the OpenAI warrants alone would be worth approximately $11.7 billion if fully vested — a stake that gives OpenAI substantial economic interest in Cerebras’s public market performance while simultaneously being its largest commercial customer. That relationship is not arms-length, and investors in CBRS are underwriting it whether they know it or not.

The backlog figure is genuinely unusual. Total remaining performance obligations — contracted revenue not yet recognized — stood at $24.6 billion against $510 million in trailing revenue, a ratio of roughly 48 times. D.A. Davidson analyst Gil Luria pegged a fair value near the backlog level and suggested a target around $115 per share, well below the IPO price. Renaissance Capital flagged the valuation as “quite high even out to 2028.” Neither view stopped the book from going 20 times covered.

The Arm Approach Nobody Expected

The morning of May 13, the day before trading began, Bloomberg reported that Arm Holdings and its parent SoftBank had made a preliminary acquisition approach to Cerebras in the weeks prior. Cerebras rejected it. Representatives for all three companies declined to comment.

The approach is significant in both directions. For Arm and SoftBank, it signals strategic ambition beyond the chip architecture licensing model — a push toward owning full-stack AI silicon capability at a moment when the company was shaping its own public market pitch. For Cerebras, the decision to reject the approach and proceed to public markets at a valuation that briefly cleared $100 billion on a fully diluted basis is a clear statement about how the company’s leadership assesses its independence value versus its acquisition value. The $350 opening price validated that judgment in real time, at least on the first day.

As we wrote in April when the S-1 was first filed, the central question for Cerebras was never whether the wafer-scale technology works. It is whether inference-specialized silicon can hold a durable position against Nvidia’s platform — CUDA, networking, software, and enterprise distribution — plus AMD’s renewed data-center push and Google’s and Amazon’s custom ASIC programs, all in a market where Nvidia still commands an estimated 81% share of AI accelerator spend.

What the First Day Taught Institutional Investors

The 68% first-day pop followed by a 10%-plus decline on day two is a familiar pattern for high-profile AI offerings — strong institutional demand at pricing, retail enthusiasm at the open, and then a valuation reality check once the lock-up calendar and customer concentration get priced properly. The stock traded as low as $275 during the first two days, giving back more than a third of the first-day closing gain before stabilizing.

The more useful signal for the investment case is the backlog-to-revenue ratio and what it requires. For CBRS to justify even its IPO price of $185 at a reasonable forward multiple, Cerebras needs to convert a substantial portion of that $24.6 billion in remaining performance obligations into recognized revenue on an accelerated timeline, do so with improving gross margins, and reduce its operating losses while simultaneously funding the infrastructure buildout that the OpenAI and AWS agreements require. Each of those variables is subject to the pace of OpenAI and AWS deployment decisions — which are in turn subject to their own capital allocation priorities and competing infrastructure commitments.

The hyperscaler capex context matters here. As we covered in our analysis of the $690 billion infrastructure cycle, the 2026 buildout is running at historically unprecedented levels. That is the demand tailwind Cerebras is riding. It is also the environment in which Nvidia, AMD, and custom ASIC programs from Google, Amazon, and Microsoft are all competing for the same infrastructure dollars. The tailwind does not guarantee Cerebras a share of it.

For the broader IPO pipeline, the Cerebras debut serves the function we anticipated when SpaceX’s confidential filing dropped in April: it tests public market appetite for AI infrastructure stories at private-market valuations before the much larger SpaceX and OpenAI offerings arrive. The result — massive demand at pricing, a 68% first-day pop, then immediate pressure on valuation — suggests the market has appetite for AI chip stories but not unlimited patience for the concentration risk hiding inside them. That is useful data for anyone pricing the next one.

What to Watch Next

• Cerebras’s first earnings report as a public company. The S-1 disclosed trailing revenue and a backlog; the quarterly print will show whether OpenAI and AWS volume is converting into recognized revenue on the timeline the prospectus implies. Any revenue shortfall against the backlog pace will reprice the stock materially.
  
• OpenAI’s deployment disclosures. The $20 billion MRA commits OpenAI to purchase capacity, but delivery is staged and the terms allow for adjustment. Watch any OpenAI commentary on compute infrastructure for signals about whether the Cerebras relationship is expanding or being managed conservatively.
  
• AWS Bedrock integration timeline. The March term sheet put AWS in position to become the first hyperscaler deploying Cerebras chips in its own data centers. A formal launch date and any disclosed pricing would provide the first indication of whether the relationship generates commercial volume beyond the OpenAI account.
  
• Customer concentration metrics in subsequent filings. The 86% UAE revenue figure is the single variable that most constrains the institutional investor base willing to hold CBRS. Any quarterly filing showing Western enterprise revenue growing as a share of the total is the clearest path to multiple expansion.
  
• Lock-up expiration dynamics. The 180-day lock-up places insider and early investor selling pressure in the November 2026 window. Fidelity controls approximately 11% of shares and Benchmark holds approximately 9%. How the stock absorbs that potential supply will be the second-most-important pricing event after the first earnings report.

That chapter is closed. G42 has since been removed from Cerebras’s primary shareholder structure to satisfy U.S. regulators, and the company has spent the months since building a customer base that looks nothing like the one in that first S-1. In January 2026, Cerebras signed a $10 billion compute deal with OpenAI, pledging 750 megawatts of computing capacity through 2028. In March, it announced a partnership with Amazon Web Services to deploy its CS-3 systems inside AWS data centers, available through Amazon Bedrock. The company is now valued at $23.1 billion after a February Series H round and is targeting a roughly $2 billion raise on the Nasdaq, with Morgan Stanley as lead underwriter, according to Bloomberg.

The public S-1 has not yet been filed as of this writing. But the architecture of the deal — the timing, the customer lineup, the deliberate sequencing of announcements — reads like a company that understands exactly what a prospectus needs to say.

The Chip That Doesn’t Fit the Nvidia Model

To understand why Cerebras matters to investors, you need to understand why it is structurally different from every other AI chip company trying to go public right now. Nvidia’s dominant GPU architecture works by connecting hundreds or thousands of discrete chips — each physically small — through high-bandwidth memory and fast interconnect. The bottleneck in that approach is data movement: getting information from one chip to another, from memory to processor, fast enough to keep pace with the model’s demands.

Cerebras built the WSE-3 from the other direction. The chip is a single processor the size of an entire 300mm silicon wafer — roughly 56 times the physical area of Nvidia’s H100. It contains 4 trillion transistors, 900,000 AI-optimized cores, and 44 gigabytes of on-chip SRAM with 27 petabytes per second of internal memory bandwidth. Because the model weights live on the chip itself rather than in external memory, there is no bottleneck to solve. The machine simply runs faster — particularly in inference tasks where an AI system is generating responses to live queries, rather than training on new data.

The practical result: Cerebras delivered Llama 4 Maverick inference at more than 2,500 tokens per second per user on its CS-3 system, compared to roughly half that on Nvidia’s flagship DGX B200 Blackwell running the same 400-billion parameter model. For applications like agentic coding tools — where a developer is waiting for multi-step AI reasoning in real time — that difference is meaningful.

“Every customer large or small is on AWS, from individual developers to the largest banks in the world. The deal will make it easy as a click to get on Cerebras.”
— Andrew Feldman, CEO, Cerebras Systems, Reuters, March 13, 2026

The Amazon Deal Changes the Distribution Equation

For most chip startups, hardware reach is the hardest problem. You can build the fastest processor in the world and still lose if your customers can’t access it through the infrastructure they already use. The AWS partnership, announced March 13, addresses that directly. Under the arrangement, Cerebras CS-3 systems sit inside AWS data centers and operate alongside Amazon’s own Trainium3 chips in a so-called disaggregated inference architecture — Trainium handles the prefill stage of a query, Cerebras handles the decode. AWS calls the result five times the high-speed token capacity in the same hardware footprint. The service, running on Amazon Bedrock, is expected to launch in the second half of 2026.

The significance for Cerebras is distribution at a scale no startup can build independently. AWS serves customers ranging from individual developers to global financial institutions. When David Brown, Vice President of Compute and ML Services at AWS, said publicly that the Trainium-Cerebras solution will deliver “inference that’s an order of magnitude faster and higher performance than what’s available today,” that is not a press release formality. It is a co-endorsement from the world’s largest cloud provider, delivered weeks before an IPO roadshow.

Cerebras has also inked IBM and the U.S. Department of Energy as customers, alongside OpenAI, Cognition, and Mistral. The customer concentration risk that sank the first S-1 story has been structurally dismantled. The question is whether the new customer roster can support the valuation.

The Valuation Math Is Tight

At the $23 billion figure established in the February Series H, Cerebras would debut as one of the ten largest semiconductor IPOs in history, priced ahead of its current revenue. Estimated 2025 revenues exceeded $1 billion according to multiple analyst reports, but the company’s cost structure — proprietary water-cooled hardware, TSMC wafer manufacturing, and a software stack that requires developers to leave Nvidia’s CUDA ecosystem — is not cheap to operate.

The CUDA problem is worth understanding. Nvidia’s developer ecosystem is the deepest competitive moat in the chip industry. Tens of thousands of enterprise AI teams write code specifically for CUDA; switching to Cerebras’s software stack requires retraining and re-tooling. The company’s inference API — which lets developers access wafer-scale performance through a standard cloud interface without buying hardware — is designed to lower that barrier. But it does not eliminate it. For institutional investors pricing the IPO, the question is how many enterprise customers will opt for Cerebras performance at a premium over Nvidia compatibility at a discount.

The Amazon integration changes that calculus somewhat. If developers can access Cerebras hardware through a standard Bedrock API call — the same interface they already use for other AWS AI services — the switching cost drops considerably. That may be the single most important structural fact about the March 13 announcement, and it is likely to feature prominently in the S-1.

What the Second Attempt Gets Right

Cerebras has learned from the timing mistake of the first filing. The original S-1 landed in September 2024 into a national security review it could not resolve quickly. The company tried to wait it out, raised capital to extend its runway, and ultimately withdrew. This time, the regulatory pathway was cleared before the filing, the key customer relationships were announced in sequence — OpenAI in January, Amazon in March — and the underwriter was selected before the formal S-1 submission.

The IPO window for Q2 2026 is not guaranteed to stay open. Market volatility and macro uncertainty can compress or shut the calendar quickly, as the broader IPO market has demonstrated multiple times in the last 18 months. The company’s Nasdaq ticker reservation — CBRS — has been held since the first filing. Whether it gets used in April or slides to June will depend on when the public S-1 drops and how investor appetite looks after the bank earnings season that begins the week of April 13.

What is clear is that Cerebras is no longer asking the market to fund a technical bet on an unproven architecture. It is asking the market to value a company that OpenAI, Amazon, IBM, and the U.S. Department of Energy have already paid to use.

WHAT TO WATCH NEXT

• The public S-1 filing on SEC EDGAR — expected in late April or early May before any roadshow; it will contain the first audited revenue figures, cost structure, and TSMC manufacturing dependency disclosure.
• AWS Bedrock launch date for the Trainium-Cerebras disaggregated service — the second-half 2026 window is wide; an earlier-than-expected rollout would strengthen the IPO narrative heading into pricing.
• Nvidia’s response — Reuters reported Nvidia is expected to combine its own GPU chips with Groq (acquired for $17 billion in December 2025) in a similar disaggregated inference architecture. That announcement, if it arrives before Cerebras prices, directly affects how investors frame the competitive risk.
• OpenAI contract execution milestones — the $10 billion agreement runs through 2028, but delivery is staged; any disclosure of compute capacity actually deployed versus committed will be the most meaningful revenue signal in the S-1.
• TSMC wafer allocation — Cerebras uses nearly an entire 300mm wafer per chip and competes directly with Apple and Nvidia for TSMC manufacturing capacity; any tightening in that supply chain is a direct production risk.