The revenue is not speculative. Salesforce reported that its Agentforce platform had reached $800 million in annual recurring revenue as of its fiscal Q4 close on January 31, 2026, up 169% year-over-year, with 29,000 total deals closed — a 50% jump quarter-over-quarter. Combined Agentforce and Data 360 ARR exceeded $2.9 billion, more than doubling in a year. Microsoft’s AI business reached an annualized run rate approaching $13 billion as of its Q2 FY2026 results, with Azure growing 39% year-over-year and AI products cited as a primary driver. ServiceNow disclosed that its Now Assist agentic AI product surpassed $600 million in annual contract value, with net new ACV more than doubling year-over-year in Q4 2025, and management targeting over $1 billion in Now Assist ACV for 2026.

None of those numbers are projections. They are reported figures from primary earnings releases and official transcripts. The problem is not that investors don’t believe the revenue. It is that the frameworks traditionally used to value enterprise software — ARR multiples, seat counts, net revenue retention — were built for a model that agentic AI is actively dismantling.

What Agentic AI Actually Does to the Per-Seat Model

The SaaS era ran on a simple, elegant assumption: more employees meant more software subscriptions. A company with 500 salespeople needed 500 CRM seats. A company with 200 support agents needed 200 helpdesk licenses. Revenue expanded with headcount, multiples compressed around that predictability, and the entire asset class was priced accordingly. Salesforce, Workday, ServiceNow, and their peers collectively built hundreds of billions of dollars in market value on that formula.

Agentic AI severs the link between headcount and software spending. An AI agent that autonomously handles customer support tickets, drafts sales outreach, or manages IT workflows does not consume a human seat. It replaces several. The enterprise does not stop paying for software — in many cases it pays more, because agents that deliver measurable business outcomes command higher prices than productivity tools — but the unit of pricing has to change. Consumption-based models, outcome-based contracts, and flat-fee agentic enterprise licenses are all competing to fill the gap that per-seat pricing is leaving behind.

AlixPartners projected in its 2026 enterprise software report that usage- and outcome-based models will account for over 40% of AI software revenue by the end of this year, with AI-native companies commanding valuations roughly five to six times higher than their SaaS-era peers. The same report noted that ARR multiples, historically the dominant valuation method for enterprise software, are projected to explain significantly less of valuations as investors grapple with the costs and structural differences of AI delivery. The framework is not broken — it is simply incomplete.

The February Selloff Was a Pricing Argument, Not an Earnings Crisis

On February 3, 2026, the enterprise software sector erased roughly $285 billion in market capitalization in a single session, in what analysts quickly labeled the “SaaSpocalypse.” The proximate trigger was a cluster of agentic AI product launches — including autonomous agent tools capable of operating enterprise software interfaces directly — which prompted investors to reprice the seat-compression risk that had been a theoretical concern since 2024. Workday, ServiceNow, Salesforce, and Adobe all saw significant intraday declines despite reporting strong earnings in the weeks prior.

What made the selloff instructive was what it revealed about the valuation gap. Salesforce’s Q4 report, released February 25, beat expectations handily — Q4 revenue of $11.2 billion, up 12% year-over-year; adjusted EPS of $3.81 against a consensus estimate of $3.05; Agentforce growing at triple-digit rates. The stock still fell roughly 5% in after-hours trading. Bernstein maintained an underperform rating, citing what it called “a mature business in a mature market.” Goldman Sachs retained a buy, emphasizing that differentiated outcomes with Agentforce would be the key driver. The disagreement was not about the Q4 numbers. It was about whether the transition from per-seat revenue to outcome-based revenue would expand or compress total addressable market and long-term margin profiles — a question that a quarterly earnings release cannot fully answer.

“We’ve rebuilt Salesforce to become the operating system for the Agentic Enterprise. Agentic AI is a tailwind for our business, and we’re well on our way to $63 billion in revenue in FY30.”
— Marc Benioff, Chair and CEO, Salesforce, Q4 FY2026 Earnings Release, February 25, 2026

Microsoft’s Conversion Problem Complicates the Narrative

Microsoft’s AI story is structurally different from Salesforce’s, and that difference matters for how investors should read the valuation debate. Azure’s AI workload growth is unambiguous — 39% year-over-year, with AI contributing an estimated 13 to 16 percentage points of that growth rate. The commercial remaining performance obligation reached $625 billion as of Q2 FY2026, up 110% year-over-year, with roughly 45% of that driven by OpenAI commitments. By any infrastructure measure, demand is real and contracted.

The consumer-facing product story is more complicated. After eight consecutive quarters of declining to disclose a seat count, Microsoft finally confirmed in its January 28 earnings call that Microsoft 365 Copilot had 15 million paid seats — representing 3.3% of the 450 million commercial Microsoft 365 users. Growth was strong, up more than 160% year-over-year, but the absolute penetration rate raised questions about whether the flagship AI product was converting as quickly as the infrastructure investment required. Independent survey data from Recon Analytics found Copilot’s share of the paid AI subscriber market had dropped from 18.8% to 11.5% in the six months through January 2026, even as total paid seats grew.

CFO Amy Hood’s response on the earnings call was notable. She argued that judging AI spend solely on Azure or Copilot revenue “is the wrong yardstick” — that the infrastructure creates competitive positioning across Microsoft’s entire business that no single revenue metric can capture. That framing has merit. It is also the kind of argument that buys time while adoption numbers catch up. At a $150 billion annualized capital expenditure run rate — unprecedented in corporate history — the return horizon matters.

The Metrics That Actually Matter Now

For investors trying to evaluate enterprise AI companies with available frameworks, three signals from the current earnings cycle are worth tracking more carefully than headline revenue growth. The first is the ratio of new bookings coming from existing customer expansion versus new logos. Salesforce disclosed that more than 60% of Agentforce and Data 360 Q4 bookings came from existing customers expanding their commitments — a land-and-expand dynamic that suggests genuine deployment rather than trial-stage procurement. That pattern is harder to fake than a logo count.

The second is token and workflow volume as a proxy for production deployment versus sandbox testing. Salesforce’s disclosure that its platform processed nearly 20 trillion tokens cumulatively and delivered 2.4 billion agentic work units — tasks where AI was not reasoning but completing — is precisely this kind of metric. A customer processing tokens at scale is a customer with agents in production, which is a customer whose renewal probability is structurally higher than one still evaluating.

The third, flagged explicitly in PwC’s February 2026 analysis of AI’s impact on software valuations, is gross revenue retention disaggregated from net revenue retention. In a world where AI agents are reducing seat counts while AI add-ons are increasing per-account spending, NRR can look healthy while the underlying seat base is contracting. GRR stripped of AI upsell reveals whether the core product is holding or eroding — and that separation is the most direct test of how durable any given platform’s moat actually is.

The companies that have the clearest answers to those three questions — expansion ratios, production deployment evidence, and clean retention disaggregation — are the ones most likely to survive the valuation reset with multiples intact. The ones that can only point to ARR growth rates without that supporting evidence are the ones the market is right to reprice, regardless of how impressive the headline number looks.

WHAT TO WATCH NEXT

• Salesforce FY2027 Q1 earnings — the first quarter under new guidance; management committed to organic revenue re-acceleration in H2 FY27, and Q1 results will show whether early momentum supports that trajectory or reveals a gap between Agentforce deal closings and recognized revenue.
• Microsoft Q3 FY2026 earnings (expected April 29) — Azure guidance of 37–38% growth was provided for the quarter; any miss or commentary on capacity constraints will directly affect how the market prices AI infrastructure demand. The Copilot paid-seat trajectory will also be updated for the first time since the January disclosure.
• ServiceNow’s path to $1 billion in Now Assist ACV — management stated a clear target for 2026; quarterly progress reports will be the most specific test of whether enterprise agentic AI contracts are scaling as predicted or plateauing after early adopters.
• The pricing model transition — Salesforce’s Agentic Enterprise License Agreement, described as a flat-fee “all you can eat” structure for customers ready to scale, is the clearest real-world test of whether enterprises will pay more per account under outcome-based pricing than they did under per-seat models. Deal size data from quarterly reports will show whether the math works.
• The AlixPartners prediction on valuation frameworks — the consulting firm projected that hybrid AI leverage ratio models will begin replacing pure ARR multiples as the standard for enterprise software valuation by end of 2026. Watch for the first major analyst firm to publish a formal framework shift; that moment will likely move sector multiples.

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.

None of that is the story. The story is what investors don’t have yet.

A confidential filing — standard practice for large listings under the JOBS Act — allows SpaceX to work through SEC comments privately before its financials become public. Under SEC rules, the company must publicly file its full prospectus at least 15 days before the investor roadshow begins. With a June target, that prospectus is expected to land in April or May. Until it does, every valuation figure in circulation is a number derived from secondary market trades, analyst estimates, and sources who spoke to reporters anonymously. No audited revenue breakdown. No subscriber acquisition cost. No disclosed cash burn attributable to xAI. No formal cap table post-merger.

The Starlink Question Is the Only One That Matters

The $1.75 trillion valuation is not primarily a bet on rockets. Starlink ended 2025 with 9.2 million subscribers and generated over $10 billion in revenue, with analysts projecting that figure could reach anywhere from $15.9 billion to $24 billion in 2026 depending on subscriber trajectory and pricing mix. The wide range in analyst projections — a gap of more than $8 billion in a single year — is itself the problem. Nobody has seen the actual unit economics.

What does an average Starlink subscriber generate per month, net of satellite operations, customer support, and terminal subsidies? How much of revenue is consumer versus maritime versus aviation versus government? What is churn? These are the figures that underpin any credible discounted cash flow model, and right now none of them exist in public form. SpaceX’s total 2025 revenue has been estimated at $15 billion to $16 billion with roughly $8 billion in profit, but those are figures sourced from Reuters citing unnamed sources, not from audited financials.

For a company targeting a valuation higher than every S&P 500 constituent except Nvidia, Apple, Alphabet, Microsoft, and Amazon, the margin of uncertainty here is unusual.

The xAI Integration Adds an Entirely New Layer of Risk

In early February 2026, SpaceX completed an all-stock acquisition of xAI, Musk’s artificial intelligence startup, in a deal that valued SpaceX at $1 trillion and xAI at $250 billion — the largest merger in history by combined entity size. The stated rationale was building orbital data centers: using Starlink’s satellite constellation to move AI compute workloads into space, where solar power is abundant and cooling is free.

The concept is technically ambitious. Whether it is financially credible is something the prospectus will need to answer. At the time of the merger, xAI was burning roughly $1 billion per month, according to Bloomberg, as it raced to build out infrastructure against OpenAI and Google. That burn rate, absorbed by SpaceX at the moment of the filing, now sits inside the entity investors are being asked to value at $1.75 trillion.

“The valuation jump from $1.25 trillion at merger close to $1.75 trillion at filing reflects expectations around the combined entity’s space and AI ambitions — not disclosed financials.”
—TheStreet, April 2, 2026

Then there is the talent picture. xAI co-founders Jimmy Ba and Tony Wu departed in February, shortly after the merger closed. By early March, Zihang Dai and Guodong Zhang had also left. Of the 12 people who co-founded xAI with Musk in 2023, only two remain — Manuel Kroiss and Ross Nordeen. Musk publicly acknowledged on X that xAI “was not built right first time around” and needed to be rebuilt from its foundations. That statement, made weeks after the largest merger in history closed at a $250 billion xAI valuation, will almost certainly surface in the prospectus’s risk factors — and will need to be reconciled with the financial picture investors are presented.

The Governance Structure Has No Precedent

SpaceX’s governance heading into a public offering is unlike any company that has come before it. Musk simultaneously holds the CEO role at SpaceX, Tesla, and leads DOGE in a senior advisory capacity. Tesla disclosed that it invested $2 billion of shareholder money into xAI in its Q4 2025 earnings report — and then SpaceX acquired xAI days later. Legal analysts tracking the deal note that consummating a conflicted merger between founder-controlled private companies before going public defers — but does not eliminate — the scrutiny around related-party dynamics. That scrutiny will arrive when the S-1 is public and securities lawyers start reading it.

SpaceX is also reportedly considering a dual-class share structure that would preserve insider voting control, and plans to allocate up to 30% of shares to retail investors — roughly three times the typical allocation in a large IPO. The retail allocation is an unusual move that deserves attention: it broadens the investor base and generates enthusiasm, but it also introduces a class of shareholders with limited governance recourse under a dual-class structure.

What the 21-Bank Syndicate Actually Signals

SpaceX has lined up at least 21 banks for this offering, with Goldman Sachs, JPMorgan Chase, Morgan Stanley, Bank of America, and Citigroup in senior underwriting roles. The deal is internally codenamed “Project Apex.” A syndicate of this size indicates that demand management is already the primary concern — no single bank wants the concentration risk of placing $75 billion in a single offering. It also means the roadshow, when it happens, will be one of the most heavily orchestrated investor events in market history.

One detail worth noting for index-aware investors: Nasdaq recently updated listing rules in a way that could allow SpaceX to join the Nasdaq 100 within 15 days of listing. That would trigger forced buying from index-tracking funds — a mechanical demand wave that has nothing to do with the underlying fundamental case. Investors who intend to evaluate SpaceX on its merits should be aware that initial price behavior after listing may reflect index inclusion mechanics more than actual price discovery.

Stacking Trades has covered the broader conditions shaping the IPO market this spring, including how macro headwinds in early 2026 rattled the IPO window — context that matters when assessing whether a $1.75 trillion listing can clear the market in June regardless of how good the underlying story is.

The Number That Investors Actually Need

The prospectus, when it arrives, will contain information that no source has yet provided: audited revenue by segment, Starlink subscriber economics, the precise mechanics of the xAI acquisition and how xAI’s assets and liabilities are carried on SpaceX’s balance sheet, the terms of the dual-class structure, and the risk factors Musk’s lawyers have deemed material enough to disclose to the investing public.

That document will be the first moment at which a credible, independently verified number can be placed next to the $1.75 trillion figure. Until then, every model being built against that valuation is working from incomplete information. That does not make the offering unattractive. It means the starting gun has been fired — and the race to price it properly has not yet begun.

WHAT TO WATCH NEXT

• The public S-1 prospectus: Expected in April or early May. The Starlink revenue breakdown, per-subscriber economics, and xAI integration terms are the first numbers worth modeling against the $1.75 trillion target.
• xAI’s disclosed cash burn post-merger: At roughly $1 billion per month at acquisition close, any updated figure in the prospectus will materially affect the combined entity’s free cash flow profile.
• Dual-class share structure details: The voting power Musk retains — and what that means for board independence and related-party transaction governance — will be the section securities lawyers read first.
• Nasdaq 100 inclusion timing: If confirmed at listing, forced buying from index funds could distort early price discovery. Watch for Nasdaq’s formal ruling.
• OpenAI and Anthropic IPO timelines: Bloomberg has reported both companies are weighing public offerings in 2026. SpaceX’s pricing and reception will set the reference point for that entire class of mega-listings.

The timing is not coincidental. Every major technology company with a balance sheet is currently trying to solve the same problem: where do you get gigawatts of firm, always-on, carbon-free power for AI data centers that run every hour of every day? Renewables can’t do it alone. Natural gas doesn’t satisfy the zero-carbon commitments most of these companies have made. That leaves nuclear — specifically, a new generation of smaller, faster-to-build reactors that don’t require the decade-long permitting fights of their 1970s predecessors. X-energy is making a direct bid to be that answer.

What X-Energy Actually Builds

Founded in 2009 by aerospace entrepreneur Kam Ghaffarian and now led by J. Clay Sell — a former U.S. Deputy Secretary of Energy — X-energy designs the Xe-100, a high-temperature gas-cooled small modular reactor that produces 80 megawatts of electricity per unit, scalable to a four-pack generating 320 megawatts. The reactor runs on TRISO-X fuel, a proprietary tri-structural isotropic particle fuel the Department of Energy has called “the most robust nuclear fuel on Earth.” Unlike conventional light-water reactors, TRISO fuel cannot melt down in the traditional sense — the ceramic coating on each fuel particle provides its own containment.

The reactor design’s key commercial appeal is modularity. X-energy describes its Xe-100 as road-shippable, with accelerated construction timelines and more predictable costs than conventional nuclear. Whether that promise holds when the first reactor actually goes into the ground at Dow Inc.’s Seadrift, Texas site remains to be seen. But it is the premise on which the company has assembled a remarkable roster of partners.

Amazon, Dow, and the Orderbook Behind the Filing

The commercial logic of X-energy’s IPO rests heavily on two anchor relationships. The first is Dow Inc., which signed a joint development agreement in 2023 to deploy a four-reactor Xe-100 plant at its Seadrift Gulf Coast chemical complex — the first grid-scale advanced nuclear reactor deployed to serve an industrial site in North America. The NRC docketed the construction permit application in May 2025 and published an 18-month review timeline, with environmental review running concurrently. Construction is expected to begin in 2026.

The second is Amazon, which anchored a $500 million Series C-1 round in October 2024 through its Climate Pledge Fund. Beyond the equity investment, Amazon has options to deploy more than 5 gigawatts of Xe-100 projects across the United States by 2039, beginning with the Cascade Advanced Energy Facility in Washington state, built in partnership with utility Energy Northwest adjacent to the Columbia Generating Station. A separate agreement with UK energy firm Centrica outlines plans for roughly 6 gigawatts of Xe-100 capacity in Britain.

X-energy’s total disclosed orderbook now exceeds 11 gigawatts across approximately 144 advanced reactors, according to the company’s own disclosures at the time of its Series D close. That is a pipeline, not a backlog — no revenue flows until reactors are built and operating. But for a pre-commercial nuclear company, having Amazon and Dow as co-development partners is a categorically different risk profile than having nothing but a design.

“X-energy’s technology will help satisfy historically unprecedented electricity demand growth, driven by the development of AI and associated data center infrastructure.”
—X-energy, Inc. S-1 registration statement — March 20, 2026

The Federal Backstop That Most Investors Miss

X-energy is one of two companies selected by the Department of Energy in 2020 as winners of the Advanced Reactor Demonstration Program, which provides up to $1.2 billion in federal cost-share matching to develop, license, and demonstrate an operational advanced reactor and fuel facility. By December 31, 2025, the company had received approximately $438 million in ARDP reimbursements, effectively subsidizing a significant portion of its development spend.

The TRISO-X fuel fabrication facility in Oak Ridge, Tennessee — called TX-1 — received an NRC Special Nuclear Material License in February 2026 and is designed to produce enough fuel to support the first 11 Xe-100 reactors at steady state. In April 2025, TRISO-X received the first allocation of high-assay low-enriched uranium, or HALEU, from the DOE’s HALEU Availability Program, securing fuel feedstock for the initial core loads of the Texas project. These are not minor milestones. The fuel chain is the single hardest logistical problem in advanced nuclear deployment, and X-energy has meaningfully de-risked it relative to most competitors.

The Financials: What the S-1 Shows

This is where the story requires discipline. X-energy reported $109 million in revenue and grant income for 2025, down 9% from the prior year. That decline matters. Stripping out grants, revenue from operations was approximately $94 million, against a net loss of roughly $390 million, per Bloomberg’s reporting on the S-1. The prior year’s net loss was $126 million on $84 million in revenue. The loss nearly tripled year-over-year as development spending accelerated.

The company employs roughly 916 people as of March 2026. It is in what Seeking Alpha’s IPO Edge analysis called “a high cash burn phase, with sharply rising losses, heavily negative cash flow.” That is accurate and investors should sit with it. Pre-commercial nuclear development at this stage of licensing and construction activity is extraordinarily capital-intensive. The IPO proceeds are intended to fund continued reactor licensing, construction activities, and supply chain expansion — not to make the company profitable in the near term. Profitability is a post-2030 story contingent on the Seadrift plant coming online.

 KEY RISKS FOR INVESTORS

Regulatory timeline risk. The NRC’s 18-month review of the Seadrift construction permit runs through approximately late 2026. Any extension or public comment complication pushes the construction start and, consequently, first revenue from operations.

Technology execution risk. No Xe-100 has ever been built. The reactor design is sound on paper and well-advanced in licensing, but first-of-a-kind construction always surfaces cost and schedule surprises.

HALEU supply chain risk. High-assay low-enriched uranium has historically been sourced primarily from Russia. X-energy received its first domestic HALEU allocation in April 2025, but fuel supply chain security for a scaled fleet remains a geopolitical dependency.

Valuation uncertainty. Share price and total raise are undetermined. Investors are being asked to price a company whose revenue case depends entirely on commercial deployments that won’t generate meaningful income until the early 2030s.

Why the Power Demand Argument Is Real

The thesis underneath X-energy’s IPO is not difficult to understand. Goldman Sachs estimates that data center electricity demand could rise 160% by 2030. The Federal Energy Regulatory Commission projects U.S. data center electricity demand will climb from 19 gigawatts in 2023 to 35 gigawatts in 2030. Goldman Sachs has further estimated that 85 to 90 gigawatts of new nuclear capacity would be needed to meet all projected data center power demand growth by 2030 — and less than 10% of that will be available globally by then.

That gap is the market X-energy is building into. It is not a manufactured thesis. Microsoft signed a 20-year power purchase agreement to restart Three Mile Island. Google contracted with Kairos Power for a fleet of SMRs targeting 2030 deployments. Meta issued an RFP for one to four gigawatts of new nuclear generation. These are not exploratory conversations. They are signed agreements and capital commitments by companies with unlimited energy budgets and existential reasons to secure reliable baseload power. X-energy has two of those companies as equity investors and development partners.

What This IPO Is Really Selling

X-energy is not a bet on current earnings. It is a bet on a structural energy transition that is already underway, on a technology that has cleared more regulatory hurdles than any other advanced reactor design in the U.S., and on a management team that includes a former Deputy Secretary of Energy and a CFO who ran corporate development for Amazon’s climate investment portfolio. Those are credentialed people with institutional backing executing against a plan that has federal cost-share and Fortune 500 anchor partnerships behind it.

The risk is time. Nuclear projects that slip slip expensively. If the Seadrift NRC review extends, if TX-1 fuel production encounters delays, or if HALEU supply chain constraints re-emerge, the cash burn against a thin revenue base will pressure the company in ways that IPO proceeds alone may not fully cushion. Investors who understand that this is an infrastructure company in the licensing phase — not a software company with a growth multiple — are the appropriate audience for this offering.

The roadshow timeline is not yet public. Based on standard SEC review periods and the March 20 filing date, trading could begin as early as May 2026. The amended S-1, which will carry the price range, is the next document to watch.

WHAT TO WATCH NEXT

• The amended S-1 with pricing terms. The first filing carried no share count or price range. The amended filing will be the first hard data point for valuation analysis and the clearest signal of where institutional demand is pricing the offering.
• NRC review progress on the Seadrift permit. The 18-month review clock started in May 2025, putting a decision around late 2026. Any NRC request for additional information or public hearing triggers would extend that timeline and reprice the risk of first commercial operations slipping past 2030.
• TX-1 fuel facility milestones. The Oak Ridge TRISO-X facility received its NRC license in February 2026. Watch for announcements of commercial fuel production beginning — that milestone converts the fuel supply thesis from regulatory to operational.
• Comparable nuclear IPO performance. X-energy will price into a market that has seen NuScale Power face cost overruns and schedule delays. How institutional investors price the sector premium or discount relative to NuScale’s public market experience will shape the deal’s reception.
• Tech company nuclear procurement announcements. Any new hyperscaler power purchase agreement or SMR development partnership announced in the pre-roadshow window will serve as real-time demand validation for X-energy’s commercial thesis.