Two giants landing in the public market simultaneously will test whether investors are willing to bear the dual pressure of a trillion-dollar valuation and ongoing losses for the AI infrastructure narrative.

Written by: BiyaNews

On May 20, 2026, SpaceX officially submitted its S-1 filing to the U.S. Securities and Exchange Commission (SEC), disclosing audited financial data after the merger with xAI. Almost within the same time frame, OpenAI was also preparing to secretly submit its IPO draft. The two giants reshaping the aerospace and artificial intelligence industries are testing a question not yet answered by the public market: Are investors willing to support the AI infrastructure narrative with a trillion-dollar valuation?

Two prospectuses, same market window

SpaceX secretly submitted its S-1 on April 1, 2026, and publicly disclosed it on May 20. The document shows that following the all-stock acquisition of xAI completed in February 2026, the merged entity achieved revenue of $18.67 billion in 2025, with a net loss of $4.94 billion and adjusted EBITDA of $6.58 billion. According to Bloomberg and Reuters, its IPO valuation target range is between $1.75 trillion and $2 trillion, with a financing scale potentially reaching $75 billion. SpaceX plans to go public on the NASDAQ on June 12, under the stock code SPCX, with Goldman Sachs as the lead underwriter, and Morgan Stanley, Bank of America, Citi, and JPMorgan Chase participating jointly.

Regarding OpenAI, the financing round in March 2026 was completed with a post-money valuation of $852 billion, with participation from institutions such as Amazon, NVIDIA, and Softbank. CFO Sarah Friar confirmed in a January 2026 blog that the annualized revenue run rate at the end of 2025 had surpassed $20 billion, but the actual revenue for the full year is estimated to be around $13.1 billion. This difference is crucial: the audited full-year financial data will be publicly disclosed for the first time in the S-1. On May 18, the jury ruled that Elon Musk's lawsuit against OpenAI was dismissed due to the statute of limitations in California, and Musk has announced an appeal, but this ruling temporarily removes the legal shadow over the IPO process.

Comparison of financial data: the divide between audit and estimate

The above table summarizes the audit data from SpaceX's S-1, the statements confirmed by OpenAI's CFO, and estimates from industry analysis agencies. OpenAI's forward-looking revenue projections are based on a year-end ARR base of $20 billion, not derived from the submitted prospectus. Both companies' IPO valuation targets are reported numbers by the media, and the final issuance terms will be disclosed in their respective public S-1s.

The four panels reveal structural differences between the two IPOs. Panel 1 distinguishes between audited data and estimated data: SpaceX's revenue of $18.67 billion in 2025 comes from the S-1 audit, while OpenAI's $13.1 billion is an estimate of actual full-year revenue, and the year-end annualized run rate of about $20 billion comes from CFO Friar's disclosure in January. Panel 2 shows that at the midpoint of the reported valuation range of $1.75 trillion to $2 trillion, SpaceX would rank just behind NVIDIA ($5.2 trillion), Alphabet ($4.8 trillion), Apple ($4.3 trillion), Microsoft ($3.1 trillion), and Amazon ($2.9 trillion). Panel 3 indicates that SpaceX's consolidated net loss mainly comes from its SpaceXAI business line, while Starlink and launch services together contributed $15.5 billion in revenue and have achieved profitability as independent operations. Panel 4 shows that OpenAI's estimated cash burn trajectory (expected to reach $27 billion in 2026 according to Sacra) diverges significantly from its revenue growth curve, with breakeven not expected until 2029 to 2030.

What each company is selling

SpaceX's prospectus is essentially a listing document for Starlink, accompanied by capital-intensive assets. The satellite internet business contributed $11.4 billion in 2025, accounting for 61% of consolidated revenue, with subscription users exceeding 10 million by February 2026. Quilty Space estimates that Starlink users will grow from 9 million at the end of 2025 to 16.8 million by the end of 2026. Launch services contributed $4.1 billion, supported by a $5.9 billion Pentagon NSSL Phase 3 contract through 2029. Both of these businesses are contract-supported, with clear growth paths visible.

The issue lies with SpaceXAI. This business line generated $3.2 billion in revenue in 2025, but recorded an operating loss of $2.47 billion in just the first quarter of 2026, becoming a major drag on consolidated profitability. Anthropic is purchasing the computing power of the SpaceX Colossus 1 data center at a price of $1.25 billion per month, with the contract extending to 2029, providing a well-known anchor customer, but this business line is still in the early stages of capital consumption. The risk factor section of the S-1 acknowledges that the Grok chatbot is facing investigations from eight regulators regarding unauthorized synthetic images.

OpenAI presents a starkly different value proposition. The company has about 900 million weekly active users and 50 million paid subscribers, with an annualized revenue run rate growing from approximately $6 billion in 2024 to over $20 billion by the end of 2025. According to Bloomberg, the company internally expects this speed to continue until achieving over $280 billion in revenue by 2030. Structural resistance is also significant: Microsoft holds about 27% of its post-conversion shares following a capital restructuring in October 2025 and secures rights through a revenue-sharing agreement with payments capped at $38 billion, which, although renegotiated, is still expected to require approximately $6 billion in payments in 2026 according to Sacra. OpenAI's gross margin is constrained by the costs of inference computing, maintaining around 33%, with inference computing costs projected to reach $14.1 billion in 2026. Cash flow breakeven is not expected before 2029.

The debate that investors need to resolve is not whether these companies are growing, but what the public market is being asked to capitalize: the operating profits of Starlink, the infrastructure consumption of xAI, the revenue speed of OpenAI, and the economics of large-scale AI computing in the future.

Capital market background: unprecedented financing scale

The scale of these offerings has no direct comparable objects in the history of modern capital markets. Saudi Aramco's IPO in 2019 raised about $26 billion and remains the largest single IPO to date. The financing target reported in SpaceX's document reaches up to $75 billion, which, if completed at the upper end of the reported range, would be nearly three times that of Saudi Aramco. Goldman Sachs is leading both transactions, and according to NBC News's review of the documents, the SpaceX transaction involves a total of 23 banks and investment institutions.

There are identifiable headwinds in the macro background. Federal Reserve policies and high long-term discount rates are compressing the terminal value assumptions for unprofitable growth companies—which is precisely the category in which SpaceX's AI business line and OpenAI fall. OpenAI's financing round in March 2026 valued at $122 billion indicates that institutional investors have demand at private market valuation levels. However, converting this demand into a higher reported IPO target in the public market requires successful roadshow execution and a prospectus narrative that sufficiently addresses profitability timelines.

The tension in valuation multiples is the analytical bridge missing from both documents. SpaceX's reported IPO range implies a multiple of about 95 to 107 times the 2025 revenue on a consolidated basis, significantly higher than any comparable aerospace or satellite company. OpenAI's reported target of over $1 trillion implies a multiple of approximately 75 times the estimated total revenue for the year 2025. Both multiples are difficult to defend solely through traditional profitability yield analysis. Both require investors to accept: that the connectivity of AI infrastructure and the terminal value of AI software adoption are sufficiently large to justify the current losses being capitalized at trillion-dollar scales. The complete S-1 documents will be the first time these assumptions are subjected to the public market disclosure standards pressure test.

Market scenario and key points of attention

SpaceX's S-1 was publicly disclosed on May 20, 2026. Key data points in the prospectus that need to be scrutinized include: the unit economics of Starlink's subscription users and average revenue per user trajectory, capital commitments and timelines for spacecraft development, independent operating cash flow and breakeven path for SpaceXAI, and the dual-class stock and voting control clauses that will dictate governance post-IPO. The roadshow is expected to begin in the week of June 4, with a pricing target of June 11 and the first trading day on June 12. These timelines are for reporting purposes and have yet to be confirmed. The gap between the disclosed financial data and reported valuation ranges will be the focal point of investor debates during the roadshow.

For OpenAI, the secret submission means that the public S-1 will not be visible for about 60 to 90 days after submission; the earliest public prospectus is likely to be late July or August 2026. Key disclosures that investors need to see include: the terms of the Microsoft revenue-sharing restructuring and their impact on reported revenue and net income, governance rights and equity holding mechanisms of the OpenAI Foundation, capital expenditure timelines related to Stargate and other infrastructure commitments, and specific assumptions behind the company's profitability timeline. Musk's appeal will progress through the California appellate court system, and the importance of this procedural risk will gradually decrease in the absence of an initial injunction restoring uncertainty from previous lawsuits.

The ability of both companies to complete their IPOs according to the reported timeline will be a test of whether the liquidity depth of AI infrastructure equity is sufficient. The last valuations from both companies' private market financings differ from the public market clearing prices. If either offering price is significantly below the reported target range or performs poorly on the first day, the impact on other companies in the AI IPO pipeline could be significant—reports suggest Anthropic has already had early discussions with Goldman Sachs, JPMorgan Chase, and Morgan Stanley regarding its own IPO.

From a professional perspective, I believe the current market environment poses a dual test for such mega-scale IPOs: on one hand, investors need to digest unprecedented financing scales; on the other hand, the uncertainty of the Federal Reserve’s policy path subjects the terminal value discounting of growth stocks to greater pressure. Based on years of experience, capital of a similar scale surged during the Internet bubble of 2000, but the infrastructure construction costs at that time were far below the current investments required for AI computing power. Historical data shows that when IPO financing scales exceed 10% of the market's average daily trading volume, first-day and subsequent performance often faces greater volatility. Investors should focus on the feedback signals from institutional investors during the roadshow process and the discount extent of final pricing relative to the reported range, as this will be a direct reflection of market demand. Of course, investment decisions must consider individual circumstances; the market always contains uncertainty.