Microsoft Just Said a Lot About SpaceX

Four AI announcements in eight days made the structure of the AI market visible. Google has the consumer search-AI substrate locked through Apple. Anthropic is winning enterprise AI at scale. Microsoft just confirmed both by detethering from OpenAI. SpaceX's $28.5 trillion TAM, with 90 percent attributed to AI, prices against a market that has now been measured — and xAI sits two orders of magnitude behind Anthropic and OpenAI on revenue. The disclosure standard should require the decomposition.

On April 27, 2026, as jury selection began in the Musk v. Altman trial, Microsoft and OpenAI announced a restructured partnership. Microsoft's exclusive license to OpenAI's technology becomes non-exclusive through 2032. Microsoft will no longer pay revenue share to OpenAI. OpenAI will continue revenue-share payments to Microsoft through 2030, subject to a cap. The AGI clause that would have triggered partnership termination if OpenAI declared general intelligence is removed. OpenAI can serve products on any cloud provider rather than only on Azure.

Coverage read this as a commercial repositioning between two large companies managing a complicated relationship. That reading is correct. The reading that matters more is what the disclosures attached to the repositioning tell us about the AI market that Microsoft has been measuring for six years.

What the Detethering Disclosed

Microsoft has the most informed external view of OpenAI's actual economics that any party outside OpenAI has. Microsoft sees the workload patterns, the inference costs, the training-run economics, the capacity utilization, and the customer-deployment patterns at the granularity that running the infrastructure produces. Microsoft is detethering. The decision to release OpenAI from exclusivity, and to release Microsoft from exclusive build-out commitment for OpenAI's growth, was made with substantially more information about OpenAI's actual economics than OpenAI's investors have. The data-informed conclusion: exclusive build-out at the scale OpenAI's narrative implies falls short of the capital commitment threshold. Microsoft is staying open to a winner in both consumer and commercial AI applications. The detethering signals that the foundational work, the work whose economics actually justify capital commitment at scale, is happening at companies like Anthropic and Palantir.

Microsoft is also the best-positioned external observer of OpenAI's attempt to monetize through advertising. OpenAI rolled out ad-supported tiers in late 2025 and early 2026. CFO Sarah Friar projected $2.6 billion in annual ad revenue. The actual contribution as of April 2026 is approximately 0.5 percent of OpenAI's total revenue, per public reporting. Microsoft sees the workload patterns associated with ad-served versus subscription-served queries at the granularity that running OpenAI's infrastructure produces. Microsoft has parallel data through Bing search advertising, the Microsoft Copilot integration across Windows and Microsoft 365, and the LinkedIn advertising business. Microsoft's revenue-share arrangement with OpenAI through 2030 captures value from OpenAI's advertising revenue alongside subscription revenue, giving Microsoft direct stake in whether the advertising experiment succeeds.

Four Announcements in Eight Days

This was the fourth AI infrastructure announcement in the eight days running from April 20 to April 27, 2026. On April 20, Anthropic and Amazon announced a deepening partnership: $5 billion in immediate investment, up to $20 billion in milestone-tied additional capital, and a $100 billion AWS commitment from Anthropic over ten years, at a $380 billion Anthropic valuation. On April 24, Anthropic and Google announced an additional partnership: $10 billion in immediate investment, up to $30 billion in performance-tied additional capital, and 5 gigawatts of TPU capacity over five years, at a $350 billion valuation. Bloomberg subsequently reported secondary-market interest in Anthropic at valuations approaching $1 trillion. On April 27, Microsoft restructured its OpenAI relationship to non-exclusive licensing.

The four announcements describe three distinct AI markets, each with structural configurations the contemporary record has now made specific.

Three Markets, Three Configurations

The advertising-subsidized AI market is closed. Google paid Apple approximately $20 billion annually for default search placement because Apple's distribution captures premium-rate query volume. The Gemini arrangement announced January 12, 2026 extends that economics into the AI substrate through the same bilateral relationship. AI-as-search-substitute will route through the Apple-Google channel by default at the scale of 1.5 billion daily iPhone users. The advertiser infrastructure, the consumer distribution where ad-support is the established norm, and the targeting data infrastructure that make advertising-subsidized AI economically viable at scale all sit inside Google's existing position. The Mehta antitrust remedies in September 2025 prohibited exclusive default agreements but permitted the economic arrangements that produce comparable outcomes. The January 12 announcement fits those constraints. The market's outcome was determined before the AI transition began. OpenAI's attempt to capture share through ad-supported tiers is producing economically minor results against the CFO's $2.6 billion annual projection. The structural reading and the contemporary evidence point in the same direction.

The enterprise-and-developer AI market is being captured by Anthropic at scale that exceeds what most enterprise software companies have produced in their first decade. Anthropic's annualized revenue exceeded $30 billion as of April 2026, up from approximately $9 billion at the end of 2025 and approximately $1 billion at the end of 2024. That is 30x growth in 15 months into substantial absolute scale. Both Amazon and Google have committed up to $65 billion of capital to Anthropic at primary-market valuations between $350 billion and $380 billion, with secondary-market interest at multiples of those figures. Both Amazon and Google operate competing AI products. Both committed capital to Anthropic at scale anyway. Anthropic's revenue mix is approximately 80 percent enterprise, with over 1,000 customers spending $1 million or more annually as of April 2026, doubled from 500 in February. OpenAI sits at approximately $25 billion ARR, up from $20 billion at end of 2025, with revenue mix approximately 60 percent consumer and 40 percent enterprise. Both companies are growing at unprecedented rates. The structural difference is which market the growth comes from. Anthropic's enterprise-first revenue model captures customer-direct payments that do not require advertising-subsidized infrastructure. OpenAI's consumer-heavy revenue mix is exposed to the advertising-subsidized market that Google has locked up. The capital and the customer revenue are reacting to that structural difference in real time.

The deployment-and-application AI market is where actual AI tools are put to work in commercial and government settings. Palantir established this market and established what extraordinary execution within it can achieve. Palantir's 2020 IPO disclosed a $119 billion total addressable market, accepted by the SEC at the time. Palantir grew from a $20 billion IPO valuation to approximately $200 billion of market capitalization on $3 billion of revenue over six years. The trajectory reflects an AI-and-data-platform company executing against specific intelligence-community relationships, specific defense contracting work, and specific commercial deployments that compounded over fifteen years. Palantir is the load-bearing comparable for what a company with the right structural position can achieve in the deployment-and-application AI market when the execution is extraordinary. Microsoft's detethering of OpenAI is, in part, a signal that this is the market where the foundational AI work is producing economics that justify capital commitment.

Three markets, three structural configurations, three sets of winners that are now legible. Companies whose AI position maps onto a market with the structural prerequisites in place capture meaningful economic scale. Companies whose AI position lacks the structural prerequisites for any of the three markets are operating against contemporary signals that the public valuations have begun to price against.

Where SpaceX Actually Fits

SpaceX's S-1, per Reuters reporting on April 23, 2026, claims a $28.5 trillion total addressable market in support of its $1.75 trillion expected IPO valuation. More than 90 percent of the TAM is attributed to the AI sector — approximately $25.6 trillion. The figure is approximately 240 times Palantir's disclosed TAM at IPO. It exceeds even Morningstar's aggressive 2033 projection for Palantir by 20 times. It exceeds McKinsey and Bain's 2030 AI market projections by an order of magnitude.

SpaceX's actual revenue runs approximately $15.5 billion in 2025, with Starlink contributing $11.4 to $11.8 billion and launch services and government contracts contributing the balance. Quilty Space forecasts $20 billion in total SpaceX revenue in 2026, with $14 billion EBITDA and $8.1 billion of free cash flow, driven primarily by Starlink subscriber growth from 10 million to a projected 16.8 million. SpaceX's actual AI position runs through xAI, Cursor, Mistral, and the broader launch, computing-deployment, and federal contracting infrastructure. Mapped against the three markets:

• Advertising-subsidized AI: SpaceX has no plausible position. Google won. No new entrant — including SpaceX, OpenAI, Anthropic, or Meta — competes here. The structural prerequisites required were assembled over twenty years and are not replicable in a competitive timeframe.
• Enterprise-and-developer AI: SpaceX's position runs substantially behind both Anthropic and OpenAI. xAI's standalone annualized revenue is approximately $500 million as of end of 2025, two orders of magnitude below Anthropic's $30 billion and OpenAI's $25 billion. Even xAI's aggressive 2026 target of $2 billion in revenue, per Reuters reporting on Morgan Stanley investor materials, would leave the gap at one to two orders of magnitude. Cursor at the reported $60 billion transaction valuation establishes a developer-tools position adjacent to this market but operating at a different scale than the foundation-model leadership Anthropic has assembled. Mistral is a smaller European foundation-model provider operating below the scale of either Anthropic or OpenAI.
• Deployment-and-application AI: SpaceX has substantial structural prerequisites. The launch dominance, the Starlink global low-earth-orbit broadband network, the Space Force Phase 3 contracts, the Artemis Human Landing System work, and the Starshield classified satellite services position SpaceX in a configuration comparable to Palantir's structural position in 2020 — perhaps stronger, given the launch-and-infrastructure adjacencies that Palantir did not have. A company executing at Palantir-quality from this structural position could plausibly produce market capitalization in Palantir's range or somewhat larger. SpaceX's plausible AI economics, if executed at comparable quality, could produce $200 billion to $500 billion of value in this market over a decade.

That is a substantial AI position. It is also approximately one to two percent of the $25.6 trillion AI TAM allocation the S-1 attests to.

The Disclosure Standard the TAM Sidesteps

The decomposition above demonstrates that the work is performable using contemporary disclosed information. The components of the AI TAM are identifiable: advertising-subsidized AI under Google's structural lock, enterprise-and-developer AI under Anthropic's emerging leadership with OpenAI participating at scale, and deployment-and-application AI where Palantir established the trajectory and where SpaceX's structural prerequisites apply. SpaceX's plausible capture in each component, evaluated against documented analogs and known market positions, sums to a figure substantially below the aggregate claim. The decomposition uses aggressive assumptions about SpaceX's ability to realize and execute against these positions. A different analyst might produce a different decomposition. What the decomposition shows is that documented analogs exist for each component market, and that the aggregate figure cannot be reconciled with the structural prerequisites SpaceX brings to each component without specific disclosure to bridge the gap.

The disclosure standard should require the decomposition. A total addressable market claim at the scale of $28.5 trillion, supporting an IPO valuation of $1.75 trillion, with 90 percent attributed to a sector where multiple distinct markets operate under different structural configurations, requires substantive disclosure that breaks the figure down in such a way that the parts can be compared — favorably, unfavorably, or objectively — against analogs that exist and known and validated market positions. The current standard accepts the aggregate figure without requiring such decomposition. The decomposition above demonstrates that the work is performable using contemporary disclosed information. The disclosure standard should require SpaceX to do that work itself.

A consistent observation runs across the four pieces this series has now produced. Platform-scale incumbents extend their existing positions through new substrates by deploying the structural mechanisms the existing positions provide. Apple and Google extended bilateral search-distribution economics into the AI substrate through the Gemini arrangement. Anthropic established foundation-model leadership through customer-direct revenue at scale and parallel cloud-provider partnerships. Microsoft detethered from OpenAI exclusivity to stay open to winners in both consumer and commercial AI applications, signaling that the foundational work is happening at Anthropic and Palantir-shaped configurations and that OpenAI's advertising experiment is producing structurally minor results. SpaceX is extending federal contracting primacy and launch dominance through TAM aggregation that puts AI inside the IPO without specifying which market the AI capture occurs in.

The structural prerequisites SpaceX holds are real and substantial. The launch dominance is unmatched. The federal contracting position is primary. The Starlink global broadband infrastructure is the only operating asset of its kind. These prerequisites support a substantial deployment-and-application AI position with potential market capitalization in the Palantir range or somewhat larger. They do not support a $25.6 trillion AI TAM allocation.

What the Coming Months Will Resolve

The SpaceX IPO is expected to price in mid-June 2026. The Q2 earnings cycle for Apple, Microsoft, Google, and Amazon will report through April and into May 2026. The OpenAI IPO is expected within months under post-restructuring economics. The Apple-Google Gemini integration will deliver to consumer iPhones through iOS 26.4. The Anthropic infrastructure deployment will produce performance information against the $30 billion ARR baseline. Each of these events will produce information that confirms or contradicts the reading the four pieces have established.

That reading is now specific. The advertising-subsidized AI market is closed under Google. The enterprise-and-developer AI market is being captured by Anthropic at scale, with OpenAI participating substantially through enterprise share. The deployment-and-application AI market produces Palantir-trajectory outcomes for companies with the structural position to capture share. AI total addressable market claims that aggregate these markets into narrative-scale figures, attributing capture to entities whose structural prerequisites support a fraction of the implied capture, are operating against contemporary signals that the next several months will resolve.

The disclosure framework that permitted Apple's two characterizations of the Google relationship to coexist is the same framework that will evaluate the SpaceX S-1's $28.5 trillion TAM. The April 2026 announcements have made the structural reality the SpaceX TAM is operating against substantially more legible than it was on April 20, 2026. The S-1 attests to the figure. The disclosure standard does not require it to specify which markets the figure represents or which structural prerequisites SpaceX brings to each.