Meta and SpaceX Are Cashing In on Excess AI Computing Power

SpaceX has signed a $29.4 billion compute deal with Google, while Meta prepares to launch its own cloud business to monetize surplus AI capacity—marking a fundamental shift in how technology giants are extracting value from massive infrastructure investments. The two companies are pursuing radically different strategies to convert excess computing power into revenue streams, even as they race to deploy hundreds of billions of dollars in data center infrastructure. This divergence reveals competing visions for the AI infrastructure economy in 2026 and beyond.

SpaceX’s $920 Million Monthly Windfall From Google

SpaceX inked a 32-month agreement to rent AI compute capacity to Google at $920 million per month, announced in June 2026 just days before the company’s planned IPO. The deal grants Google access to approximately 110,000 Nvidia GPUs housed in SpaceX’s data centers, infrastructure originally built to support xAI’s Grok operations. This arrangement totals roughly $29.4 billion in committed revenue and represents one of the largest compute capacity agreements ever struck between technology companies.

The timing proved strategic for SpaceX’s public market debut. The company filed its S-1 registration statement targeting a $1.75 trillion valuation, with 2025 filings revealing $18.67 billion in total revenue but a $2.58 billion operational loss driven by $12.7 billion in AI capital expenditure. In the first quarter of 2026 alone, SpaceX deployed $7.72 billion toward AI infrastructure—representing 76% of the company’s total capital deployment. The Google deal provides concrete proof that SpaceX’s massive compute buildout can generate immediate, substantial returns.

Meta’s Internal Monetization Model and Cloud Ambitions

Meta CEO Mark Zuckerberg confirmed in May 2026 that launching a cloud computing business to monetize surplus AI capacity is “definitely on the table,” though he emphasized the company would only pursue this strategy once it achieved significant data center overcapacity. This represents a potential new revenue stream for Meta beyond its core advertising business, which has already begun benefiting substantially from AI improvements. Unlike SpaceX’s external model of selling compute to competitors, Meta’s strategy centers on leveraging AI Internally to drive advertising efficiency while exploring cloud services as a secondary opportunity.

Meta raised its full-year 2026 capital expenditure guidance to $125–$145 billion in April, up from a prior range of $115–$135 billion—a 58% year-over-year increase primarily dedicated to data centers and AI hardware. CFO Susan Li noted that 2026 growth would be “notably larger” due to expanding compute needs. Meta simultaneously struck a $10 billion cloud deal with Google to support six years of AI infrastructure, including servers, storage, and networking—one of the largest commitments in Google Cloud’s 17-year history.

Validation From Industry Leaders and Emerging Revenue Models

Nvidia CEO Jensen Huang endorsed Meta as “the best AI deployer in the world,” specifically validating Meta’s approach by pointing to the company’s incremental return on invested capital exceeding 20% on recent AI investments, with cash-based return on invested capital above 52%. This financial validation matters significantly: Meta’s Q4 2025 earnings revealed that AI-driven ad ranking produced roughly four times the revenue impact compared to simply increasing ad load, with recent updates increasing Facebook feed clicks by approximately 3% and Instagram conversions by 5%.

Meta also launched its first AI subscription trials in May 2026, testing services at $7.99 or $19.99 per month across Singapore, Guatemala, and Bolivia. This marks the company’s inaugural effort to charge users directly for AI features—a diversification from its advertising-dependent revenue model. Wolfe Research estimates that AI subscriptions could contribute up to $3 billion by 2027 and $16 billion by 2030, providing Meta with multiple monetization pathways for its expanding compute infrastructure.

The Contrasting Paths to AI Infrastructure Profitability

Meta’s and SpaceX’s strategies reflect fundamentally different approaches to the same challenge: how to generate returns from massive, capital-intensive AI infrastructure investments. Meta has demonstrated that internal monetization through advertising efficiency can justify enormous capex commitments, with Q4 2025 ad revenue reaching $58.14 billion—up 24% year-over-year—directly tied to AI system improvements. SpaceX, by contrast, has chosen to build infrastructure specifically designed for external consumption, treating compute capacity as a primary business line rather than a supporting function for other revenue streams.

Goldman Sachs projects that SpaceX’s AI-related revenue will surge 100-fold from $3.2 billion in 2025 to $322 billion by 2030, potentially surpassing AWS at $107 billion and Nvidia at $115 billion. The forecast predicts total SpaceX revenue of $474 billion by 2030, with AI accounting for more than half of all revenue. These projections underscore analyst confidence in the external compute rental model, though they remain subject to execution risk and competitive dynamics.

What Investors Should Monitor Next

The immediate focus falls on SpaceX’s IPO execution and the company’s ability to deliver on its Google compute commitments while scaling additional capacity for other enterprise customers. Meta’s cloud business launch timeline and initial adoption rates will determine whether the company can establish itself as a credible infrastructure competitor to established cloud providers. Additionally, the success of Meta’s AI subscription trials in emerging markets will indicate whether consumers will pay directly for AI features, a critical variable in projecting the company’s long-term revenue diversification.

Both companies are making multi-year bets that massive AI infrastructure can generate substantial returns—Meta through internal advertising efficiency and emerging subscription services, SpaceX through direct compute rental agreements. The next 12 to 18 months will reveal which approach proves more sustainable and profitable as competitive pressures intensify and demand for AI computing resources continues accelerating across the technology sector.