Has Google reached a computing power partnership with SpaceX? Analyzing the new paradigm of Starlink edge AI and space-based computing power

On June 5, 2026, a regulatory document disclosed one of the largest computing power collaborations in the history of the global technology and aerospace sectors: Google and SpaceX officially signed a roughly three-year cloud computing service agreement. According to the agreement, starting from October 2026 until June 2029, Google will pay SpaceX $920 million per month to access approximately 110k Nvidia GPUs along with supporting CPUs, memory, and other computing infrastructure. The total payment over the contract lifecycle is approximately $30 billion.

The timing of this agreement is no coincidence. On June 12, SpaceX is set to go public on NASDAQ at a fixed price of $135 per share, with ticker SPCX, aiming to raise $75 billion, corresponding to a valuation of about $1.77 trillion—making it the largest IPO in U.S. stock market history. Locking in Google as a major client just before this milestone has strategic significance for SpaceX’s valuation narrative and Google’s AI infrastructure deployment.

The edge nodes of the Starlink low Earth orbit satellite constellation can help fill gaps in traditional cloud providers’ infrastructure; securing a major client before the IPO enhances SpaceX’s valuation credibility; and Google’s strategic layout of satellite-based AI computing beyond data centers. Based on this, the transaction impacts the stock prices of GOOGL and SPCX in both directions, and also involves analyzing the trading functions of Gate stocks in capturing asset volatility.

Transaction Overview: Contract Terms and Key Data

Core Terms

According to regulatory filings submitted by SpaceX to the U.S. SEC, the agreement was officially signed on June 5, 2026. The key points are as follows:

| Term Item | Details | | --- | --- | | Effective Date | Full monthly payments start from October 2026; prior to that, capacity ramp-up period with lower monthly fees | | Contract Duration | Until June 2029 | | Monthly Fee | $110k/month (from October) | | Total Lifecycle Value | About $30 billion (full payment over 32 months) | | Computing Resources | About 110k Nvidia GPUs plus CPUs, memory, and related components | | Termination Clause | After December 31, 2026, either party may terminate with 90 days’ prior notice | | Delivery Guarantee | If SpaceX fails to deliver the promised GPU quantity before September 30, Google can terminate the agreement or reduce payments proportionally after a one-month grace period | | Intellectual Property | Google retains all ownership of its content, AI models, and related data |

Second Major Client Contract

It’s noteworthy that this is the second large-scale computing power leasing deal SpaceX has signed in the short term. In May 2026, AI company Anthropic reached an agreement with SpaceX to lease all computing output of SpaceX’s Colossus data center cluster at a monthly rate of $1.25 billion. This data center was previously built and operated by xAI. Together, these two contracts generate approximately $2.17 billion in monthly computing leasing revenue for SpaceX, with an annual recurring revenue (ARR) of about $26 billion. If both contracts are executed until maturity, their combined total value will exceed $70 billion.

First Dimension: Edge Node Coverage of Starlink Low Earth Orbit Satellite Constellation

Starting point: Why doesn’t Google build more data centers itself?

Google is not a company lacking in computing power. Industry estimates suggest that, thanks to its self-developed TPU chips and continuous iteration, Google is one of the world’s largest single AI computing power holders. In 2026, Google committed to investing between $175 billion and $185 billion in capital expenditures for global data center expansion. Recently, Alphabet announced an $80 billion equity financing plan to support this spending.

However, the rapid growth in AI computing demand continues to outpace industry expectations. Google Cloud’s official statement on this deal is: “This is a short-term and timely agreement designed to ensure we have transitional capacity to meet the surge in demand for our Gemini Enterprise AI platform, which exceeds our expectations.” Google Cloud’s latest financial report shows that its backlog (signed but not yet recognized revenue contracts) nearly doubled from the previous quarter, exceeding $460 billion.

The core issue is: ground data centers are constrained by energy supply, land resources, cooling capacity, and electrical infrastructure bottlenecks. Against the backdrop of continuously expanding AI model parameters, the expansion speed and marginal costs of traditional data centers are no longer able to match the demand growth curve.

Unique Asset of Starlink: Coverage in Non-Cloud Infrastructure Regions

Starlink has deployed over 9,600 low Earth orbit satellites, covering more than 160 countries and regions, with 10.3 million users. In 2025, this segment achieved revenue of $11.4 billion and an operating profit of $4.4 billion. But Starlink’s strategic value extends far beyond broadband access.

Traditional cloud providers (AWS, Azure, Google Cloud) concentrate their data centers in densely populated, power-rich, well-connected regions. Globally, many areas—including remote mining sites, ocean routes, polar research stations, and desert oil fields—are not economically or technically feasible for access to these data centers. In these scenarios, edge AI inference computing at the satellite or ground terminal level is increasingly in demand.

Starlink’s satellite network inherently offers global coverage. Combined with Google’s AI inference capabilities, both parties can deploy edge AI computing tasks on Starlink satellites or ground terminals—such as real-time analysis of remote sensing images, AI-assisted navigation on ships, or maintaining AI emergency response capabilities in disaster zones with disrupted communications. This edge AI architecture does not require raw data to be transmitted back to ground data centers thousands of kilometers away, enabling effective AI model operation in low-bandwidth, high-latency environments.

Moving from Edge to Orbit: The Long-term Vision of Space-based Computing Power

The surface layer of this deal is computing power leasing, but the deeper layer points to Google and SpaceX’s shared long-term strategic direction: orbital data centers. In May 2026, preliminary discussions about jointly building orbital data centers were disclosed, aiming to leverage Starlink’s satellite network and near-Earth platforms to overcome ground data center limitations related to energy, geography, and bandwidth.

The core logic of orbital data centers is: deploying AI computing power in low Earth orbit, powered almost continuously by solar energy, and using inter-satellite laser links (ISL) for high-speed data transfer between satellites. Currently, Starlink’s laser terminals support single-channel transmission speeds of 25Gbps. This architecture aligns closely with Google’s ongoing “Suncatcher” space-based data center project—aiming to build a constellation of solar-powered satellite data centers, with launch targets in 2027.

Market research indicates that in 2025, the global space-based data center market was about $1.28 billion, expected to grow to $3.81 billion by 2034, with a compound annual growth rate (CAGR) of 12.96%. The broader market for space computing (edge computing) is even larger, reaching $168.91 billion in 2025 and projected to grow to $345 billion by 2034. This market potential provides a long-term narrative foundation for Google and SpaceX’s collaboration.

Second Dimension: Locking Major Clients Before SpaceX IPO and Valuation Credibility

SpaceX’s Valuation Logic: A Negotiated Pricing

SpaceX plans to issue 555.6 million shares at a fixed price of $135 per share, raising $75 billion, with an estimated valuation of about $1.77 trillion. This valuation level involves significant debate:

• Supporters’ view: In its S-1 filing, SpaceX states that its total addressable market (TAM) is approximately $28.5 trillion, covering broadband, mobile communications, AI infrastructure, and enterprise applications.
• Skeptics’ view: Morningstar analysts used discounted cash flow (DCF) models to value SpaceX at about $780 billion—roughly 44% of the IPO price. Morningstar explicitly states, “We believe the company’s valuation is significantly overstated, and investors will have opportunities to buy at more attractive prices after the IPO.”

Financial data reflect this valuation debate. SpaceX posted a net loss of $4.28 billion in the latest quarter, with a full-year 2025 net loss of $4.94 billion, accumulating deficits of $41.3 billion. The AI division xAI’s quarterly operating loss is about $2.5 billion. Based on 2025 revenue of $18.7 billion, a valuation of $1.77 trillion implies a price-to-sales ratio of about 92—extremely high even for growth companies, and especially risky for a conglomerate still incurring massive losses.

Supporting Valuation Narrative with Computing Power Leasing Contracts

Against this backdrop, the two annualized $26 billion computing leasing contracts with Google and Anthropic provide two key supports for SpaceX’s valuation narrative:

First, revenue visibility. SpaceX’s IPO filing shows three main segments: Starlink ($11.4 billion in 2025), space launch (~$4 billion+), and AI/xAI (~$3.2 billion). The combined annualized revenue from the two leasing contracts ($26 billion) already exceeds the total revenue of these three segments. This means that even if rocket and satellite businesses stagnate, the leasing of computing power alone can underpin the underlying revenue basis for the valuation.

Second, strategic signal of business model transformation. Traditionally, markets view SpaceX as a “rocket company + satellite broadband operator,” which struggles to justify a valuation multiple higher than Nvidia’s. The leasing contracts indicate that SpaceX is repositioning itself as an “AI infrastructure-as-a-service” provider—transforming its asset-heavy data center infrastructure (originally built for training large models like Grok) into a commercial asset for third-party leasing. This shift turns data centers from cost centers into profit centers.

This transformation has profound implications: it means SpaceX’s AI infrastructure no longer solely supports its own large model R&D but also generates scale economies by providing compute to giants like Google. A Google Cloud spokesperson described this cooperation as “bridging capacity demand,” implying a time lag between their own data center expansion cycle and AI demand explosion, with SpaceX providing mature compute resources within this window.

Issuer Risk Warning: Contract Termination Options

It’s important to note that these contracts are not irrevocable long-term lock-ins. The agreement explicitly states: after December 31, 2026, either party can terminate with 90 days’ notice; if SpaceX fails to deliver the promised GPU quantity before September 30, Google can terminate immediately or reduce payments proportionally after a one-month grace period. This clause introduces some uncertainty about the enforceability of the agreement beyond 2027. For potential investors, fully incorporating the total lifecycle value (over $70 billion) into valuation models requires careful consideration of these exit options.

Third Dimension: Google’s AI Computing Strategy Beyond Data Centers

Asymmetric Competition in Computing Power Supply

In the current global cloud market, AWS holds about 30%, Azure about 25%, and Google Cloud about 13%. While Google Cloud is catching up in market share, it has built a differentiated competitive advantage in AI computing power through its self-developed TPU chips and the Gemini ecosystem. At Google Cloud Next 2026, Google announced the eighth-generation TPU—optimized for training (TPU 8t) and inference (TPU 8i)—and for the first time, offered direct hardware sales to select customers (previously only via cloud leasing).

However, AI computing supply is entering an “asymmetric competition” phase. Traditional cloud providers’ advantages are based on data center scale, but physical constraints—land, power, cooling, and carbon emissions—limit expansion. Google’s “Suncatcher” space-based data center project, a fully solar-powered satellite constellation, aims to break these physical constraints. If successful, in-orbit data centers will no longer be limited by terrestrial power grids, land availability, or cooling water, enabling a fundamental differentiation in the supply chain.

Starlink as the “Precursor Infrastructure” for Space-based Computing

Before “Suncatcher” is fully operational, Starlink already provides a ready-made space-based infrastructure.

Structurally, Starlink’s low Earth orbit satellite network is more than just a communication system. Its satellites are equipped with inter-satellite laser links, onboard computing capabilities, and high-precision attitude control—forming the technical foundation for future in-orbit data centers. Elon Musk has publicly stated plans to expand Starlink V3 constellations and develop in-orbit data centers to meet the surging global compute demand driven by AI.

Another key validation comes from Starcloud (formerly Lumen Orbit), which has ordered over 50 Starlink mini laser terminals from SpaceX to build its orbital data center network. The first batch hardware is expected to be launched into orbit within a year. This demonstrates that Starlink’s inter-satellite communication infrastructure already has the technical feasibility to support third-party orbital data center deployment.

For Google, piloting edge AI computing via Starlink can provide operational experience, technical data, and business model validation ahead of large-scale “Suncatcher” deployment. This “test first, expand gradually” approach is more cautious technically and more feasible commercially.

Bidirectional Impact on GOOGL and SPCX Stock Prices

Impact on GOOGL (Alphabet)

Short-term: modest positive financial impact. The $920 million monthly computing expenditure accounts for less than 0.3% of Alphabet’s 2025 revenue of over $350 billion, so the immediate financial impact is limited. But strategically, this alleviates capacity bottlenecks for Gemini Enterprise. The rapid growth in Google Cloud backlog (signed but not yet recognized revenue) indicates accelerating AI service demand, and sufficient compute supply helps convert backlog into recognized revenue—supporting Google Cloud’s revenue growth.

Medium-term: changes in strategic cost structure. Google’s choice to outsource part of its AI infrastructure reflects a “hybrid model” of self-owned plus external procurement. If Google can acquire compute more quickly and at lower marginal costs via external suppliers, capital efficiency improves; over-reliance on SpaceX could impact long-term supply chain controllability and costs.

Risk warning: The 90-day termination clause means Google retains flexibility to exit. If Google’s own data center capacity expands faster than expected, or its TPU inference costs outperform external procurement, it might reduce or terminate the contract after 2027. Currently, Google describes this as a “bridge capacity,” indicating a transitional supplement rather than a long-term dependency.

Impact on SPCX (SpaceX)

IPO valuation support. The $26 billion annualized revenue from the two leasing contracts is a key narrative supporting SpaceX’s IPO valuation at about 92 times sales. Without this revenue, the estimated annual revenue would be around $18.7 billion, pushing the valuation multiple to even more extreme levels. Therefore, signing these contracts substantively validates the IPO’s valuation rationale.

Sustainability challenges. But note that Morningstar’s DCF valuation of SpaceX’s core business (rockets + Starlink) is about $611 billion, with AI business valued at roughly $170 billion (probabilistic scenarios), totaling about $781 billion. This is roughly $1 trillion below the $1.77 trillion IPO valuation, requiring post-IPO performance to bridge the gap. The actual realization depends heavily on the execution of AI contracts and SpaceX’s ability to expand third-party compute clients.

Comparable companies: Nvidia’s current market cap is about $5.3 trillion, with annual revenue around $250 billion (multiple ~21x). AI startup Anthropic’s recent valuation was about $965 billion. SpaceX’s ~92x sales ratio far exceeds these, indicating the IPO price includes long-term expectations of “sustained growth of Starlink + scaled AI profitability + orbital data centers.” The window for these expectations to materialize is projected around 2027–2029.

Gate Platform Stock Trading Features and Operational Path

For investors seeking exposure to GOOGL and SPCX, Gate offers a one-stop stock trading solution.

Core advantages of Gate stock trading

• Unified multi-asset trading: Seamlessly switch between cryptocurrencies and US stocks within a single account, avoiding cross-platform transfers.
• 24/7 trading window: Unlike traditional brokers limited to US stock trading hours (21:30–04:00 Beijing time), Gate provides round-the-clock order management.
• Fragmented unit trading: Supports minimum units of $1, enabling small, diversified investments even in high-priced stocks like GOOGL (~$200/share) and SpaceX IPO (~$135/share).
• Real-time quotes and depth info: Access live prices, pre/post-market data, company announcements, and analyst reports—helping users stay updated on relevant market events.
• Multi-market coverage: Supports trading of listed stocks in US, Hong Kong, and other major markets, catering to global asset allocation.

Operational guide for GOOGL and SPCX

• Account setup: Register and complete KYC verification on Gate; deposit USD or USDT as margin.
• Trading GOOGL: Search “GOOGL” on the stock page; choose market or limit orders based on quarterly performance and Gemini Enterprise subscription signals.
• Trading SPCX: SpaceX plans to list on NASDAQ under “SPCX” on June 12; users can participate in IPO trading via Gate’s stock section.
• Risk management: Stock trading involves market risks; users should develop position strategies aligned with their risk appetite to avoid over-concentration.

Conclusion

The $30 billion computing power contract between Google and SpaceX is essentially a strategic coupling of two giants—each with deep complementary resources—against the backdrop of “post-Moore’s Law” era computing bottlenecks.

Starlink’s low Earth orbit network provides the physical platform for edge AI computing in non-cloud regions, addressing the structural gaps in traditional data center coverage; Google’s AI ecosystem and Gemini demand create a high-value commercial outlet for SpaceX’s data center assets. Looking further ahead, this deal offers both technical validation and business logic for the long-term vision of orbital data centers.

For investors, GOOGL represents a stable AI application layer, while SPCX offers a highly flexible option in space-based infrastructure. When constructing a portfolio, it’s important to consider macro variables such as AI demand growth, data center expansion efficiency, and regulatory policies. Gate’s stock trading features facilitate the implementation of such combined strategies with convenient tools and fractional operation support.