Google pays SpaceX $920 million per month for AI compute

The timeline of an unprecedented agreement

On June 5, 2026, TechCrunch reported The official news: Google will pay SpaceX $920 million per month for compute capacity. A spokesperson for Google confirmed the deal, describing the partnership as a direct response to unexpected demand generated by the company's recently launched AI products.

In absolute terms, we are talking about over $11 billion annually. Therefore, it is a financial commitment that exceeds the GDP of several medium-sized European countries. However, the figure should not distract from the more relevant aspect: the choice to use SpaceX — and not a traditional hyperscaler — as a provider of computational infrastructure.

Therefore, the question the market is asking is not just «how much does it cost,» but «why SpaceX specifically.» The answer lies in the Starlink architecture and the ability to distribute compute at the edge on a global scale, reducing dependence on terrestrial data centers.

Why SpaceX and not a traditional data center

SpaceX has a constellation of satellites in low Earth orbit (LEO) through the Starlink program. This infrastructure enables geographic coverage that no terrestrial data center can replicate with the same capillarity. Furthermore, latency in LEO orbit is significantly lower compared to traditional geostationary orbit satellite solutions.

According to recent analyses by Gartner on the evolution of cloud computing, the edge distribution of compute represents one of the most relevant trends for the 2026-2028 period. In particular, the ability to process data close to the source—rather than centralizing it in a few hubs—becomes critical for real-time AI models.

Consequently, Google obtains something with this agreement that it cannot build in a short time: a globally distributed, already operational, and scalable network of computational nodes. Similarly, SpaceX diversifies its revenues well beyond space transport and consumer connectivity.

Winners, losers, and those who silently observe

The most obvious winner is SpaceX. The agreement transforms Starlink from a connectivity service into compute-as-a-service infrastructure. Among other things, it definitively legitimizes the space business model as a component of the enterprise cloud ecosystem.

Google, for its part, finds itself in an ambivalent position. On the one hand, it resolves an urgent infrastructure bottleneck. On the other, it publicly acknowledges that it did not anticipate the demand generated by its own AI products—an admission that financial markets and competitors will not fail to analyze.

The quietest losers are the traditional cloud providers: AWS, Microsoft Azure, and those segments of Google Cloud that rely on conventional terrestrial infrastructure. Indeed, if the SpaceX-Google model proves superior in efficiency, it could redefine the industry's benchmarks. Similarly, traditional data center builders could see demand for new physical facilities slow down in the coming years.

Those who are observing carefully, finally, are the European governments. The dependence on compute infrastructure controlled by non-European private entities raises questions of digital sovereignty that the EU regulator cannot ignore for long.

SHM Studio Reading: Compute as a Strategic Variable

We of SHM Studio We interpret this agreement as a signal of discontinuity. This is not technical news reserved for system engineers. On the contrary, it concerns anyone who uses cloud services for marketing, data analysis, or automation activities.

In particular, Italian SMEs that have adopted AI-as-a-service tools — from language models to platforms digital marketing based on machine learning — indirectly depend on the compute infrastructure that Google uses to deliver those services. Therefore, the quality, latency, and availability of those services are directly affected by agreements like this.

For this reason, those planning a medium-term digital strategy today—whether oriented towards SEO, all Google Ads campaigns and to the development of websites with AI components — you must consider the stability of the cloud infrastructure as a variable, not a given.

Concrete implications for service latency and availability

The practical question for an SME is simple: will this agreement improve or worsen the performance of the services I use? The honest answer is: it depends on the time horizon.

In the short term—2026—the agreement addresses a capacity issue, not perceived quality. Therefore, the end-user will not notice immediate differences. However, in the medium term—2027-2028—the integration of satellite compute nodes could reduce latency for users in geographic areas with limited data center coverage.

According to Harvard Business Review, The reduction of latency in AI services has a direct impact on the productivity of enterprise users. In fact, even variations of a few tens of milliseconds affect the user experience of tools like real-time language models.

In addition to this, redundancy must be considered. A compute network distributed across LEO orbit is intrinsically more resilient than a cluster of geographically concentrated data centers. Consequently, cloud service availability could improve in failure or overload scenarios.

What nobody is saying: the price of compute could go up

There is an aspect that official communication tends to downplay. A $920 million monthly agreement has a cost that, sooner or later, will be transferred down the value chain. Therefore, it is not unreasonable to assume that the prices of AI-powered cloud services will face upward pressure in the coming quarters.

Google has already increased the prices of some Gemini APIs in 2025. Therefore, the agreement with SpaceX only adds structural pressure on margins. SMEs that currently use these services at low costs should start evaluating alternative scenarios or negotiating multi-year contracts while prices are still stable.

In this context, technological diversification becomes a strategic choice. Relying on a single cloud provider for all activities of artificial intelligence, Automated copywriting e LinkedIn campaign exposes the company to repricing risks that could impact the ROI of digital activities.

Next moves: what to evaluate in the next 12 months

In light of the analysis, some operational priorities emerge for Italian SMEs with a structured digital presence.

• Cloud cost audit Map the AI-as-a-Service offerings currently in use and estimate their price sensitivity. Specifically, identify which activities depend on Google APIs and which could migrate to alternative solutions.
• Multi-year contracts where possible, lock in current prices through long-term agreements with providers. Additionally, review the price adjustment clauses in existing contracts.
• Latency assessment if the company operates in geographic markets peripheral to European data centers, monitor the evolution of Starlink's performance as a potential future opportunity.
• Vendor mix diversification do not concentrate the entire digital infrastructure on a single cloud ecosystem. Then, explore hybrid solutions that combine Google Cloud, Azure, and European providers such as OVHcloud or Hetzner.

To further explore these assessments within the specific context of their company, the team SHM Studio is available for consultation. Finally, those who want to stay updated on the evolution of these scenarios can follow the in-depth analyses published in our blog.