How Google limited Meta: What exactly happened

In March 2026, Google delivered unpleasant news to Meta: its cloud division, Google Cloud, would not be able to deliver the full capacity Meta had requested for access to Gemini AI models. The limitation affected other customers, but Meta felt the impact particularly strongly due to its exceptionally high demands for computing power.

The consequences were immediate and practical. Meta's engineering teams had to start strictly conserving AI tokens — units that measure the volume of text processed in language models. Some internal AI projects were delayed or had to be put on hold.

Sundar Pichai publicly confirmed this situation during an earnings call on April 29, 2026. Google Cloud then reported revenues of over $20 billion for the first quarter of 2026 — a year-on-year increase of 63%. Yet, Pichai admitted that Google was "compute constrained" and that revenues would have been even higher if it had been able to meet all demand. Meanwhile, the backlog of signed contracts had almost doubled to over $460 billion.

The irony of the situation is further highlighted by another fact: Meta itself is a massive investor in AI infrastructure, yet it has to purchase capacity from a direct competitor. In April 2026, it announced the creation of Meta Superintelligence Labs — an elite team for developing advanced AI. Meanwhile, according to reports from summer 2025, Meta's leadership was actively considering whether to deploy Gemini or OpenAI models for various functions of its products.

Why spending billions isn't enough

The Google–Meta case illustrates a structural problem that a simple increase in investment won't solve overnight. Training and operating advanced AI models requires:

• Specialized chips — especially GPUs from Nvidia, whose manufacturing capacities have physical limits
• Energy-intensive data centers — a single facility can consume hundreds of megawatts
• Years of construction — even with billions in investment, building new capacity takes months to years

Sundar Pichai directly named this paradox: demand exists, money exists, but the physical infrastructure simply cannot grow as fast as needed. And we are talking about Google — a company with one of the largest cloud infrastructures in the world.

Europe: 5% of global power, 100% ambition

If the situation is complicated for Google and Meta, it is even more critical for Europe. According to available data, the European Union owns approximately 5% of the world's computing power for AI, while the United States holds roughly 75%.

McKinsey estimates that demand for data center capacity in Europe will grow from 10 GW of IT power in 2024 to 35 GW by 2030 — a threefold increase, almost exclusively due to AI. However, infrastructure is not keeping pace with this growth.

The reasons are systemic:

• Electricity price — European rates are significantly higher than in the USA, making data center operations more expensive
• Grid connection — in some European countries, obtaining a grid connection for new large facilities takes up to ten years
• Outdated transmission network — the European electricity system was not designed with AI requirements in mind, where a single facility can suddenly demand hundreds of megawatts
• Dominance of American hyperscalers — the trio of Google Cloud, AWS, and Microsoft Azure holds approximately 70% of the European cloud services market, while European providers hold only 15%

This means that European companies are dependent on American infrastructure when building AI applications and products. And as the Meta case shows, even money does not guarantee access to capacity when there is a global shortage.

What Europe is doing and what is not enough

The European Commission is aware of the threat. In 2025, it launched the InvestAI initiative with the aim of mobilizing 200 billion euros for AI infrastructure and research. This sounds impressive, but in the context of the global race, it looks more modest: OpenAI itself announced investments of 500 billion dollars (about 422 billion euros) in AI infrastructure over four years.

On the legislative front, the Commission introduced the Cloud and AI Development Act (CADA), which aims to reduce Europe's strategic dependence on American providers and strengthen the sovereignty of European AI. How quickly this translates into real capacity remains to be seen.

Positive signals are coming from the private sector. French startup Mistral AI secured approximately 830 million euros in debt financing to purchase about 13,800 Nvidia chips and build its own data center near Paris. The capacity is expected to be available in the second quarter of 2026. This is a step in the right direction, but it is still a drop in the ocean of global AI infrastructure.

What this means for Czech companies and developers

Access to more powerful AI models — Gemini, GPT-4o, Claude Opus — remains available to Czech companies via standard APIs for now. Limitations like the one that affected Meta primarily target customers with extremely high consumption. The average Czech startup or developer will not directly feel the restrictions.

In the long term, however, an unfavorable scenario looms: if demand for AI computing power globally exceeds supply, hyperscalers will have to choose between customers. Larger players with long-term contracts and greater financial strength will have priority. Smaller companies — including Czech ones — may encounter longer waiting times, higher prices, or restricted access to the latest models.

This is precisely why European sovereignty in AI infrastructure is a matter of industrial policy, not just a technological debate. Dependence on two or three American platforms is a strategic risk for the entire continent — as demonstrated by Google's seemingly internal dispute with Meta.

Conclusion: Computing power is the new oil

The Google–Meta case is more than a technical curiosity. It is a warning. AI computing power has become a strategic resource — much like oil in the 20th century. And just like then, whoever owns or controls the infrastructure will have enormous influence on the economy and geopolitics.

Europe has a chance to change the situation, but time plays a crucial role. Investments in data centers, modernization of the electricity grid, and support for domestic AI infrastructure players are not luxuries — they are existential necessities for Europe's competitiveness in the era of artificial intelligence.