Desert electricity for pennies and a plan for the cheapest data centers in the world. Europe may have a serious problem

Ashley Davis04/01/2026

0 94 4 minutes read

In the AI economy, the winner is increasingly not the one who has the most data, but the one who delivers computing power to the user at the cheapest rate. The cost of using models (inference, i.e. querying the system and generating answers) in practice comes down to two components. This is correspondingly expensive equipment and the bill for the energy needed for the operation of the chips and cooling. It's difficult to save on equipment without reducing performance, so electricity becomes the most important battlefield — and here Saudi Arabia wants to play in a different league.

The Shuaibah project (Shuaiiba 1) reports a record low energy cost of 3.9 halala per kWh, and significant new PV capacity was brought online in the country in 2024, building economies of scale on the energy supply side.

“The Economist” reports that this one cheap energy is presented as the cornerstone of the plan and is intended to power large-scale data centers for AI and lower the cost of generated tokens (accounting units in AI services) below the prices to which the market is accustomed.

The next part below the video:

Read also in BUSINESS INSIDER

This is important because commercial models are priced per input and output token. If someone can produce tokens cheaper than global suppliers assume in their price lists, they can become an inference wholesaler for the entire world. This is also what Riyadh's offer will look like – transfer the model launch to the Saudis, and the cost of producing a response will drop so much that it will be possible to compete aggressively on price.

Check also: This is how money circulates in the AI market. A short history of the great transfer

Humain and the state investment orchestra

This strategy is supported by state institutions. Humain – a new entity building capabilities for the development of AI – is to centralize activities, and its CEO Tareq Amin combines the technological narrative with the pressure of “Vision 2030”, i.e. moving away from oil dependence.

At the same time, Saudi Arabia plays with advantages that are currently a bottleneck for Europe. That is, the availability of land, a fast administrative path and the possibility of infrastructure planning with state support (including energy nodes near renewable energy sources).

The most difficult element is the chips. The geopolitical game of export control is about the best deals. And here in 2025 something important happened, because The American administration began to allow the sale of advanced semiconductors to entities from the Gulfbut under conditions of safety and appropriate reporting.

In practice, this means that Riyadh sees a window to buy the latest GPUs (including on a large scale), and not just build a hall and cheap energy.

In addition, there is capital and operational work. Humain announced a partnership with AirTrunk (backed by Blackstone) for approximately USD 3 billion to build a data center campus in Saudi Arabia.

In the following months, there were also announcements about cooperation with global technology and hardware players and about building an “AI operating system” offer for organizations, which it is intended to create internal demand and attach partners to the local ecosystem.

See also: Chance of min. 50 percent profit in 2026. These companies can deliver great results

Europe towards offshoring of computing power

For Europe, the Saudi plan is an uncomfortable mirror. The continent has talents, regulations and a market, but it is increasingly facing energy and network limitations. The Ember report showed that in 2023, data centers accounted for a huge part of energy consumption in key hubs – in Amsterdam, London and Frankfurt it was several dozen percent of demand, and in the Dublin area the scale was even greater.

At the same time, European energy prices for business remain a competitive challenge and their level and structure (including taxes and fees) vary greatly between countries.

In this situation, Europe has three realistic paths to respond – and each has a political cost.

The first is to accelerate your own AI infrastructure despite barriers. The EU is already trying. The AI Factories programs and the plan for large “AI gigafactories” will build computing capacity and data storage in Europe, with priority for start-ups, SMEs and the research ecosystem. However, without parallel unblocking of the network, permits and sources of stable energy (RES and storage, nuclear, system flexibility), this plan will be more expensive than in regions that have cheap electricity and space.

The second one is a harder doctrine of sovereignty. Europe may want to keep some AI applications – especially in regulated sectors and government – within EU jurisdiction, even if it means a higher token cost. This direction is reinforced by cloud sovereignty initiatives and the legal framework around data control and dependence on non-EU providers. In this variant, Saudi Arabia becomes a price competitor, but not necessarily a place where sensitive calculations are “released”.

Read also: An AI factory will be built in Munich. Nvidia and Deutsche Telekom will increase Germany's computing power by half

The third path is the most pragmatic and most controversial, i.e. selective inference offshoring. Europe could decide that for many applications (marketing, customer service, content generation, some analytics) what matters is cost and speed, and not necessarily the location of the server room — and then Saudi data centers become a factory for cheap AI answers, while sensitive data and critical models remain in the EU. This, however, would require very precise rules for data transfer, audit and supply chain control, because in AI geopolitical risk quickly translates into operational risk.