Saudi Arabia’s ambitions in AI have come as a shock to the West.
In less than a year, the kingdom has moved from broad statements about diversification to signing multibillion-dollar data centre deals, securing access to advanced chips and outlining a plan to sell AI services cheaper than almost anywhere else on earth.
Saudi Arabia now plans to create a unique model that could easily surpass what the rest of the world has been doing so far.
That is to turn cheap electricity into cheap AI, and export the results.
Power first, everything else second
At the centre of Saudi Arabia’s AI strategy is electricity. Not talent, not software, not even chips. Power.
Large solar projects on the Red Sea coast are producing electricity at close to one $0.01 per kilowatt-hour.
That is a fraction of the cost in Europe or parts of the United States.
For artificial intelligence, this matters most at the inference stage, because training large models is expensive and sporadic.
Running them every day, answering billions of prompts, is where long-term costs sit.
Inference costs are driven by two things. Hardware efficiency and electricity prices. And since Saudi Arabia cannot manufacture advanced chips, it can run them on some of the cheapest reliable power available.
This strategy has turned into national policy. According to a PwC report, AI could help boost Saudi Arabia’s GDP by more than 12% by 2030.
So the government has prioritised AI infrastructure under Vision 2030 and placed execution under a single vehicle, HUMAIN.
The goal is not to host a few data centres. It is to reset the cost base of AI services by doing the computing where power is abundant and cheap, then exporting the output digitally.
Source: PwC
HUMAIN and the full-stack gamble
HUMAIN is owned by the Public Investment Fund and led by Tareq Amin, a telecoms and infrastructure executive rather than a traditional software founder.
HUMAIN is structured as an operating company, not a holding firm. It is building data centres, cloud platforms, large language models and applications under one roof.
The company’s plans are large by any standard. Public statements point to a target of up to six gigawatts of data centre capacity by the mid 2030s.
Early site surveys reportedly identified more than 200 potential locations with access to power, land and permits. In most countries, power availability is the bottleneck. In Saudi Arabia, it is the selling point.
HUMAIN is also pushing beyond infrastructure. It has launched an AI-driven operating system aimed at enterprises, where tasks are executed through natural language rather than menus and icons.
Internally, the company says it already uses AI agents to run payroll, legal and administrative processes.
This indicates that Saudi Arabia does not want to be seen only as a place to park servers.
It wants to show what happens when AI is deployed at scale inside organisations.
The partners tell the story
Saudi Arabia’s AI strategy becomes clearer when looking at its partners. HUMAIN has signed up Amazon Web Services as its preferred global cloud partner and is working with Nvidia to supply the accelerators that power modern AI.
Together, they are planning a dedicated AI Zone in Riyadh, designed to handle both training and inference workloads and to give customers direct access to foundation models through AWS services.
On the physical side, HUMAIN has partnered with AirTrunk, backed by Blackstone and the Canada Pension Plan Investment Board.
The initial commitment is about three billion dollars for a data centre campus.
This is not venture capital. It is long-term infrastructure money looking for stable returns.
Chip access has been more politically sensitive. Early deals focused on inference-specific architectures from companies like Groq.
More recently, Saudi Arabia secured licences to import tens of thousands of top-tier Nvidia chips, reportedly costing around one billion dollars.
That volume would not fill multiple hyperscale campuses, but it marked a shift. The kingdom is no longer locked out of frontier compute.
Selling tokens, not electricity
Saudi Arabia’s commercial model is unusual. It does not plan to sell electricity abroad. Power is hard to move and expensive to transport. Data is cheap to move.
HUMAIN’s pitch to AI developers is straightforward. Run your models on Saudi infrastructure. Use Saudi electricity. Generate output tokens at a lower cost than anywhere else.
In practical terms, this could mean selling AI inference output at prices well below current market rates.
Industry sources quoted in recent reporting suggest output tokens could be priced at roughly half prevailing levels, while end users continue to pay standard fees.
The margin gap would be captured by whoever controls the compute.
If this holds, it would put pressure on cloud providers and AI application companies with high inference costs.
It would also create a natural split in the market. High-latency sensitive tasks stay close to users. Bulk inference moves to wherever power is cheapest.
What could change global AI economics?
Saudi Arabia’s approach carries risks. Data centres run hot. Cooling in a desert climate raises water and efficiency questions that remain unresolved.
Talent shortages are real and cannot be fixed with capital alone. Access to advanced chips depends on geopolitical decisions beyond Riyadh’s control.
But the strategy is internally consistent. It treats AI as an industrial commodity.
It focuses on the part of the value chain where national advantages matter most. And it is backed by capital that can absorb long payback periods.
For global markets, the most important implication is pricing. If Saudi Arabia succeeds in setting a lower global reference price for AI inference, margins will compress elsewhere.
Companies that rely on expensive power markets will face a choice. Pass costs on, accept lower profits or move workloads.
There is also a change in bargaining power. By aggregating multi-gigawatt demand and sovereign capital, Saudi Arabia becomes a buyer that chipmakers and cloud firms cannot ignore.
That leverage extends beyond pricing into localisation, data governance and long-term partnerships.
Saudi Arabia is not guaranteed to become the world’s AI hub. But it is attempting something few others can.
It is turning electrons into intelligence at scale and selling the result.
The post Can Saudi Arabia really undercut the world on AI with low-cost electricity? appeared first on Invezz
