In a darkened room in Saudi Aramco’s headquarters, a green avatar appears on a screen. It introduces itself as “Ariva,” and says in a deep voice that it can solve whatever technical problem the company is facing. When an engineer asks how to increase drilling speeds at one of the thousands of oilfields in the kingdom, Ariva spits out an answer without pausing, listing the adjustments technicians need to make, and what the precise results will be.
Talking computers are old-school, of course. But Aramco’s presentation, during a rare two-day visit by Fortune to its base in eastern Saudi Arabia, aimed to reveal how the world’s biggest oil producer is piecing together a new long-term growth strategy. The company is in the midst of a massive rollout of artificial intelligence—crucial, it believes, in turning this 86-year-old fossil-fuel giant into an AI powerhouse that can outlast the golden age of oil.
In Aramco’s Dhahran headquarters and in Abqaiq—the world’s biggest oil-processing facility, an hour south into the desert—engineers and executives detail the rapid rollout of AI in both oil and gas production and research. That includes its use in experimentation—for example, how to best capture carbon from the atmosphere—as well as numerous uses that have already been deployed.
The company anticipates enormous benefits. With meticulous production records for each well, dating back more than eight decades, Aramco can use generative AI models to maximize its production with minimal human intervention, and to explore more remote, deeper reserves.
Aramco uses AI, for example, to help detect potential blockages and leaks ahead of time, across thousands of miles of pipelines. From the company’s Exploration and Petroleum Engineering Advanced Research Center, or ExpecARC, Aramco sends seismic robots (they look like the Mars Rover) into the desert, where they use AI to analyze geological formations deep underground—saving the company from sending out exploration teams on sometimes fruitless missions.
In one building in Dhahran, data scientists say they now capture the data from each oil well in real time, using 10 petaflops of computer power—capable of 10,000 trillion calculations a second. That cuts human hours on oil rigs, and allows the company to reach more remote reserves that were too complicated or costly to drill before.
Audits on oil wells, which previously took weeks to prepare, can now be generated from field notes within a day, using large language models, or LLMs, trained on years of auditing.
“AI will give certain companies a huge advantage over others,” Aramco’s executive vice president for innovation and technology, Ahmad Khowaiter, said in an interview in Dhahran. “And this is where we believe we are going to be able to take the lead.”
There is a reason for Khowaiter’s confidence: Aramco’s deep pit of money. While countless companies are adopting AI in their businesses, the company is among the few with the means to finance an all-out push into the technology, and to scale its uses globally. The $1.8 trillion company pumps about 9 million barrels of oil a day—9% of world consumption—and generated $440.8 billion in revenue last year, with nearly $23 billion in free cash flow this quarter. Aramco raised another $11.2 billion this month by floating 0.7% of its government-held shares on the Saudi stock exchange, known as the Tadawul, selling out the offering within hours. (Aramco’s first public offering, for 1.5% of the company, raised $30 billion in 2019.)
Aramco, whose revenues make up about 50% of Saudi Arabia’s economy, is not the country’s only AI investor. The Saudi sovereign wealth fund, called the Public Investment Fund, or PIF (funded largely by Aramco profits and run by Aramco chairman Yasir Al-Rumayyan), has been discussing a potential deal with VC firm Andreessen Horowitz, to launch a $40 billion AI fund in the kingdom; such a fund would make Saudi Arabia among the world’s biggest AI investors. Sizable AI investments are emerging, increasingly binding the Saudi royals to the likes of Amazon, Alphabet’s Google, and Silicon Valley’s VC firms.
Deploying billions in oil profits to build AI
In Aramco’s Dhahran complex, execs were keen to show how they had committed billions to building algorithms in-house, purchasing some of the world’s most powerful supercomputers, and luring AI know-how from tech giants abroad.
Khowaiter, for example, says Aramco is already using “several thousand” Nvidia chips for its regular computing, and that “we have hundreds on order” (the California company’s AI microchips cost more than $30,000 a piece). Khowaiter says Aramco is also eagerly awaiting Nvidia’s new-generation products.
Among the many recent signs of Aramco’s involvement in Saudi AI strategy:
In late May, Aramco purchased a 200-qubit quantum computer from French startup Pasqal, in which Aramco Ventures, its VC arm, invested $100 million last year. The French company says it has an “ambitious vision to establish a powerhouse for quantum research within Saudi Arabia.”
In March, Aramco Digital—a subsidiary launched last year by Aramco’s $7.5 billion VC arm—introduced an AI-powered operating system called Nour OS. The system has about 12 use cases, including financial and legal services, and human resources: While those are being used internally at Aramco, they’re potentially marketable beyond the company. Aramco Digital CEO Tareq Amin, former chief technology officer of Japan’s Rakuten, says that when Aramco hired him last year, he was told “to establish nothing less than the Google of the Middle East.”
Also in March, the oil company launched a gen AI large language model called Aramco Metabrain, which it claims is the first industrial LLM for oil and gas. Trained on nearly nine decades of carefully logged Aramco data, the model’s 250 billion parameters can predict factors like output and drilling speeds. Engineers in Dhahran told Fortune they will begin using a 1 trillion parameter version later this year.
Big Oil gets religion on computing
Aramco is hardly the only oil company to seize on AI and quantum. Exxon Mobil says its collaboration with IBM, signed in 2019 to develop uses for quantum computing, allows it to simulate nature-based energy sources. Shell has been experimenting with AI to boost production and predict problems since 2016, according to Tom Siebel, CEO of C3.ai, who says his company’s technology was “monitoring every device on every offshore oil rig in real time.”
But the sheer immensity of Aramco’s operations could set its AI commitment apart. “There’s a difference between the big oil companies and the smaller companies,” says Samta Kapoor, EY Americas AI lead in San Francisco. “Building an LLM from scratch takes a lot of resources, a lot of compute, a lot of very, very highly qualified data scientists [who] are not easy to find.”
In Aramco’s upstream innovation center, its avatar Ariva calculated the exact drilling rates best suited to increase output at a well site, using years of data on water use, carbon emissions, and oil reserves. “We don’t even have to drill it,” Khowaiter says. “We know exactly what the geology will be.”
Still unclear, for now, is how much money it will cost Aramco, and the Saudi royals, to use its wealth of data to become a global AI force.
