From data centers and chips to sovereign models and autonomous vehicles, the Kingdom is building an AI ecosystem designed to control more of the technology stack and capture more of its economic value. 

[For more news, click here]

Saudi Arabia opened the fifth edition of LEAP in Riyadh on 31st August with a proposition that goes beyond digital transformation, it wants to turn its biggest structural advantage, abundant power, into an export business built around artificial intelligence. Across eight hours of main stage programming, LEAP framed AI infrastructure as a national industrial asset, with the Kingdom building the power, compute, models and applications to produce intelligence at scale. 

“Today we export energy to the world. The new export of energy now translates tokens, tokens as these words. The more words that people use with AI, the more net energy export of intelligence that we will become.” Tareq Amin, CEO, HUMAIN 

Saudi Arabia's AI bet starts with power 

Amin put the global picture at 82GW of AI capacity against a forecast need of 220GW, and called that an underestimate. HUMAIN has 14GW secured on the grid and 211 land banks across seven AI campuses, against a 6GW target he compared with three-quarters of Singapore's power consumption. 

Permitting for the first data center took 14 days, which Amin called the biggest surprise of his career after building in the US, China, India and Japan. Luma AI's Amit Jain argued that cutting grid connection times from seven years to 18 months gives the Kingdom an advantage competitors will struggle to overcome. At NEOM's Oxagon, HUMAIN and DataVolt have broken ground on a 360MW AI factory with a 1.5GW ambition, billed as the world's first fully renewable data center. 

The compute is being sold before the buildings exist 

HUMAIN has pre-sold capacity to global AI companies, AMD 500MW, xAI another 500MW, Together AI 250MW, AWS 100MW and Luma AI 75MW. 

“We're not building empty factories. The factories need customers, need tenants.” Tareq Amin, CEO, HUMAIN 

“Inference has overtaken training as a primary workload.” Dr. Lisa Su, Chair and CEO, AMD 

AMD's largest inference cluster outside the US is now live in the Kingdom, with 13,000 GPUs targeted by the fourth quarter. xAI's first deployment outside the US starts at 50MW and scales to 500MW, and Groq racked and powered 19,000 LPUs in under seven days. Four silicon architectures came online in a single year, and customers in 120 countries now consume inference from Saudi infrastructure. 

HUMAIN's Node platform abstracts the underlying silicon, model and tuning requirements, letting customers work across American, European and Chinese open models alongside its own. Voice was built from one million hours of Arabic speech, and M3 is an open weight model commissioned with MiniMax and trained on one trillion tokens of regional data, which Amin called a gift to the Arab world. 

Building a sovereign AI stack 

HUMAIN is pursuing sovereignty beyond keeping data inside national borders. HR, finance, legal and payroll now run on HUMAIN One, which Amin said was the first enterprise operating system built anywhere in the world, and Microsoft has integrated its Office suite into it, billed as a world first. HUMAIN is also running Adobe Create workloads in-Kingdom on Qualcomm silicon and building an air gapped agent marketplace with Turing. 

Amin's diagnosis of why enterprise AI fails elsewhere is architectural rather than technological, being decades of silos and stacked data lakes. At HUMAIN, agents already carry out 40% of existing work, and sovereignty here is less about where a model is trained than who owns the system through which work gets done. 

Sovereignty moves from slogan to strategy 

Officials and executives gave the word different meanings but pointed to the same shift. Control over infrastructure, models, data and critical systems is becoming a national economic strategy. Cisco's Chuck Robbins noted that “everybody's definition is slightly different.” 

“Digital sovereignty is not a choice between global and local technology. True sovereignty means no one else can turn off your critical systems.” Muhammad Ali, SVP and Head of Consulting, IBM 

Turing CEO Jonathan Siddharth defined sovereignty as owning the learning loop, since enterprise knowledge often sits outside the training distribution of frontier models, making employee corrections valuable institutional data. He gave the day's most concrete argument for controlling your own stack. 

“Earlier this year, there was an export directive due to which a frontier model just went dark. If your operations were relying on that frontier model in your day-to-day, your operations would have gone dark too.” Jonathan Siddharth, CEO, Turing 

Enterprise AI faces the harder test 

If infrastructure was the easiest part of the story to measure, enterprise adoption was harder. Aramco processes more than 10 billion data points a day and runs inhouse models across tens of thousands of employees, while its CyberMind system handles roughly one billion potential threats a year. 

“While consumers can be impressed if a model is fast or creative, in industry, a model must be secure, it must be reliable, and it must be explainable.” Ahmad Al-Khowaiter, EVP, Technology & Innovation, Aramco 

Sami Al Najmi described Aramco's progression as a ladder running from descriptive to predictive, generative, agentic and finally autonomous operations, but autonomy depends on the quality of the underlying data. 

“To build towards a real autonomous operation, the data maturity has to be near 100%. We cannot take any risk into autonomous operation.” Sami Al Najmi, SVP, Digital and Information Technology, Aramco 

Human oversight remains for high-risk decisions, with systems tested against digital replicas first. PwC's Joe Atkinson warned that roughly 20% of companies account for around 74% of total AI gains, while Salesforce's Mohammed Al Qahtani, citing McKinsey, put enterprise AI success at 6%, blaming companies for buying data centers and models without first determining what the business needs. He expects this to be “the last generation who will witness 100% human only workforce.” 

The Ministry of Interior's digital identity platform has onboarded 32 million people and completed 6.5 billion authentications, and agentic AI on passport control completed 98% of services without human interaction. The more useful finding came from analyzing the conversations; citizens were not asking about services; they were asking about outcomes. The ministry now runs an orchestrator that reads intent and distributes tasks across departments and is extending the approach to field responders through a system called Meydan. 

“Having more data to the decision maker doesn't help. Having the right information in that moment, and the contextual data, will make all the difference.” Assistant Minister of Interior for Technology Affairs, Saudi Arabia 

From physical AI to the next funding gap 

HUMAIN's joint venture with Applied Intuition targets thousands of autonomous trucks on Saudi roads by 2030, followed by robotaxis, ports and mining. 

“This is physical AI at a national scale. It's the first time that's being done ever in the history of mankind.” Qasar Younis, Co-founder and CEO, Applied Intuition 

In the US and Germany, trucks, robotaxis, ports and mines are separate customers with separate deployments. In Saudi Arabia, HUMAIN is the common counterparty across all four. Younis noted that mining employs roughly 1% of the global workforce but accounts for around 8% of workplace fatalities. 

The other major test is capital. Jeff Hoffman, chair of the Global Entrepreneurship Network, delivered the day's only substantial outside criticism after his microphone was cut mid answer, arguing that founders and investors remain disconnected, and that it is not the culture here to stand up and say look at me, so the media has to say it for them. 

“So step number one to becoming a global ecosystem is: do not cut off the microphone of foreigners.” Jeff Hoffman, Chair, Global Entrepreneurship Network 

Three investors independently echoed a related diagnosis: the gap is no longer at the earliest stage, but when companies need to scale. 

“The massive gap now in Saudi Arabia is the Series C to Series D.” Omar Al Majdouie, Founding Partner, Raed Ventures 

The international test of Saudi Arabia's model 

Pakistan's H.E. Shaza Fatima Khawaja said her government had moved from “fancy big projects” to foundational infrastructure. Governments should create the conditions for technology rather than become technology companies, she argued, because “the moment a government starts running projects, it starts losing its way.” 

“Pakistan's national stack is built on open source. And when it is complete, we will share it, free, with every frontier and emerging nation that wants it, not sold, shared.” H.E. Shaza Fatima Khawaja, Federal Minister for IT and Telecommunication, Pakistan 

Palestine signed the DCO Charter and launched a Digital Prosperity Plan targeting 10,000 people trained and market access for more than 250 startups by end-2028. 

“Do not look at Palestine only through the lens of what it needs. Look also at what Palestine can build.” Abdul Razzaq Al-Natsheh, Minister of Communications and Digital Economy, State of Palestine 

Austria provided the final note of caution. Its state secretary acknowledged the country remains dependent on other nations for digital infrastructure and came looking for partnerships that preserve control. 

“I don't want to wake up in 10 years and the digital world was built without Austria.” Alexander Pröll, State Secretary for Digitalization, Constitutional Affairs, the Civil Service and Coordination, Federal Chancellery of Austria 

Saudi Arabia is not trying to catch up with a digital world built elsewhere. It is building the power, compute, models, software, talent and capital to help shape what comes next and to own more of the value along the way. 

Related articles 

Saudi Arabia Bets On Concrete to Win the AI Race, And a New Partnership Shows How 

HUMAIN and Accenture Are Taking Saudi Arabia's AI Ambitions From Blueprint to Operation 

Why Saudi Arabia Could Deploy Autonomous Trucks at National Scale Before the US