HUMAIN
Last reviewed
May 31, 2026
Sources
18 citations
Review status
Source-backed
Revision
v2 ยท 2,074 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Last reviewed
May 31, 2026
Sources
18 citations
Review status
Source-backed
Revision
v2 ยท 2,074 words
Add missing citations, update stale details, or suggest a clearer explanation.
HUMAIN is a Saudi Arabian artificial intelligence company launched in May 2025 and owned by the Public Investment Fund (PIF), the kingdom's sovereign wealth fund. It is chaired by Crown Prince Mohammed bin Salman, who also chairs PIF. The company describes itself as a full-stack AI business, meaning it aims to work across the whole technology chain, from data centers and AI infrastructure through cloud services and large language models to consumer and enterprise applications. HUMAIN sits at the center of Saudi Arabia's push to turn itself into a global AI hub under the Vision 2030 economic plan. [1][2][3]
The launch was announced on 12 May 2025, days before a visit to Riyadh by United States President Donald Trump and a Saudi-US investment forum at which HUMAIN unveiled a set of large chip and infrastructure deals with American technology firms. [2][4]
HUMAIN is owned by the Public Investment Fund, which manages close to a trillion dollars on behalf of the Saudi state and has become the main vehicle for the kingdom's economic transformation. Placing Crown Prince Mohammed bin Salman as chairman signals how closely the company is tied to government strategy rather than treated as an ordinary commercial startup. In late 2025 PIF and the national oil company Aramco signed a non-binding term sheet for Aramco to take a significant minority stake in HUMAIN, with PIF keeping majority ownership and Aramco contributing AI assets and talent, including its Aramco Digital unit. [1][3][5]
Vision 2030 is the long-running plan, launched in 2016, to reduce Saudi Arabia's dependence on oil revenue by building new industries in areas such as tourism, manufacturing, and technology. HUMAIN is the AI piece of that plan. The stated goal is for Saudi Arabia to host large amounts of AI compute, train models in Arabic, and supply AI services across the Middle East and beyond, rather than relying entirely on data centers and models built elsewhere. This ambition is often described as building sovereign AI, meaning capabilities that a country controls within its own borders. [1][3]
The approach mirrors moves by neighboring states. The United Arab Emirates has pursued a similar path through G42 and related ventures, and the two Gulf programs are often compared because both depend on access to American chips and both frame AI as a matter of national strategy. [6]
When HUMAIN launched, PIF described four areas of activity that together make up the full-stack model. [1][7]
| Layer | What it covers |
|---|---|
| AI infrastructure | Building and operating large data centers and the compute hardware inside them |
| AI cloud | Offering cloud computing and access to AI compute as a service |
| Data and models | Developing models, including Arabic large language models, and the data behind them |
| Applications and solutions | Building tools and products for government, business, and consumers |
The infrastructure layer is the most capital intensive and the part most dependent on foreign suppliers, since the advanced processors that train and run large models are made by a small number of mostly American companies. The cloud layer is meant to package that compute so customers in the region can use it without building their own facilities, putting HUMAIN in competition with global providers such as Amazon Web Services, Microsoft Azure, and Google Cloud while also partnering with some of them. HUMAIN has said it wants Saudi Arabia to become one of the largest AI compute providers in the world. [1][7][8]
The model layer centers on ALLaM, a family of Arabic large language models. ALLaM was first developed by the Saudi Data and Artificial Intelligence Authority (SDAIA) through its National Center for AI, and an early 13 billion parameter version trained on Arabic and English was released on IBM watsonx in 2024. HUMAIN later built a larger flagship version, ALLaM 34B, which it has called the most capable Arabic model built in the Arab world and which it trained on one of the largest Arabic datasets assembled to date. The models are built to handle Arabic and its dialects well, an area where many leading Western models have historically been weaker because most training data is in English. [9][10][11]
In August 2025, HUMAIN released HUMAIN Chat, a consumer chatbot powered by ALLaM 34B and presented as the kingdom's first sovereign Arabic AI assistant. It can hold conversations across Arabic dialects, switch between Arabic and English in the same exchange, take voice input, and run real-time web search. It was positioned as an Arabic-first option for the more than 400 million Arabic speakers underserved by general assistants such as ChatGPT, with the data hosted inside Saudi Arabia. The app launched first in the kingdom on web, iOS, and Android, with regional and global expansion planned. [11][12]
HUMAIN has also promoted HUMAIN ONE, described as an enterprise operating system for autonomous AI agents. Built on the company's orchestration engine and on ALLaM, and developed with AWS, it lets businesses run tasks across functions such as human resources, finance, and procurement through a single natural-language interface. The product was shown at the Future Investment Initiative in Riyadh in October 2025 and developed further into 2026. [13]
The most attention around HUMAIN came from a cluster of partnerships announced during the Saudi-US investment push in mid-May 2025. The figures below are as stated by the companies and reported at the time, and several describe multi-year commitments or targets rather than money already spent. [4][8]
| Partner | Announced deal |
|---|---|
| NVIDIA | Sale of about 18,000 GB300 Grace Blackwell chips to power data centers with roughly 500 megawatts of capacity, part of a plan to deploy several hundred thousand GPUs over the following years in what NVIDIA called AI factories |
| AMD | A collaboration valued at up to 10 billion dollars to deploy about 500 megawatts of AI compute over five years using AMD Instinct GPUs, EPYC processors, and networking, across sites in Saudi Arabia and the United States |
| Qualcomm | An agreement to develop AI data centers in Saudi Arabia using Qualcomm data center processors and inference technology, alongside a design center to build local semiconductor talent |
| Amazon Web Services | A plan to invest more than 5 billion dollars in a dedicated AI Zone in Saudi Arabia, with up to 150,000 AI accelerators and access to NVIDIA and AWS Trainium chips |
| Cisco | A collaboration on AI infrastructure and networking, plus an AI institute at King Abdullah University of Science and Technology and training for up to 500,000 people |
NVIDIA chief executive Jensen Huang appeared at the forum, and the NVIDIA agreement was the headline item because it used the company's newest Blackwell-generation systems. The AMD collaboration was notable for spanning both Saudi Arabia and the United States and for using AMD's open compute approach as a counterweight to NVIDIA's dominance. The Qualcomm deal reflected that company's move beyond mobile chips into data center processors, and a later agreement targeted 200 megawatts of Qualcomm inference systems from 2026. HUMAIN also worked with Groq to host inference workloads in its data centers. The buildout is meant to scale over time, and HUMAIN has set targets of about 1.9 gigawatts of capacity by 2030 and roughly 6 gigawatts by 2034, though much of that remains a plan rather than installed hardware. [4][8][14]
The deals were possible in part because of a shift in United States policy. Advanced AI chips are subject to American export controls, and under the previous administration a framework known as the AI diffusion rule would have placed tiered limits on how many high-end AI accelerators could flow to countries such as Saudi Arabia and the UAE. On 12 May 2025 the United States Commerce Department announced it was rescinding that rule, around the same time as the Riyadh forum. Approvals still came with conditions. Reporting indicated an initial clearance for the Gulf state firms to buy on the order of 35,000 advanced chips worth roughly a billion dollars, subject to security and reporting requirements and government-to-government assurances. [15][16]
The arrangements drew scrutiny in Washington. Some lawmakers and analysts worried that concentrating large amounts of cutting-edge American compute in the Gulf could create a route for the technology to reach China, given commercial and diplomatic ties across the region, and they pushed for security conditions on the exports. Supporters argued that anchoring Gulf AI on American hardware and standards keeps these states inside the United States technology sphere rather than pushing them toward Chinese suppliers. This tension, between commercial opportunity and control of strategic technology, is the main geopolitical backdrop to HUMAIN's growth. [16][17]
HUMAIN is led by chief executive Tareq Amin, who took the role at launch. Amin previously ran Aramco Digital, the technology arm of the Saudi oil company, and earlier held senior roles building telecommunications networks at Reliance Jio in India and at Rakuten in Japan, so his background is in large-scale digital and network infrastructure rather than in AI research. That profile fits a company whose first major moves have been about data centers and chips. Crown Prince Mohammed bin Salman serves as chairman, reinforcing the link to PIF and to state strategy. [3][18]
HUMAIN is one of the clearest examples of a state using sovereign wealth to try to buy a place in the AI industry quickly. By combining cheap energy, abundant capital, and political backing, Saudi Arabia hopes to host a meaningful share of global AI compute and to lead AI development in Arabic. If the announced infrastructure is built, the kingdom would gain capacity that few countries outside the United States and China can match. [1][8]
The limits are also real. Most of the core technology, the chips, the chip designs, and much of the model expertise, comes from abroad, so HUMAIN's plans depend on continued American export approvals and on relationships with a handful of suppliers. Building and filling gigawatts of data centers takes years, and headline deal values are commitments that can change. The company's models, while strong in Arabic, are newer and less proven than the leading systems from larger labs. How far HUMAIN moves from announcements to operating capacity, and whether it competes on model quality rather than only on scale and access to compute, will determine its place in the field. [8][16]