India’s Bold Bet on UAE-Backed AI Supercomputer to Counter Big Tech Cloud Control

India just took a calculated step to loosen the hold of U.S. cloud giants on its artificial intelligence future. On May 15, Abu Dhabi-based G42 formalized terms with the Indian government to deploy Condor Galaxy India. The system packs 64 Cerebras CS-3 supercomputers. It delivers 8 exaflops of compute power. All of it lands on Indian soil.

Operators from a G42 unit will install, run and maintain the cluster. Cerebras supplies technical backup. Data stays subject to Indian rules. The Centre for Development of Advanced Computing partners with G42’s Core42 subsidiary. This arrangement marks a pragmatic bid for control over critical AI resources without full isolation from global suppliers.

The deal arrives at a telling moment. India already holds pledges worth $45 billion from Amazon, Microsoft and Google. Those commitments center on Nvidia-based clouds. Microsoft alone plans $17.5 billion over four years. Google committed $15 billion. Amazon Web Services earmarked $12.7 billion. India’s national AI mission currently offers 34,000 Nvidia processors to researchers and companies. Officials aim to reach 100,000 by year-end.

Yet reliance on rented capacity from three American firms leaves policy makers uneasy. Governments everywhere now grasp a basic truth. Access to advanced AI can vanish with a single foreign policy shift. Recent restrictions by Anthropic on its latest models for non-U.S. users drove that point home. Indian tech leaders and officials immediately debated how to reduce such exposure. TechCrunch reported the episode as a wake-up call for strategic autonomy.

So India turned to the UAE. G42, backed by Mubadala sovereign wealth fund, had already built deep ties with Cerebras. The two companies launched the Condor Galaxy network in the United States. Clusters operate in California, Texas and Minnesota. That track record gave New Delhi confidence the hardware would arrive ready to run.

Cerebras takes a different approach from dominant players. Its wafer-scale engine creates the largest AI chip on the market. One dinner-plate-sized piece of silicon handles tasks that require thousands of smaller Nvidia processors linked together. The design shines at fast inference. It runs trained models at speed. That fits India’s stated priorities in healthcare, agriculture, language applications and public services. Training massive new models from scratch matters less right now than deploying AI across a population of 1.4 billion.

Cameron Kerry noticed the logic. The former acting U.S. commerce secretary and current Brookings Institution fellow described the move to Rest of World. “This is an example of India’s pragmatic approach to AI sovereignty, using the power of its scale to adapt what’s available from other countries, whether AI leaders like China and the U.S. or others, to adapt to its own needs.” He co-authored a February Brookings report that argued no nation controls every layer of AI. Governments must combine pieces from multiple partners.

Chris Miller sounded a note of caution. The Tufts University professor and author on semiconductor geopolitics told the same outlet that G42 faces steep competition. Amazon, Microsoft and Google bundle hardware, software, tools and support. “A key question will be whether they can offer competitive software and other services to their data-center customers,” Miller said.

G42 calls its broader effort the Intelligence Grid. It builds, owns and operates AI facilities for national governments. India signed on first. Talks continue with others, though names stay private. The UAE company also partners at home with the same American firms it challenges abroad. Microsoft pledged $15.2 billion for UAE data centers through 2029, routing some work through G42 subsidiary Khazna. Amazon runs a full cloud region there.

Cerebras itself went public on Nasdaq on May 14, 2025. The offering raised $5.55 billion. It stood as the largest U.S. tech listing since Uber in 2019. G42 and Mohamed bin Zayed University of Artificial Intelligence together drove 86 percent of Cerebras revenue in 2025, according to SEC filings. The wafer-scale pioneer counts those two UAE customers as its biggest.

But control questions linger. India’s Digital Personal Data Protection Act of 2023 permits transfers to most countries. It blocks only a short blacklist. Whether the government wrote tighter conditions into this specific arrangement remains unclear. G42 declined to disclose financial terms or final hardware ownership. Kerry warned the deal may grant less operational independence than it first appears.

Still, the message lands. More governments want physical machines under their jurisdiction instead of API calls to distant clouds. That pressure will force Amazon, Microsoft and Google to adapt. “The better they respond, the better for the U.S. position in the world regardless of what the U.S. government does,” Kerry told Rest of World.

Recent funding news underscores India’s momentum. On June 15, Bengaluru-based Sarvam AI raised $234 million at a $1.5 billion valuation. The round, led by HCLTech, makes it the country’s newest AI unicorn. TechCrunch noted the investment reflects growing focus on domestic capabilities amid sovereignty concerns. Sarvam and peers now have fresh capital to build on top of the new Cerebras infrastructure.

The Condor Galaxy India project builds directly on earlier India-UAE agreements. It follows the fifth strategic dialogue in December 2025 and a visit by UAE President Sheikh Mohamed bin Zayed Al Nahyan in January 2026. Those talks produced a broad technology cooperation framework. Officials announced the supercomputer on the sidelines of the AI Impact Summit in New Delhi. G42’s official release described it as a foundational asset for national AI goals. PR Newswire carried the details.

Technical specifics impress. The 8-exaflop cluster extends the Condor Galaxy concept beyond U.S. borders into a major emerging market. Cerebras CS-3 systems rely on wafer-scale technology that avoids some bottlenecks of traditional GPU clusters. Energy efficiency and inference speed stand out as advantages for countries with massive populations and varied use cases.

Yet success hinges on execution. G42 must prove it can match the integrated offerings of U.S. hyperscalers. Indian developers need familiar tools, not just raw flops. Local researchers at C-DAC will test whether the system truly accelerates applications in Hindi, regional languages and domain-specific tasks.

And the broader pattern grows clearer. Nations from the Gulf to South Asia now treat compute infrastructure as strategic infrastructure. They refuse to outsource every layer. Full independence stays impossible. Supply chains for chips, energy, talent and software cross too many borders. But selective partnerships that place hardware domestically and keep data governance local offer a middle path.

India’s scale gives it bargaining power few others possess. The country can demand customized deployments. It can require local operation. It can direct compute toward public priorities. That combination makes the G42-Cerebras deal more than one contract. It signals a template other governments may follow.

Big Tech will watch closely. So will Washington. Export controls, national security reviews and alliance management all intersect here. The White House previously authorized advanced AI semiconductor exports to G42, easing earlier tensions. Those approvals enabled parts of this architecture.

Results will take time to measure. Installation, integration, software optimization and real-world pilots lie ahead. But the direction is set. India no longer accepts a future where its AI ambitions rest entirely on foreign clouds. It now owns a powerful alternative on its own territory. That changes the equation. And it won’t be the last such shift.