India's UAE AI pact threatens Amazon, Microsoft, and Google

India is making a clear move against the default AI cloud stack. In partnership with the United Arab Emirates, it is trying to ease the grip that Amazon, Microsoft, and Google have had on artificial intelligence computing, and the vehicle is G42, the Abu Dhabi company backed by sovereign wealth fund Mubadala. The headline detail here is not just that a foreign partner is involved. It is that G42 will deploy a U.S.-designed supercomputer in India, which creates a very different model from the usual setup where companies and governments simply rent compute from the big cloud platforms and live inside someone else’s infrastructure.

That matters because AI is only as powerful as the hardware behind it, and access to compute has become one of the most valuable bottlenecks in the whole market. If a government can own or control that hardware, it is no longer fully dependent on Amazon Web Services, Microsoft Azure, or Google Cloud for the guts of its AI ambitions. Rest of World’s reporting frames this as more than a procurement deal. It is a potential template for governments that want to build AI capability without handing the keys entirely to the largest U.S. cloud providers. For decision-makers, that means the AI infrastructure market may not stay organized around a few dominant vendors forever, especially if state-backed alternatives can offer countries more control over where the hardware sits and who can use it.

The big strategic shift is about leverage. Cloud giants have spent years turning compute into an easy on-ramp: spin it up, pay for it, scale when needed, and let the vendor handle the rest. That model is wildly efficient, which is why it became standard. But it also concentrates power. The more a country depends on a handful of providers for advanced AI infrastructure, the more exposed it is to pricing, access, and policy decisions made elsewhere. A partnership like India’s with the UAE points to a different instinct: keep the convenience of cutting-edge infrastructure, but reduce the dependence. In plain English, it is an effort to own the pipes, not just rent them.

G42’s role is what makes the arrangement especially notable. The company is not being described here as a generic vendor. It is backed by Mubadala, Abu Dhabi’s sovereign wealth fund, which means this is not just a commercial play. It has the weight and patience that state capital can bring to infrastructure bets. That can matter a lot in AI, where the buildout is expensive, strategic, and often justified by long-term national interest rather than immediate profit. When a government collaborates with another government-linked entity to deploy major AI hardware, it can move faster on strategic goals than a standard private-sector procurement might allow. It can also send a message: AI infrastructure is now a matter of sovereignty, not just software purchasing.

The source also points to a subtle but important wrinkle: the supercomputer is U.S.-designed. That tells you the global AI supply chain is not neatly splitting into separate blocs. Instead, the parts are being rearranged. Design, ownership, location, and control can all live in different places. A U.S.-designed machine deployed in India by a UAE-backed company shows how countries can work around the old binary of either buying from U.S. cloud companies or building everything from scratch. This kind of hybrid setup may become attractive to other governments that want advanced capability without ceding too much control to one foreign vendor or one foreign jurisdiction.

For Amazon, Microsoft, and Google, the stakes are obvious even if this one deal is still limited in scope. Their cloud dominance has rested not only on technology, but on inertia. Once compute, data workflows, and enterprise contracts settle into a platform, they are hard to dislodge. A government-backed alternative that offers a credible path to own AI hardware could chip away at that inertia, especially in markets where national control, data sensitivity, or procurement politics matter. Even if the deal does not immediately dent their global revenue, it gives other buyers a reference point: there is more than one way to build AI capacity, and the biggest cloud names do not have a monopoly on the answer.

For executives, the practical takeaway is that AI infrastructure is becoming a geopolitical decision as much as a technical one. Boards evaluating where to place workloads, how to structure partnerships, and whether to depend on one cloud provider should read this as an early sign of a broader market pattern. Governments are no longer passive customers in the AI race. They are trying to shape the hardware layer itself. If that trend spreads, vendors will have to compete not just on price and performance, but on sovereignty, flexibility, and who gets to own the critical machines that run the future.