AI sovereignty push accelerates as nations race to break cloud dependency

India and the UAE are partnering on a joint AI sovereignty initiative designed to bypass dominant US cloud providers such as Google and Microsoft, while Europe is racing to train frontier models on its own compute. The moves reflect a growing unease with the concentration of AI infrastructure among a few American hyperscalers.

Sarvam AI, a Bengaluru based startup, became India's newest AI unicorn on June 15, 2026, after raising $234 million in a funding round led by HCLTech. HCLTech invested $150 million in the round. The company develops large language models for Indic languages and aims to train them on domestic hardware.

Europe builds a mesh for sovereign compute

A grassroots project called Euromesh, hosted on GitHub, is attempting to answer whether Europe can train a frontier AI model using only compute resources it owns. The project inventories European research clusters and national supercomputers to aggregate capacity. It has drawn 33 points on Hacker News and active discussion.

Key factors driving the sovereignty push include:

• US export controls on advanced AI chips to certain nations have accelerated Europe's desire for independent training capability.
• India and the UAE's partnership through G42 and Cerebras aims to deploy wafer scale clusters outside US jurisdiction.
• The prompt to app loop popularized by platforms such as Replit, Lovable, and Base44 runs on hyperscaler infrastructure, creating de facto vendor lock in for AI generated applications.
• Brussels has explicitly cited the risk of model access being revoked by US regulators as a catalyst for European investment.
• Total global investment in sovereign AI cloud infrastructure is projected to exceed $40 billion by 2028, according to industry estimates.

Brussels warns of geopolitical risk

The Register reported on June 15, 2026 that a US clampdown on Anthropic's model availability in Europe has sent EU sovereignty initiatives into overdrive. European Commission officials stated that the incident proves Europe needs greater technological independence and that reliance on US cloud providers for AI training is a national security vulnerability.

What comes next is a fragmented AI cloud landscape. The Middle East and South Asia are building their own stacks. Europe is pooling academic supercomputers. US hyperscalers face a future where their largest growth market may also be their most determined competitor. For now, the AI applications most users interact with continue to run on AWS, Azure, or GCP, but that may not be true for long.