Samsung just told South Korea’s president it plans to spend 1,000 trillion won — roughly $648 billion — over the next decade on chip factories, AI data centres, batteries, and displays. The same week, Trump’s administration asked OpenAI to restrict its next model to “trusted partners” only. Anthropic got limited approval to release its most capable model to select companies and government agencies. None of these stories made the front page together, but they tell one story: governments are now actively steering who gets access to the most powerful AI in the world, and the answer is increasingly “not everyone.”
The Trusted Partners Era Has Begun
OpenAI’s decision to limit its next major model release to vetted partners, approved case-by-case by US government agencies, is not a business choice — it is a policy decision made for the company by Washington. The same pattern is playing out with Anthropic, whose most capable model remains restricted pending export review. For years the AI industry’s default was move fast and ship broadly. That default is gone. The companies building frontier models are now operating closer to how defence contractors operate — capability exists, but who gets to use it is a government decision, not a product decision.
This matters enormously for any business or developer outside the small circle of “trusted” markets. If your company sits in a country or sector that does not make the approved list, you are not getting the newest model — regardless of your budget. The practical effect is a two-speed AI world: insiders building on the frontier, everyone else building on what was frontier eighteen months ago.
Why Samsung Is Betting $648 Billion Right Now
Samsung’s decade-long commitment is the clearest signal yet that the chip and data centre arms race has no ceiling in sight. The investment spans chip fabrication in South Korea’s southwest, new AI data centres, and battery manufacturing — three pieces of the same puzzle: control the physical supply chain that the AI industry depends on. South Korea is making a national bet that whoever controls advanced memory and AI infrastructure manufacturing controls genuine geopolitical leverage for the next two decades, not just corporate profit.
The memory shortage already biting Apple and Microsoft is the early warning sign. Smaller hardware companies without Samsung-scale supply agreements are facing what industry insiders are calling an existential crisis on memory chip access. If you run a business that depends on consumer electronics, server hardware, or any device with meaningful onboard memory, expect cost pressure to continue climbing through the rest of the year.
What This Means If You Are Not Silicon Valley
For Indian businesses and developers, the practical takeaway is to diversify model dependency now rather than later. Relying entirely on a single frontier model provider that may restrict your access based on geography or government approval is a real operational risk in 2026 — not a hypothetical one. OpenRouter and similar aggregators that let you switch between multiple AI providers are worth building into your stack specifically to hedge against this. China’s Zhipu AI is reportedly closing the capability gap with Western frontier labs precisely because it operates outside US export restrictions — a reminder that restriction in one market often accelerates capability in another. Also read our breakdown of GitHub Copilot’s pricing shock for the parallel story of AI access becoming more expensive even where it is not restricted.
KickassOpinion Verdict
AI access is becoming geopolitical, not just commercial. The $648 billion Samsung is committing and the gatekeeping OpenAI and Anthropic are now operating under are two sides of the same coin: physical infrastructure and software capability are both being nationalised in everything but name. If your business strategy assumes unrestricted access to the best AI tools indefinitely, this is the week to question that assumption. Geopolitical AI Risk Rating: 8/10 — plan for it now.