AI Infrastructure Newsletter

The not-usually-so-contentious trade and tech relationship between the U.S. and South Korea is top of mind this week, with the convergence of several developments.

AI infrastructure continues to drive eye-popping numbers: Right now, the headliner is Microsoft’s deal for GPUs from Nebius (which was spun out of Russian internet company Yandex two years ago).

Today’s top new focuses largely on AI expansion: Mistral rakes in a 1.3 billion Euro investment in support of European AI sovereignty, while LG lands a deal to provide cooling tech for data centers in the Middle East.

AI infrastructure is rapidly evolving across multiple fronts, as shown our top stories for today, which underscore both the supply constraints and massive investment commitments happening in the space.

AI infrastructure is shifting gears fast. OpenAI is moving beyond Nvidia, locking in a $10 billion deal with Broadcom to mass-produce its own chips from 2026.

Tech sovereignty and supply-chain control continues to be a major front in the global battle over AI. This really boils down to two approaches: Build up the domestic ecosystem options — as we see SAP doing in Europe, with a €20 billion investment to support cloud-based data sovereignty and AI adoption — while also trying to undercut your competitors, if you have the trade clout to do so.

That may be overstating it, but AI’s rapid acceleration is definitely reshaping global technology strategies. Today, we have three examples.

In today’s lead stories, we look at some of the big questions around AI and AI infrastructure. First — where are all these GPUs going, anyway? Nvidia’s disclosure that just two customers account for nearly 40% of its revenues sheds some light on why GPUs are so expensive and hard to come by.

The AI data center chip market is still surging, with Omdia projecting $286 billion by 2030, though growth is already slowing after a rapid 2022–2024 run and spending is expected to peak by 2026 before easing in subsequent years.

Global giants like Disney, Foxconn, Hitachi, SAP, and TSMC are adopting Nvidia’s new RTX Pro enterprise servers to retool conventional computing clusters into “AI factory” infrastructure.