AI race boosts chips, cloud and power suppliers

For investors, the implication is straightforward: the AI trade is no longer only about software monetization. It is about the toll roads of the AI economy — chipmakers, equipment vendors, memory, foundry capacity, cloud infrastructure, power and grid buildout, and the allies that sit inside those supply chains. Nvidia, Microsoft and AMD remain central exposures, but the bigger message is that geopolitical fragmentation is adding another layer of demand for domestic and allied capacity.

The market is already paying attention. Nvidia has held above its 50-day and 200-day moving averages, with the shares recently around $220 after touching the low- to mid-$200s, while AMD has rebounded to about $514 and sits well above both long-term averages. Microsoft, after a deep summer selloff, has clawed back toward $496, though it remains below its 200-day line. The technical picture is consistent with a market that still wants the AI story, but is increasingly differentiating between beneficiaries of capex and companies exposed to export controls, customer delays and regulatory risk.

That split is exactly what this new geopolitical order creates. The U.S. bloc, anchored by Japan, South Korea and the Netherlands, has the most critical choke points in chip design, lithography, manufacturing and advanced packaging. China’s answer is to offer infrastructure, training and open-model access to emerging markets that want AI capability without total dependence on Washington. Countries in the Global South may not be choosing sides on ideology so much as on access: capital, compute and autonomy. India’s decision to sign the U.S.-led pact while preserving strategic flexibility captures the middle path many nations will try to follow.

This is where the market underestimates the second-order effects. The real beneficiaries are not only the obvious semiconductor leaders, but also the industrial and infrastructure names that supply land, power, cooling, networking and grid equipment. Nvidia’s own filings warn that expanding land, power and energy to meet demand is a multi-year challenge. Broadcom and Oracle have also flagged that AI buildouts can be delayed by capital spending, supply constraints and geopolitical frictions. That tells you the bottleneck is widening, not narrowing.

The loser set is equally clear: companies dependent on unrestricted global access, hyperscalers facing higher compliance costs, and chip designers vulnerable to tighter export controls. AMD’s filing makes plain that further semiconductor restrictions could curb exports. Microsoft has warned that tariffs, shifting AI export-control policy and sanctions disputes can raise costs and accelerate sovereignty efforts abroad. In other words, the AI race is becoming less global just as the capex cycle becomes more expensive.

The strategic takeaway is that the most attractive upside may now sit in the picks-and-shovels of AI sovereignty. I would rather own the infrastructure enablers, power beneficiaries and domestic supply-chain winners than chase every model narrative. The next phase of returns should come from the companies that make AI physically possible — not just the ones that make headlines.