Investors widen AI capex bets

The AI stock story is getting wider. For two years, the cleanest trade was basically “buy Nvidia” and let hyperscaler spending do the rest. But the latest move in the market says investors are starting to price a different idea — that the real prize is the whole supply chain around AI infrastructure, not just the GPU at the center of it. This week, that showed up in sharp moves across Intel, AMD, Micron and SanDisk, while Nvidia lagged the newer excitement. ### Why are investors suddenly looking past Nvidia? Because the spending wave itself is getting easier to see. When cloud giants and platforms raise AI infrastructure budgets, the money does not stop at GPUs. It flows into high-bandwidth memory, custom chips, networking gear, advanced packaging, storage and plain old server CPUs. Meta’s April 29 update was the cleanest recent signal — it raised 2026 capital expenditure guidance to $125 billion to $145 billion, citing higher component pricing and more data center buildout. (cnbc.com) ### Why is memory the first big second-order winner? Because AI servers are memory-hungry in a way normal computing was not. Training and inference clusters need huge amounts of high-bandwidth memory right next to accelerators, and they need lots of DRAM and flash around the system too. Micron has become the poster child for that shift. In March, it posted record fiscal Q2 2026 results and said the numbers reflected the “strategic value of memory in the AI era.” Earlier, it said its entire calendar 2026 HBM supply was already sold out on price and volume. (fool.com) ### So why are SanDisk and storage names moving too? Because once companies move from training giant models to actually deploying AI systems, storage starts to matter more. Agents, retrieval systems and enterprise AI workflows all create more data movement and more demand for fast flash. That is why storage names have started outperforming in this latest leg, with Western Digital and Seagate also drawing attention alongside SanDisk. It is the less glamorous layer, but it sits underneath real-world deployment. (investors.micron.com) ### Where do AMD and Intel fit in? Mostly in the “not every AI dollar buys a GPU” bucket. AMD still has accelerator exposure, but investors also like its broader data-center footprint. Intel’s rally is more about CPU demand and the idea that agentic AI workloads can expand server spending outside the classic Nvidia stack. CNBC described the move as a “changing of the guard” trade, with money rotating into chipmakers that were not the first-wave AI winners. (msn.com) ### What about networking and custom silicon? That is another big reason the trade is widening. Broadcom said fiscal Q1 2026 AI revenue hit $8.4 billion, up 106% year over year, driven by custom AI accelerators and AI networking. That is a huge clue. The market is rewarding the connective tissue now — not just the compute engine. Think of it like a gold rush where shovels were phase one, but now investors want the railroads, warehouses and power lines too. (cnbc.com) ### Why do foundries matter more in this version? Because somebody has to actually manufacture and package all of this. TSMC just posted another record quarter and said AI demand keeps growing. The bottleneck is no longer only chip design. It is also advanced packaging, yield, capacity and how fast suppliers can expand. That makes foundries and packaging capacity part of the AI capex thesis, not just background infrastructure. (broadcom.com) ### Is this a healthier market story? In one sense, yes. A broader trade usually means investors think demand is durable enough to lift multiple layers of the stack. But the catch is valuation. Once a theme spreads, everything starts getting labeled an AI beneficiary, and not every company will capture real pricing power. The winners will be the ones sitting at actual bottlenecks — HBM, advanced networking, custom silicon and packaging capacity. (cnbc.com) ### Bottom line The market is still betting on AI capex. But the bet is getting more specific — and more industrial. Nvidia remains the center of gravity, but investors are increasingly treating AI as a full buildout story, where memory, networking, storage and foundry capacity can all matter almost as much. (investors.micron.com)