After a GPU-fueled surge, investors are rotating into enterprise AI software and platform plays. What that means for valuations, risk, and your watch list.
Illustration by IMF Alpha editorial · Reviewed by Pedro Marini
Nvidia changed what a chip company can look like: enormous margins, tight customer lock-in, and pricing power that behaves more like a subscription business than old-school silicon sales. That helped turn AI into a household investing theme. But the next wave of returns will probably come from software, data, and recurring revenue — the things hardware sellers risk missing if they stop at chips.
Money is moving away from pure-play silicon toward firms that actually monetize AI for customers. Think:
This isn’t an either-or. Big cloud providers and chip makers still matter. But expect relative valuation pressure on hardware if investors start pricing recurring revenue more highly.
Pay attention to these metrics rather than the next PR cycle:
One last, slightly counterintuitive point
Nvidia’s win is as much about networks and customer lock-in as it is about GPUs. So the long-term winners may not be only chip designers but the companies that stitch compute into everyday business processes and make it billable, repeatable revenue. The smarter trade over the next five years isn’t chip versus software. It’s who turns compute into measurable, repeatable cash flow.
From data co-ops to synthetic markets, American firms are treating training sets like strategic assets — and investors are paying attention.
Startups and incumbents rush to replace risky customer datasets with synthetic alternatives, promising privacy, scale and cost savings — but trade-offs are real.
From privacy-first assistants to faster replies offline — why manufacturers, chipmakers and app developers are racing to squeeze LLMs into pockets, and what it means for users and markets.