Investors are asking whether the GPU party keeps running or if a quieter rotation is starting toward inference chips, cloud AI services, and software monetization.
Illustration by IMF Alpha editorial · Reviewed by Pedro Marini
Short version
Nvidia re-priced the market. That doesn’t mean every future AI dollar will live in one stock. We’re moving into a phase where money chases specialized inference silicon, cloud providers that turn compute into steady revenue, and AI-native software — not just raw GPU horsepower.
Nvidia owned the early narrative. Predictable story: dominance attracts alternatives, regulators, and tougher valuation questions. Over the next 12–24 months smart money will concentrate selectively — favoring business models that turn compute into recurring revenue rather than blind hardware bets.
GPU incumbents remain central for top-end training, but margin pressure and competition open doors for AMD and custom accelerators. Intel and Qualcomm are quietly refocusing on inference and edge AI to claw back data-center relevance. Meanwhile the big cloud players aim to convert raw compute into subscription-style offerings — that predictable revenue is the lever that moves long-term multiples.
Valuation is a sentiment thermometer, not the same thing as a business model. A 10x revenue multiple on recurring cloud AI income is not equivalent to 10x on cyclical chip sales.
Think of Nvidia like the steam engine: indispensable, obvious. But someone has to build the rails, the towns, and the factories where value actually gets realized. Those rails are software, cloud platforms, and inference silicon.
Pure hardware bets without services are brittle. It’s like buying crude oil futures and ignoring refineries and distribution. Integration and recurring service revenue matter a lot.
Nvidia rewrote expectations, and that creates both richer opportunities and sharper risks. For long-term investors, a sensible posture is selective diversification: favor companies that reliably turn AI compute into recurring revenue, look for durable moats in software and services, and treat pure-play hardware positions as tactical, not foundational.
Pedro Marini
From clean rooms to simulated customers, financial firms are racing to create usable datasets for generative AI while dodging privacy pitfalls
Smartphones and PCs are starting to run generative models locally. That shifts power to chipmakers, changes app economics, and gives privacy a new marketing lifeline.
From privacy-by-default budgeting to instant fraud checks, on-device generative models are reshaping fintech. Here’s what consumers, banks and investors should watch next.