Investors are looking beyond the GPU king. Chips, cloud platforms and software plays offer different risk-return tradeoffs — here’s a compact map of who could win next.
Illustration by IMF Alpha editorial · Reviewed by Pedro Marini
The short version: Nvidia still stands as the obvious poster child for AI investing, but rich valuations and concentration risk are nudging capital into second-tier chipmakers, networking ASICs and cloud-native software. This isn’t a repudiation of GPUs — more a search for broader exposure across the AI stack.
Why this matters now
GPUs are central to generative AI; that part is obvious. Markets, though, look for the next lever that can amplify returns without requiring the same frothy multiples. That search creates both opportunity and traps. Treating AI as a single bet on one name risks missing smaller, faster-moving winners tied to data-center spend, cloud monetization, and enterprise AI adoption.
Where money is flowing — and why
AI-adjacent GPU plays (AMD, Intel): These firms gain if GPUs shift from scarcity to volume. AMD has closed much of the technical gap in data centers; Intel is trying to stitch CPUs, x86 tweaks and specialized accelerators into a coherent stack. Both carry execution risk and cyclicality.
Networking and infrastructure ASICs (Marvell, Broadcom): AI isn’t only compute. Moving huge datasets matters a lot. Chips that speed networking, switching and storage can enjoy steady demand as operators push for better throughput and lower latency.
Cloud and software platforms (Microsoft, Amazon, Meta): For many investors this is the lower-friction route. These platforms sell compute, tooling and monetization channels — turning capex into recurring revenue and often enjoying deeper data advantages.
Vertical and niche AI software: Startups and smaller public firms building domain-specific tools can post higher gross margins and faster sales cycles. The tradeoff: customer concentration and a long, uneven road to scale.
A few practical signals to watch
Risks that can derail the trade
This isn’t one-way traffic. Execution problems (chip yields, slower software uptake), macro shocks that freeze capex, or regulatory pushback on models and data use can all compress multiples fast. And if one vendor figures out a major efficiency breakthrough, value could reconcentrate quickly.
A practical allocation framework
Final take
This rotation is less about abandoning Nvidia and more about hedging a narrative that’s become crowded. Think of AI as layered: raw compute, data fabrics, application software. Each layer moves on different timeframes, has distinct margin dynamics and faces different regulatory pressures. With headline valuations stretched, some of the next returns may come from the plumbing that lets AI scale — not only the chips that run the models.
Note: ticker examples are illustrative, not trading advice. Do your own research or consult a professional before making investment decisions.
How cloud giants, startups and synthetic-data vendors are packaging, selling and protecting the raw material powering generative AI — and what it means for investors.
Regulatory risk, licensing fights and mounting privacy pressure are pushing U.S. companies to buy and build synthetic datasets — and investors are paying attention.
Tiny LLMs, phone NPUs and smarter chips are turning smartphones into private AI assistants. Here’s what that means for privacy, apps and investors.