Most Software Firms Won’t Survive the AI Squeeze, Says Fund that Crushed Peers

With input from Bloomberg and Forbes.

Nick Evans sold software stocks before the stampede — and now he’s warning bargain hunters: most names are still toxic. Evans, the manager behind a $12 billion tech fund that beat 99% of peers over one year (and 97% over five), says the payoff for selling early wasn’t luck — it was reading the threat correctly. Nick Evans

A single plugin from Anthropic wiped about $285 billion off the market in one day, and the fallout was brutal. Thomson Reuters slid 16%, while big software names like Salesforce, Adobe and HubSpot have lost between roughly 19% and 39% since January. The S&P 500 software & services gauge has been hammered, dropping more than 4% in a single session and running down about 20% year-to-date. S&P 500 Software & Services Index

Why the carnage? Because AI agents — the kind that can take over whole workflows — threaten the per-seat, subscription cash cows that many software businesses still depend on. Investors are dumping anything that looks replaceable by an autonomous agent. Big-bank research even showed the sector’s software basket suffered its worst one-day drop in months.

But Evans and others say the market is barking up the wrong tree if it only worries about which app gets replaced. The real money is in the plumbing that runs those agents — continuous inference compute, data centers, networking and custom chips — not the apps themselves.

That matters because running millions of agents nonstop burns enormous compute. The hyperscalers are racing to build that capacity: combined capital spending from the likes of Amazon, Google, Microsoft and Meta is now projected near $700 billion for 2026 — a jump of more than 60% from 2025. Throw in a planned $20 billion from Oracle and that total easily crosses the $700 billion mark.

That’s why the physical layer looks like the safer bet. Chips, data centers, high-speed networking and custom silicon are the backbone of the agent era — and demand there is predictable and huge. Big suppliers stand to benefit whether one AI model wins or another. Nvidia and Broadcom are in prime positions to grab a huge slice of the capex pie, while Arista Networks sits on a chokepoint for the GPU clusters that power inference.

Wall Street takes note: some banks say the software selloff was overdone, but not everyone is convinced that the threat is temporary. JPMorgan warned the panic may have overshot; at the same time, analysts at Deutsche Bank point to the scale of the infrastructure build-out as creating a meaningful moat for suppliers.

Bottom line: the AI shock isn’t a story about chatbots vs. CRM. It’s about who owns the compute, the chips and the cables. If you’re picking stocks, Evans’s blunt takeaway is simple — bet on the plumbing, not the potentially obsolete app on top.