Wall Street prices software apocalypse after AI-driven sell-off

Wall Street briefly rediscovered that software is made of software.

After a sharp sell-off that erased more than $1 trillion in Big Tech valuations, investors started muttering about a “software apocalypse,” Business Insider reports, blaming fresh AI capabilities—specifically Anthropic’s newly released industry-specific plug-in—for undermining confidence in established software names.

The public debate quickly split into two camps. One side treats generative AI as a universal solvent: why pay recurring SaaS rents when an LLM can “just build it”? The other side, led by people who have actually shipped large systems, points out the obvious: AI doesn’t eliminate software; it consumes it.

Nvidia CEO Jensen Huang called the idea that software tools are “in decline” because AI will replace them “the most illogical thing in the world,” according to Business Insider. He argued that software is the tool AI uses rather than the thing it destroys, and he name-checked ServiceNow, SAP, Cadence, and Synopsys as companies he sees as bright spots.

OpenAI CEO Sam Altman offered a more market-aware version of the same point: software isn’t dead, but the way it’s created and used will change, and volatility is likely as investors try to price that transition, Business Insider writes.

Figma CEO Dylan Field, whose own stock has been punished since its 2025 IPO, framed the moment as a Darwinian filter: it’s not enough to ship something that “works”; the winners will be those who build the right thing and integrate AI without turning product quality into an afterthought. Figma is simultaneously hedging and embracing the shift, announcing a partnership with Anthropic on tooling that converts AI-generated code into editable designs.

Zoho founder Sridhar Vembu delivered the most interesting critique because it targets incentives rather than models. SaaS, he wrote, was “ripe for consolidation” long before LLMs, and an industry spending “vastly more on sales and marketing than on engineering” was always vulnerable; AI is merely the pin that pops an inflated business model, per Business Insider.

The concrete technical risk is not that AI replaces software, but that the software stack is already a brittle dependency graph held together by optimism and CI pipelines.

Modern products are assembled from transitive dependencies, container images, build scripts, and third-party APIs. Each layer expands the attack surface: compromised package registries, poisoned updates, malicious pull requests, and CI/CD credential theft. Single-maintainer projects can become single points of failure. “Move fast” becomes “ship unsigned artifacts.”

AI accelerates this dynamic by increasing code churn and lowering the cost of adding yet another abstraction layer. It can also accelerate discovery and exploitation of vulnerabilities—while simultaneously generating plausible-but-wrong patches that pass superficial review.

The market, of course, will respond as it always does: by pricing narratives first and engineering constraints later. Governments will promise “AI safety” frameworks; enterprises will buy yet another platform; and the dependency tree will keep growing—because no one gets promoted for deleting code.

If there is an apocalypse here, it’s not software’s death. It’s software’s continued survival as an ever denser web of outsourced trust.