As generative AI rattles public markets, venture capitalist Bill Gurley says investors confronting what some are calling a “SaaSpocalypse” should distinguish between structural impairment and cyclical fear — and pay closer attention to how AI infrastructure deals are being financed.

Software-as-a-Service stocks have slid in early 2026 as investors reassess competitive durability in light of rapid advances in generative AI. New development tools, including app-building enhancements tied to Anthropic’s Claude platform, have intensified concerns that AI-native systems could reduce reliance on traditional enterprise vendors such as Salesforce, Atlassian, and DocuSign.

Appearing on Squawk Box on CNBC, Gurley, a general partner at Benchmark, acknowledged the anxiety but framed it within historical precedent. He compared the current moment to the aftermath of Facebook’s IPO, when fears about the mobile transition sent shares sharply lower before the company proved it could adapt.

“I’ve never seen a disruption that had this much anxiety and go across so many companies,” Gurley said, underscoring the breadth of the selloff.

Is AI a feature or a substitute?

At the core of the debate is whether generative AI acts as a complementary productivity layer or a direct substitute for SaaS platforms.

Bearish investors argue that AI coding agents and workflow automation tools could allow enterprises to generate bespoke internal systems, reducing demand for subscription software. If AI systems can draft contracts, manage pipelines, reconcile documents, and orchestrate workflows autonomously, the marginal value of certain SaaS features could compress.

Yet Gurley pointed to a countervailing signal: AI-native firms themselves continue to rely on legacy enterprise systems. He cited Anthropic’s use of platforms such as Workday and Salesforce, suggesting that mission-critical systems of record — HR, CRM, compliance — remain deeply embedded in operational infrastructure.

That dynamic highlights switching costs and integration depth as key variables. Enterprise SaaS platforms are often tied into billing systems, regulatory reporting, identity management, and third-party integrations. Even if AI accelerates customization, ripping out core infrastructure can be operationally and legally complex.

Valuation compression and capital cycles

The SaaS sector entered 2026 with lingering valuation sensitivity after the post-2021 reset in growth equities. Many cloud software companies had already transitioned from growth-at-any-cost models to margin expansion and free cash flow generation. The new AI wave introduces a second-order risk: capital reallocation.

Institutional investors appear to be rotating toward AI infrastructure — semiconductors, data centers, and foundational model providers — at the expense of application-layer software. If generative AI captures a disproportionate share of incremental enterprise IT budgets, SaaS revenue growth could decelerate.

Gurley’s message to investors who retain conviction is rooted in capital markets psychology. Echoing the philosophy of Warren Buffett, he argued that panic-driven price declines can create asymmetry for long-term buyers.

“You shouldn’t be blogging about what’s wrong with the prices,” Gurley said. “You should be quiet and picking them up off the floor.”

Implicit in that advice is a need for discrimination. Not all SaaS companies will integrate AI successfully. Firms that treat AI as a superficial add-on may struggle, while those that embed it deeply into workflow automation, predictive analytics, and customer engagement could defend or expand their moats.

Where Gurley expressed sharper concern was in the financial architecture underpinning AI expansion.

He described what he sees as circularity in transactions between AI model developers and infrastructure providers. Early agreements between Microsoft and OpenAI, for example, involved cloud credits and revenue flows that fed back into Microsoft’s Azure business.

A more recent example involves Meta and Advanced Micro Devices. The companies announced a deal under which Meta would purchase six gigawatts of computing capacity, with a structure that could result in Meta owning up to 10% of AMD’s stock.

Such arrangements raise questions about how revenue is recognized, how demand is signaled to markets, and whether cross-ownership could amplify volatility in a downturn. Gurley said that when he described similar deal structures to ChatGPT without naming the parties, the model generated references to past accounting scandals, including Enron and WorldCom. He did not accuse companies of misconduct but warned that, if growth assumptions falter, investors may revisit these structures critically.

He also suggested regulators are unlikely to intervene preemptively, arguing that scrutiny often intensifies after market stress exposes weaknesses.

Systemic implications

The intertwined nature of AI funding, infrastructure buildout, and equity stakes introduces systemic risk considerations. Massive capital expenditures on GPUs, custom silicon, and data centers assume sustained demand for AI workloads. If enterprise adoption slows or model efficiency reduces compute intensity faster than expected, infrastructure providers could face overcapacity.

Conversely, if AI-driven productivity gains materialize at scale, enterprise IT budgets may expand, benefiting both infrastructure and application-layer vendors. The key uncertainty is elasticity: whether AI increases total software spending or redistributes it.

On workforce impact, Gurley diverged from more alarmist narratives. He described AI as “jet fuel” for motivated individuals, echoing comments made publicly by entrepreneur Mark Cuban. He argued that generative tools dramatically compress learning curves, enabling professionals to acquire technical and domain expertise faster than at any prior point.

“You can learn faster than you could have ever learned at any point in history right now,” he said.

That perspective frames AI less as a labor substitute and more as a force multiplier — particularly for those pursuing differentiated or entrepreneurial career paths.

The “SaaSpocalypse” narrative captures a market recalibration rather than a settled outcome. Generative AI may compress margins in commoditized software categories while entrenching leaders that successfully integrate automation and intelligence into core workflows.

The challenge is analytical rather than emotional for investors as it involves evaluating unit economics under AI integration, assessing balance sheet resilience amid valuation compression, and scrutinizing infrastructure deal structures that could magnify future drawdowns.

Gurley’s broader message is that market-wide fear does not automatically equate to permanent impairment. But neither does technological enthusiasm negate financial discipline. The next phase of the AI cycle is expected to lie less on model demos and more on durable revenue quality, capital allocation, and transparent accounting.