AI pause sparks software surge as investors flee chips

The origin of the pause idea traces back to March 14, 2023, when OpenAI released GPT-4, “the successor to GPT-3.5,” in what was presented as a turning point for the field. That launch “mark[ed] a new capability threshold that triggered calls for a pause in training more powerful AI models,” a notable break from the period just before it, when AI laboratories routinely trained increasingly powerful systems without industry-wide safety assessments or public agreements on limits, reflecting a norm of unrestricted capability scaling rather than any shared willingness to slow down.

Sam Altman’s later public framing made clear how far the conversation had moved from that earlier ethos of relentless scaling toward a safety-first posture. In his September 2026 interview with Fortune, Altman said “no gamble with humanity is OK,” adding that he would stand up to investors if he needed to “pause or stop AI development altogether,” while also describing a need “to slow down while safety and alignment solutions can catch up to model capabilities,” language that directly recast frontier progress as something to be constrained when risk outpaces safeguards.

The rotation hit hardware first because investors were already questioning whether the AI buildout justified the risk and valuation embedded in chip names. As Semiconductor Stocks Plunge as Capital Shifts to Big Tech Hyperscalers put it, “semiconductor stocks are aggressively losing momentum… The SOX Index is now down 16% from its June highs,” alongside “heavy institutional distribution in peripheral chip names… capital seeks safety in the massive cash flows,” showing that money was leaving the most cyclical, expectation-heavy layer of the stack before it looked for a new home.

That capital then favored software not because investors suddenly believed in a broad application-software recovery, but because they wanted steadier models and narrower AI execution risk. UBS analyst Karl Kiersted said the move was “primarily a function of a fade in the semis trade and a broader portfolio rotation into defensive cheaper stocks,” even as he cited “pretty strong evidence that rising AI spend inside Fortune 500 enterprises is beginning to crowd out spend,” specifically that it “crowd[s] out traditional software spend” and “crowd[s] out IT services spend,” making recurring-revenue software with incremental AI integration look safer than frontier infrastructure bets.

The immediate spark for the September move was not a gradual valuation rethink but a specific mid-month news event: public calls from OpenAI and Anthropic leaders for an AI development slowdown. StockStory, citing Reuters on September 14, directly linked that mid-September 2026 message to the rotation, saying those pause appeals triggered sell-offs in semiconductor and AI-hardware names while “software stocks bucked the broader tech trend and climbed higher in early trading,” tying the frontier-AI slowdown call to a same-session move into enterprise software and SaaS rather than to any slower-building macro or earnings catalyst.

That catalyst showed up immediately and broadly across software screens the same day, reinforcing that the pause call was the trigger rather than a post-hoc explanation. StockStory reported that “a number of stocks jumped in the afternoon session” after enterprise software and SaaS shares rallied broadly, with Cloudflare up 6.3%, Paylocity 3.6%, and GitLab 4.2%; elsewhere, WIX rose 10%, BlackLine 5.5%, Amplitude 7.5%, Toast 4.6%, Sprinklr 5%, Autodesk 7.5%, C3.ai 6%, GoDaddy 5.9%, Rapid7 19.3%, and Sprout Social 5.8%.

The bid went first to enterprise software because investors decided those companies offered a clearer, steadier way to monetize AI than the costlier infrastructure layer. Jim Cramer said sentiment turned after “ServiceNow's latest earnings report on July 22,” and that since then “software companies like ServiceNow and Salesforce have climbed roughly 22% and 15%, respectively,” a reversal that followed months when SaaS names were “under pressure as investors feared AI models would disrupt their software-as-a-service business models” before the market concluded recurring software demand was more durable than feared.

That preference for steadier AI beneficiaries also helps explain interest in names like Commvault, where AI features were being layered onto an existing subscription engine rather than requiring investors to underwrite another expensive buildout cycle. Yahoo Finance noted “Rapid expansion and cross sell momentum within Commvault Systems SaaS platform is now reinforced by subscription ARR of US$1.05b, SaaS ARR of US$424m growing at 38% year over year, and almost half of SaaS customers adopting multiple products,” even as “The stock trades on a P/E of 91.2x, compared with 29.8x for the US Software industry, 25.8x across peers, and a fair ratio of 37x.”

The move was not confined to a single ticker; it was broad enough to re-rate software across categories. CNBC’s Jim Cramer said enterprise software was “one of the clearest examples” of the rebound, adding that ServiceNow’s post-earnings rally “caused an avalanche” that helped ignite a wider sector advance, while Commvault offered a concrete measure of the scale: “Commvault's stock surged 61.8% in six months to $141.82,” a gain large enough to mark it as part of a meaningful software upswing rather than an isolated bounce.

Commvault also shows how powerful the rotation was at the stock level, because the gains persisted even without spotless operating signals. StockStory noted “weak Q2 billings of $300.3M” and “only 12.2% YoY growth,” describing “soft demand amid rising competition,” yet AlphaStreet News still ran the headline “Commvault Systems Jumps 6.9% Amid Sector-Wide Rally,” underscoring that investors were lifting a broad set of software names with mid-single-digit daily moves and, in some cases, much larger multi-month advances across the group.