Hedge funds have turned decisively against enterprise software. Data from Goldman Sachs shows they slashed their long positions in the sector to the lowest level since 2019 while pouring money into semiconductor names instead. The shift comes as fears grow that agentic AI will upend the subscription model that powered two decades of growth.
Software and services stocks sit down 14 percent year to date and have shed 9 percent over the past 12 months. Contrast that with semiconductors and semiconductor equipment. Those groups stand up 38 percent this year and have more than doubled in the last year. The performance gap tells only part of the story.
Goldman Sachs examined $9 trillion in equity holdings at the start of the second quarter. Its U.S. Weekly Kickstart report laid bare the repositioning. Hedge funds and mutual funds both rotated away from software toward semis. Mutual funds now carry their widest underweight in software, excluding Microsoft, since 2012.
They added on net to Lam Research, Applied Materials and ASML. Mutual funds favored Intel and SiTime. Even Microsoft saw net cuts from both groups last quarter. CNBC captured the blunt summary: hedge funds are doubling down on AI but dumping software stocks.
But the exit runs deeper than quarterly flows. Short interest has delivered real profits. Hedge funds pocketed an estimated $24 billion betting against software names so far in 2026. The sector’s average forward earnings multiple collapsed from roughly 39 times to 21 times in a matter of months. Roughly $2 trillion in market capitalization disappeared from software companies over the past year.
Names once considered invincible took painful hits. Salesforce lost about 30 percent. Workday fell 33 percent. The S&P Software & Services Select Industry Index dropped more than 20 percent in the early months of the year. Investors began to question whether the traditional per-seat subscription could survive when AI agents handle workflows directly.
The Fortune article that first highlighted Goldman’s latest numbers framed the move as deliberate rather than panicked. Hedge fund net leverage sits at the 85th percentile of the last five years. Managers aren’t de-risking overall. They are making a concentrated bet that value in AI will accrue to the infrastructure layer, not the application layer.
Concerns center on more than just efficiency gains. AI agents promise to replace interfaces entirely. Users once clicked through multiple SaaS dashboards. Now a single super-agent could coordinate tasks across systems, reducing demand for individual tools. Pricing may shift from seats to usage or outcomes. That change threatens the predictable recurring revenue investors prized.
Enterprise customers already signal harder bargaining. CIOs and CTOs renegotiate contracts and demand proof of AI capabilities. Some threaten to replace vendors that fail to integrate agentic features. Thomson Reuters CTO Joel Hron drew a clear line. The divide falls between companies with proprietary, domain-specific data that can differentiate AI models and those that function mainly as interface wrappers.
ServiceNow executives declared the sidecar AI era finished. At its Knowledge 2026 conference, president and COO Amit Zavery told Fortune the company focuses on outcomes delivered through its Context Engine. Built on 100 billion workflows and 7 trillion annual transactions, the system aims to give agents reliable context and governance inside enterprises. “Enterprise software was never sexy,” Zavery said. “The amount of time people building software in this space spend — not just building features, but making it secured, compliant, guaranteed performance … all those things are never sexy jobs.”
Yet the bear case faces counterarguments. Goldman Sachs itself projects the global application software market will reach $780 billion by 2030. Agentic AI should expand the overall pie even as it changes how value is captured. Historical platform shifts often rewarded incumbents with strong customer relationships and data advantages. JPMorgan analysts have pointed to long-term contracts and switching costs that will not vanish overnight.
Adaptation matters. Companies that bolt on AI features, reprice around usage, and protect their data moats may endure. Those that treat AI as an add-on risk commoditization. Some observers argue the selloff overshot. Recent X discussions reflect divided sentiment. Dropbox CEO Drew Houston noted he has yet to meet a customer canceling subscriptions because of heavy ChatGPT use.
The pain extends beyond public equity. Private credit markets hold an estimated $600 billion to $750 billion in software exposure. Many loans originated during the low-rate boom when valuations justified heavy leverage. Now secondary markets show stress. Roughly $25 billion in software loans trade below 80 cents on the dollar. Distressed tech debt reached $46.9 billion.
Take Navan. The travel platform IPO’d in late 2025 with $657 million in debt against $223 million in cash. A $400 million credit facility led by Goldman Sachs formed a big part of that burden. Revenue grew. Losses narrowed on an operating basis. Yet interest expense drove a $100 million net loss in the first half of 2025. The stock fell roughly 60 percent from its debut to around $10. A law firm opened an investigation.
Figma offers another cautionary tale. It IPO’d at $33. Shares now hover near $24, down 80 percent from peak private valuations despite 40 percent revenue growth. Such examples fuel worries that private credit could face larger losses than public equity if defaults rise. UBS analysts flagged a potential 13 percent default rate in some scenarios. Portfolio managers at firms like Apollo reportedly halved software exposure.
Still, opportunity exists in the dislocation. Blackstone, Blue Owl, Goldman Sachs, HPS and others extended a $1.4 billion loan to help Hg Capital acquire OneStream at a $6.4 billion valuation. Carlyle and BlackRock reportedly bought discounted software debt. Wider spreads may reward patient lenders. Some bulls, including Thoma Bravo, see disruption below 50 percent and view the reset as temporary.
VCs feel the pinch differently. Firms with large stakes in listed technology companies watched those holdings turn less profitable. Sequoia Capital’s position in Klarna suffered repeated share price collapses. The broader venture industry faces a world where exits prove harder at prior multiples. Record SaaS M&A in 2025 offered some liquidity, yet many private companies sit in valuation limbo. Founders and investors hesitate to sell below entry prices, creating zombie-like entities that limp forward.
Goldman Sachs research adds nuance. The bank found no broad link between AI adoption and productivity except in customer support and software development — precisely the areas many SaaS tools target. In another note, it suggested FOMO rather than pure economics partly fuels the AI infrastructure boom. Hyperscalers appear willing to prioritize participation in the arms race over near-term shareholder returns. That narrative could eventually face its own test.
The rotation looks clean on the surface. Semis win. Software loses. But outcomes will depend on execution. Incumbents with deep data and compliance infrastructure may integrate agents successfully and maintain pricing power. New entrants could erode moats if vibe coding and low-cost development lower barriers dramatically. One prediction circulating on X holds that the future belongs to one super-agent per company rather than one per human, living inside tools like Slack or Codex.
Earnings forecasts reveal skepticism. Goldman expects information technology earnings to grow 31 percent in 2026. Its top-down sector EPS estimate of $92 falls well below the $106 bottom-up consensus. Such gaps highlight how models bake in caution even as some executives push forward with AI-native strategies.
Public market reaction mixed in recent weeks. Software shares showed signs of stabilization in May as some argued the selloff went too far. Yet hedge fund positioning data suggests the conviction trade against legacy SaaS remains intact. Shorts continue to press their advantage. Long-only managers trim rather than add.
No one expects software to disappear. Enterprises will always need systems of record, governance, security and integration. The question centers on economics. Will the profit pool stay with traditional vendors or migrate toward AI platforms and agents that capture usage-based fees? Goldman projects more than 60 percent of software economics could flow through agentic systems by 2030.
ServiceNow’s bet on its Context Engine illustrates one path. Others experiment with outcome-based pricing or vertical specialization that embeds domain expertise AI struggles to replicate quickly. Vertical SaaS categories may fare better than horizontal ones. Companies with sticky workflows and proprietary data enjoy stronger defenses.
Private markets face a slower reckoning. Debt maturities loom. Covenant breaches could force restructurings. Redemption pressures at private credit funds might lead to markdowns or gates. Contagion remains a risk if software weakness spreads to broader leveraged lending. Yet the same discounted prices that worry sellers attract opportunistic buyers.
History offers perspective. Previous technology shifts created winners among those who adapted. The cloud transition rewarded certain incumbents. Mobile disrupted but also expanded markets. AI may follow a similar pattern even if the speed feels different this time. Short-term pain for software stocks could give way to longer-term opportunities for those that solve the hard problems of security, compliance and integration that Zavery described as unsexy but essential.
Investors watch the next round of earnings closely. Guidance on AI monetization, retention metrics and pricing experiments will matter more than ever. Hedge funds have placed their bets. The market now tests whether those convictions prove correct or if the rotation reverses once reality sets in. So far the data points to continued pressure on traditional SaaS valuations. Adaptation, not denial, will decide who survives the transition.
WebProNews is an iEntry Publication