A single open-source release from Anthropic has triggered the worst rout in software stocks since 2008. But the real story is not about destruction – it is about what comes next.
Jefferies trader Jeffrey Favuzza needed only four words to capture the mood on 4 February 2026: “Get me out.” By the closing bell, approximately $285 billion in market capitalisation had been erased from global software stocks in a single session – the worst day for the sector since the financial crisis. The catalyst was not a recession, a regulatory shock, or a failed earnings call. It was a GitHub repository containing eleven plugins.
On 30 January, Anthropic released a set of open-source extensions for Claude Cowork, its desktop AI agent platform. Each plugin, covering legal work, finance, sales, marketing, customer support and more, demonstrated that an AI agent could autonomously execute specialised workflows previously locked behind expensive SaaS subscriptions. Within 72 hours, the market delivered its verdict with brutal efficiency.
Goldman Sachs’ US software basket fell 6 per cent. JPMorgan’s software index dropped 7 per cent in a day and 18 per cent year-to-date. The iShares Expanded Tech-Software ETF slid more than 20 per cent for the year. Software underperformed semiconductors by 20 percentage points over 20 trading days – a divergence not seen since the dot-com bubble peaked in February 2000, according to BTIG.
The carnage was indiscriminate. Thomson Reuters, whose legal data empire faced a direct challenge from the legal plugin, fell as much as 22 per cent. RELX, parent of LexisNexis, suffered its worst day since 1988, plunging 14 per cent and now sitting half its February 2025 peak. LegalZoom dropped nearly 20 per cent. Figma has lost 85 per cent from its IPO peak. HubSpot is down 39 per cent for the year. Even SAP has shed more than a third of its value from year-ago highs.
Meanwhile, Nvidia and TSMC gained 2.1 per cent on the same day – a stark illustration of where capital sees the future.
The Body Shop Model Breaks Down
Nowhere was the shockwave felt more acutely than in India, where the Nifty IT index recorded its sharpest single-day decline since March 2020, erasing approximately $23 billion in value. Tata Consultancy Services has cut 30,000 jobs in six months and announced 12,000 more. Industry analysts warn up to 500,000 Indian IT jobs are at risk.
The threat is structural. More than a third of Indian IT companies now use AI for 40 per cent of core operations. Yet fewer than one in five IT workers possess AI-relevant skills. Nasscom estimates India needs one million AI professionals by 2026 – a gap that retraining alone cannot close. Tech Mahindra’s CEO Mohit Joshi called the selloff “a market overreaction,” insisting firms are pivoting to outcome-based delivery. But the market is pricing in a future where billing by the headcount hour faces existential compression.
The Per-Seat Apocalypse
At the philosophical centre of the rout sits a simple arithmetic problem. SaaStr founder Jason Lemkin put it with characteristic bluntness: “If 10 AI agents can do the work of 100 sales reps, you don’t need 100 Salesforce seats anymore – you need 10.” That is a 90 per cent reduction in seat revenue for the same work output.
The industry is already experimenting with alternatives. Intercom charges $0.99 per successful AI resolution. Zendesk asks $1.50 to $2.00. Salesforce’s Agentforce bills $2 per conversation. Gartner forecasts that 40 per cent of enterprise SaaS will include outcome-based pricing by year’s end. The shift from per-seat to per-outcome is not a tweak – it is a wholesale reconstruction of the industry’s revenue architecture.
The Contagion Spreads
The damage extends beyond public equities. From 2015 to 2025, private equity acquired more than 1,900 software companies for over $440 billion. UBS warns that in an aggressive AI disruption scenario, private credit defaults could reach 13 per cent – nearly double the rate for leveraged loans. The bank estimates 25 to 35 per cent of the $1.7 trillion private credit market is exposed to AI disruption risk. For the advertising sector, the collateral damage was swift: WPP, Publicis and Omnicom fell 9 to 12 per cent as the marketing plugin raised the spectre of automated campaign management.
The Super Individual Rises
Yet amid the wreckage, a different story is taking shape. Anthropic CEO Dario Amodei predicted a single individual may operate a billion-dollar company by 2026. Already, solo founder Maor Shlomo sold his six-month-old AI startup Base44 to Wix for $80 million in cash. Boris Cherny reportedly built Cowork itself in 10 days using Claude Code – a recursive proof of concept if ever there was one. The “vibe coder” – a non-programmer building functional applications through natural language prompting – was named word of the year in 2025.
Not Dead Yet
The counterarguments deserve a hearing. SaaS represents more than 10 per cent of total IT spending, a market exceeding $300 billion and projected to approach a trillion. The legal data companies that suffered most, Thomson Reuters, RELX, Wolters Kluwer, are what Artificial Lawyer calls “data fortresses” with irreplaceable moats. Enterprise stickiness, compliance requirements and integration complexity remain formidable barriers. The plugins themselves require enterprise licences and technical expertise to deploy meaningfully.
Bain & Company counsels that technological revolutions are rarely binary: transitions create heterogeneous ecosystems, not wholesale replacement. BTIG’s Jonathan Krinsky acknowledges software is “probably oversold enough for a bounce” but warns “it is going to take a long time to repair and build a new base.”
Perhaps the most measured framing comes from Lemkin himself: “The 2026 crash isn’t AI killing SaaS. It’s the market finally pricing in the deceleration that started in 2021. This isn’t the death of SaaS – it’s the end of easy SaaS.”
A security researcher might add a footnote. Within 48 hours of Cowork’s launch, PromptArmor documented a critical file exfiltration vulnerability – prompt injection hidden in a Word document using one-point font, white on white. The fundamental contradiction remains unresolved: users are told to avoid granting access to sensitive files while being encouraged to let AI organise their entire desktop.
The SaaSpocalypse is not a single event but a convergence: a genuine leap in AI capability, a market belatedly pricing in three years of softening growth, and panic contagion rippling into private credit. The per-seat model is cracking. But the rubble is not yet cleared, and from it, something new is being built – one plugin, one agent, one super individual at a time.
Read our full Report Disclaimer.
Report Disclaimer
This report is provided for informational purposes only and does not constitute financial, legal, or investment advice. The views expressed are those of Bretalon Ltd and are based on information believed to be reliable at the time of publication. Past performance is not indicative of future results. Recipients should conduct their own due diligence before making any decisions based on this material. For full terms, see our Report Disclaimer.