For twenty years, enterprise software was the safest bet in technology: predictable recurring revenue, high margins, sticky subscriptions, wide moats. Then AI agents learned to do what knowledge workers do — and the market decided that a single agent replacing dozens of software licenses was not a hypothetical but an inevitability. Roughly one trillion dollars in software market value was repriced in weeks. The MSCI Software Index fell 21% by mid-February. “Black Tuesday” on February 3 saw a 13% single-day sector collapse. Goldman Sachs compared the outlook to newspapers in 2002. The fundamentals — double-digit revenue growth, record RPO backlogs, high retention — still look fine. The market doesn’t care. It is pricing a future where the per-seat model is dead.
The SaaS business model was built on a simple equation: charge per seat, per month, forever. The more employees a company had, the more licenses it needed. Growth was predictable. Margins were high. Switching costs were enormous. For two decades, this model produced some of the most reliable revenue streams in corporate history. ServiceNow, Salesforce, Workday, Adobe, Intuit — these were not speculative bets. They were compounding machines.[3]
Then came “seat compression.” AI agents can now perform tasks that once required teams of knowledge workers — analysing documents, generating reports, automating data processing, completing multi-step workflows. If one AI agent can replace the output of a dozen employees, companies need fewer seats, not more. The per-seat pricing model doesn’t just slow down. It structurally contracts. The transition that analysts have labelled the shift from “Software as a Service” to “Service as Software” arrived faster than anyone modelled for.[2]
The paradox at the heart of this case: the fundamentals have not yet broken. ServiceNow’s subscription revenue grew 21% year-over-year. Snowflake grew 30%. Salesforce’s remaining performance obligations hit $72.4 billion, up 14%. Retention rates are as high as ever.[5] But the market is forward-looking, and what it sees ahead is a pricing model under existential pressure. As Wells Fargo put it: the level of uncertainty is so high that 2027 estimates can’t be trusted, which means it’s hard to say software as a whole is cheap, even if some names will ultimately be winners.[3]
Meanwhile, capital is rotating at speed. Billions are flowing from software into industrials, financials, and energy. Caterpillar’s backlog hit a record $39.9 billion. JPMorgan’s market cap surged past $900 billion. Energy stocks rose 21% as the market bets on the massive electricity requirements of the AI economy.[2] The “virtual economy” is losing its lustre compared to the “tangible economy.” Software, once classified as low-risk and high-margin, is being reclassified as high-risk and prone to disruption by the very technology it sought to monetise.
Alphabet, Amazon, Meta, and Microsoft reveal plans to spend a combined $680 billion on AI infrastructure in 2026 — 70% higher than 2025 estimates. Investors ask: if AI is this expensive, who pays?[5]
D3 Capex AlarmThe sector benchmark collapses 13% in a single session following disappointing guidance updates from industry titans. Anthropic’s new AI-powered Cowork plug-ins automate legal administration. ~$1 trillion in market value is repriced. The rout goes global: Japanese IT firms fall 16%, India’s Nifty IT drops 5.8%, China’s Kingdee plunges 12%.[2][8]
D3 Black TuesdayAnthropic publishes a blog post about Claude Code modernising COBOL. IBM loses $40 billion in market cap in a single session — its worst day in 25 years. Accenture and Cognizant also fall. A blog post just cost IBM $30 billion.[9]
D3 + D6 ContagionGoldman strategist Ben Snider compares the software sector to the newspaper industry in the early 2000s, where share prices fell an average of 95% between 2002 and 2009. He warns of “long-term downside risk.”[1]
D3 Narrative ShiftOnly 2 of 30 names positive YTD: Akamai and Fortinet. Workday, Gartner, GoDaddy each down 30%+. Private credit funds receive 10.9% redemption requests as lenders worry about software borrower viability. Adobe CEO steps down after 18 years.[3][7]
D2 + D3 Deepening| Dimension | Evidence |
|---|---|
| Revenue / Financial (D3)Origin · 65 | ~$1 trillion in software market value repriced. MSCI Software Index −21% YTD by mid-February. S&P Software & Services Index −26% from October high, −18% YTD. Three converging pressure vectors: (1) business model threat from AI-driven seat compression; (2) valuation reset as stretched multiples unwind; (3) capital rotation into industrials, energy, and financials. Intuit −45%, Workday −39%, Salesforce −30%, ServiceNow −28%, IBM −27% in February alone. Private credit contagion: Morgan Stanley fund received 10.9% redemption requests (honouring only 5%).[1][2][3] |
| Employee (D2)L1 · 55 | Meta reportedly planning 20%+ workforce cuts to fund AI capex. Block cut 40% of its workforce while posting record gross profit — using the open-source Goose AI agent to restructure. Adobe CEO Shantanu Narayen stepping down after 18 years amid the disruption wave. Sector-wide hiring freezes as companies redirect human capital budgets to AI infrastructure. A Citrini Research report warned AI may “significantly displace workers” and called for an AI tax to cushion job losses.[3][6] |
| Operational (D6)L1 · 55 | Companies scrambling to pivot from per-seat to outcome-based pricing. Combined hyperscaler AI capex guidance of $680 billion for 2026 — 70% higher than 2025. Operational restructuring underway across the entire sector. IBM scrambled to publish a defence of its mainframe business within hours of Anthropic’s COBOL blog post. The operational challenge: rebuilding pricing models, infrastructure, and go-to-market strategies while revenue is under pressure simultaneously.[2][9] |
| Quality / Product (D5)L2 · 50 | 4% of all public GitHub commits now fully authored by Claude Code — a 42,896× increase in thirteen months. “Vibe coding” enables non-developers to create applications, diminishing demand for products from established software makers. SemiAnalysis projects AI-authored commits could hit 20% of all daily commits by end of 2026. The quality paradox: AI is simultaneously improving output quality for some use cases while making the products that charged for human-mediated quality obsolete.[6] |
| Customer / Ecosystem (D1)L2 · 45 | Enterprise customers adopting a “wait and see” approach, delaying purchases while evaluating whether AI agents can replace existing software tools. Client retention rates remain high — customers are not leaving, they are simply slowing their buying. Morningstar notes this distinction matters: the fear is existential but the evidence is still ambiguous. Global contagion — software selloff hit Japan, India, China, and Europe within hours of the US rout.[4][8] |
| Regulatory (D4)L2 · 25 | AI governance and training data rights emerging as policy questions. Canadian study found major AI systems show extensive knowledge of news reporting without compensation. Citrini Research calling for an AI tax to cushion job losses. No direct regulatory action on the software selloff itself, but the groundwork for AI employment regulation is being laid. The regulatory landscape is shifting from data privacy toward AI labour displacement.[3] |
-- The Per-Seat Funeral: 6D Diagnostic Cascade
FORAGE software_sector_repricing
WHERE market_cap_wiped > 500_000_000_000
AND seat_compression_observed = true
AND pricing_model_threat = "per-seat obsolescence"
AND capital_rotation_to_tangibles = true
AND fundamentals_still_strong = true
ACROSS D3, D2, D6, D5, D1, D4
DEPTH 3
SURFACE per_seat_funeral_cascade
DIVE INTO pricing_model_disruption
WHEN ai_agents_replace_seats AND saas_to_service_as_software
TRACE seat_compression_cascade
EMIT sector_repricing_signal
DRIFT per_seat_funeral_cascade
METHODOLOGY 85 -- sophisticated SaaS cos: recurring revenue, deep moats, enterprise relationships
PERFORMANCE 35 -- ~$1T repriced, unable to demonstrate AI-resilience, 2027 estimates shrinking
FETCH per_seat_funeral_cascade
THRESHOLD 1000
ON EXECUTE CHIRP diagnostic "~$1T repriced. Per-seat model structurally threatened by AI agent seat compression. Goldman compares to newspaper decline 2002-2009. Fundamentals paradoxically strong but market forward-pricing disruption."
SURFACE analysis AS jsonRuntime: @stratiqx/cal-runtime · Spec: cal.cormorantforaging.dev · DOI: 10.5281/zenodo.18905193
ServiceNow grew subscription revenue 21%. Snowflake grew 30%. Salesforce’s RPO hit $72.4 billion. Retention rates are at all-time highs. By every trailing metric, the sector is healthy. But the market is forward-pricing a disruption that hasn’t arrived in the income statements yet. The 2027 estimates are shrinking while 2026 estimates improve. This is exactly the pattern that preceded the newspaper collapse: the revenue looks fine — right up until it doesn’t.
On February 23, Anthropic published a blog post about Claude Code modernising COBOL. IBM lost 13% of its market value — the worst single day since October 2000, and its worst month since 1968. The blog post described capabilities that IBM itself had been shipping since 2023. The selloff was driven by narrative, not novelty. In a market already primed for AI disruption fear, the messenger mattered more than the message.[9]
The per-seat model assumed that enterprise headcount would grow, and each head would need licenses. AI agents invert both assumptions. If one agent replaces the output of ten employees, ten seats become one. If the agent works through MCP connectors rather than traditional APIs, it doesn’t need seats at all — it needs access. The shift from charging per human to charging per outcome is the most consequential pricing model disruption since SaaS itself replaced perpetual licences.[2]
Goldman Sachs strategist Ben Snider compared the software sector to newspapers in the early 2000s, where share prices fell an average of 95% between 2002 and 2009. The parallel is uncomfortable: newspapers also had sticky subscriptions, recurring revenue, and wide moats. The internet didn’t kill newspapers overnight. It took seven years of declining classified revenue, ad revenue migration to Google, and readership migration to free alternatives. The lesson: if disruption is structural, the bottom is not where you think it is.[1]
One conversation. We’ll tell you if the six-dimensional view adds something new — or confirm your current tools have it covered.