The Signal
Capital has removed more than two trillion dollars of market value from the software sector since late January, and the selling is concentrated where revenue is priced per seat. The iShares Expanded Tech-Software ETF (IGV) is down 30 percent from its September 19, 2025 peak while XLK and QQQ are roughly flat over the same window, a divergence JP Morgan's analysts call the largest non-recessionary 12-month software drawdown in more than 30 years. This is not a technology selloff. It is the market concluding that the human seat is no longer the unit of software value.
Why It Matters
The repricing began in earnest on February 3, when roughly 285 billion dollars left SaaS names in a 48-hour window after a cluster of agentic product launches made the math unavoidable: if an AI agent performs the work of ten licensed users, the vendor selling ten seats is structurally overvalued. AI-native enterprise spending is growing 94 percent year over year while traditional SaaS growth has cooled to 8 percent. The capital allocators got there before most operating budgets did.
The vendor side confirms the read. Salesforce's Agentforce reached 800 million dollars in ARR by Q4 of fiscal 2026 running consumption pricing, outcome pricing, and flat-fee agentic enterprise license agreements side by side. ServiceNow beat Q1 2026 guidance with subscription revenue of 3.67 billion dollars, up 22 percent, and its Now Assist customers above one million dollars in annual contract value grew 130 percent year over year. IDC forecasts that pure seat-based pricing will be effectively obsolete by 2028, with 70 percent of vendors refactoring around consumption, outcomes, or organizational capability. Gartner puts more than 40 percent of enterprise SaaS spend on usage, agent, or outcome pricing by 2030, with the seat-based share of vendor revenue falling from 21 percent to 15 percent.
The buyer side is where the unexamined exposure sits, and the buyer is who this brief is for. Most enterprises above 100 million dollars in revenue hold multi-year agreements written when headcount was a reliable proxy for software value. Those contracts now misprice the work in both directions at once: the enterprise keeps paying for seats its agents are emptying, while absorbing consumption charges no one modeled. A Goldman Sachs study found enterprises burning through a full year's AI token budget in three months because nothing in the contract or the operating model meters the work. The repricing the market just executed on vendor equity will arrive in enterprise operating budgets through the renewal calendar, one contract at a time.
Defensive Risk
CFOs and COOs of enterprises holding multi-year per-seat agreements in CRM, ITSM, and HCM that renew within the next 12 months are the exposed population. The mechanism is contract-value mismatch: seat counts the agents are compressing on one side, unmetered consumption spend on the other, with no single owner accountable for the combined number. The trigger window is the FY2027 budget season and the renewal cycle that precedes it, when vendors mid-transition will defend revenue by converting confused buyers to hybrid terms the buyer has not modeled. The responsible defense is an immediate inventory of every software agreement by pricing unit and renewal date, with a standing rule that no per-seat renewal above a set threshold is signed without an agentic-usage clause and a modeled consumption ceiling.
Offensive Advantage
The enterprise that knows its own unit-of-work economics is negotiating against vendors that do not yet know theirs. Hybrid pricing is the 2026 transition state, which means vendors will trade flat-fee agentic terms for multi-year commitment right now, on terms that will not be available once their own consumption data matures. The buyers best positioned to capture that trade are the ones that have redesigned workflows around agents deliberately enough to measure cost per outcome, because they can price the contract against the work rather than against the headcount. That measurement capability is an organizational design question before it is a procurement question, an argument developed in the Scaling the AI Studio body of work.
The Governance Boundary Principle applies directly: the board owns oversight of the spend architecture and the question of who is accountable for the combined software and consumption number. Management owns the contract-by-contract redesign. A board that treats this repricing as a procurement detail has placed a balance-sheet-scale question below its own line of sight.
The Read
Over the next 30 to 90 days, expect Q2 earnings calls to widen the disclosure gap between vendors reporting consumption and agentic metrics and vendors still reporting seats. More flat-fee agentic enterprise agreements will be announced as reference deals, and the vendors that stop disclosing seat counts entirely are the tell that the internal forecast has flipped. Confirmation of this read appears if net revenue retention and seat growth visibly diverge across the major platforms by the July reporting cycle. The read is invalidated if seat-based growth reaccelerates and agent consumption disappoints at scale, in which case the drawdown was an overshoot and the per-seat model has more life than the market priced. Capital has already voted. The renewal calendar is where the enterprise casts its own ballot, whether or not anyone in the building is counting.
Methodology
This signal came from sector capital flows and sell-side positioning notes, cross-verified against vendor earnings disclosures: the IGV drawdown versus XLK and QQQ, the JP Morgan drawdown classification, Jefferies sector notes, Salesforce and ServiceNow reported results, and IDC and Gartner pricing-model forecasts. It scored at threshold because it pairs a two-trillion-dollar capital reallocation with a named, datable mechanism that lands on the enterprise buyer through the renewal calendar. No other signal scanned this run combined capital-flow magnitude with direct C-suite actionability at this level.