Software shed about US$1tn in a week; seat-priced renewals are next

01Seven sessions, about US$1 trillion: what the market sold

On 30 January Anthropic open-sourced 11 plugins for Claude Cowork, its agent for office work, covering sales, finance, legal, marketing, customer support and data among others.¹ On Tuesday 3 February the legal plugin, which reviews contracts and triages non-disclosure agreements, became the market’s reason to sell. heise online counted US$285 billion of market value gone across the sell-off; Adobe and Salesforce each fell about 7% that day.² The Sunday Guardian reported the same US$285 billion and quoted Jeffrey Favuzza, via Bloomberg: “We call it the ‘SaaSpocalypse,’ an apocalypse for software-as-a-service stocks.”³

By 5 February the S&P 500 software and services index had fallen for seven straight sessions and, by Reuters’ count, shed about US$1 trillion in market value since 28 January.⁴ TheStreet put the loss at more than US$1 trillion since January three weeks later.⁵ The slide began before the plugins: the iShares software ETF (IGV) fell 4.9% on 29 January and 4.6% on 3 February.⁷

The index bottomed on 10 April, 36.6% below its September high, then rose 44% to 1 June. June gave back about half of that gain, and the names buyers actually renew did worse than the index.⁷

02By 30 June, HubSpot was down 55% and Salesforce and Adobe 41%

We took the closing price of ten listed application-software vendors on 31 December 2025 and on 30 June 2026. HubSpot closed at US$182.51, against US$401.30 at the start of the year. Salesforce went from US$264.91 to US$156.66, Adobe from US$349.99 to US$205.02. Atlassian and monday.com each lost about half. Over the same six months the S&P 500 rose 9.5%.⁷

03Revenue kept growing; the market priced the seat count

None of this was a revenue collapse. HubSpot’s revenue for January to March rose 23% to US$881 million, its customers 16% to 299,458.¹⁰ Salesforce reported US$11.1 billion for February to April, up 13%, and Agentforce annual recurring revenue of US$1.2 billion.¹¹ Adobe’s quarter to May grew 12.7%.¹² Investors were not reading the last quarter. They were reading the multiplier in the next ten.

The argument was stated plainly in February: “Seat-based SaaS compresses as headcount compresses.”⁶ Freshworks had put the same risk in its own annual report, filed on 26 February: it charges “based on the number of users they enable as ‘agents’”, its features let customers handle support “without the need for substantial human interaction”, and so “customer agent staffing requirements may be minimized, and our revenue may be adversely affected.”⁸ If an agent answers the tickets, the help desk renews fewer seats, and the vendor’s revenue falls even if the customer stays.

The two fastest growers, Atlassian and HubSpot, fell the most. The evidence against the panic came from the other side of the AI trade: Nvidia’s Jensen Huang told CNBC, “I think the markets got it wrong.”⁵ Both views can be right for a buyer. The vendors are not failing; they are under pressure to keep every seat.

04The vendors already sell a meter beside the seat

Salesforce has been building the alternative for a year. Since May 2025 it has sold Flex Credits at US$500 per 100,000, with an Agentforce action costing 20 credits, or US$0.10, beside its older US$2 per conversation; a Flex Agreement lets a customer move spend between user licences and credits.¹³ In December it told analysts it had sold 16 fixed-price Agentforce enterprise licence agreements (AELAs) in a quarter, with about 100 more in the pipeline, while its revenue chief said seat-based offers “continue to entice other customers for their spending predictability.”¹⁴ Freshworks went the other way on cost: on 5 May it announced a restructuring affecting about 500 employees, 11% of its workforce, to “streamline the Company’s organizational efforts and product development process” and make more use of AI and automation.⁹

05What to ask for at renewal

A vendor whose shares have halved is judged by investors on whether customers renew and expand. Freshworks reports a net dollar retention rate of 106%, and every true-down a customer wins lowers that kind of number.⁹ That gives the buyer something to trade: commitment for terms. The five requests below are our sequence, in order of how much they are worth.

Price protection and the right to shrink pull against each other, and the vendor will offer one to keep the other. A cap on the uplift is worth most to a company whose user count is stable; a true-down right is worth most to one about to put agents on its help desk or sales desk. Decide which you are before the first call. If the vendor proposes credits, ask for twelve months of your own usage data priced at the credit rate first: a meter you cannot forecast is a budget you cannot defend. And read the definition of an action. Salesforce charges 20 credits for one; how many actions a single customer request triggers is set by how the agent is built, not by the buyer.¹³