HubSpot lost 19% in one day. The market just told us what per-seat pricing is worth in an AI world.

I was scrolling through my phone on a Tuesday morning in May 2026 when someone I follow on X posted about HubSpot dropping 19% in a single trading session. Not 1.9%. Nineteen percent.

My first thought was: that's a massive single-day move for a company that size.

My second thought was: we have HubSpot seats.

So I clicked through to find out what had happened. No data breach. No accounting scandal. No CEO resignation. HubSpot had announced a pricing model change. They'd decided to charge customers based on when their AI agent resolves a customer case, rather than for the seat the agent occupies. The market, apparently, found that news just as alarming as a scandal.

The thing HubSpot said when it lost nearly a fifth of its market cap in a day was: "We'll charge you based on what the AI does, not how many humans you have." That sentence contains most of what you need to understand about what's happening to the SaaS industry right now.

This isn't a story about HubSpot being in trouble. Their revenue is up 23%. Their gross margins are 83%, which is extraordinary by any measure. The immediate trigger for the drop was softer-than-expected Q2 guidance and management noting the quarter had started slowly as they adjusted their agentic AI offerings. But the pricing model announcement amplified the market's reaction. The market isn't worried about this quarter. It's worried about whether AI agents extend the product or replace the customer.

The answer to that question is reshaping the economics of an entire industry.

When HubSpot Lost a Fifth of Its Value in a Day

HubSpot's "per-resolution" model for its Breeze AI agents is simple enough to understand. If an AI agent resolves your customer's issue, you pay for the resolution. You don't pay for a seat it might or might not occupy. CEO Yamini Rangan called it "the natural evolution" of pricing in an agentic world.

The market saw it differently. The market saw it as an admission.

What HubSpot was really saying, between the lines, was: "Per-seat pricing assumed there was a human behind every seat. That assumption doesn't hold anymore." And investors, who are generally good at reading between lines even when companies would prefer they didn't, took that admission seriously.

HubSpot isn't alone in making this move. Every major SaaS company appears to be separately arriving at the same conclusion right now, which is either reassuring (they're all adapting) or terrifying (they all know the old model is broken), depending on your disposition.

Salesforce introduced "Agentic Work Units" (AWUs) as a new measurement standard for tracking AI agent productivity. On the pricing side, its Agentforce customer-facing agents cost roughly $2 per conversation, a separate consumption charge sitting on top of your existing Salesforce licence. You still pay for your regular seats. The AI work gets billed on top. Think of it as a consumption meter bolted onto your subscription, where every customer interaction handled by an agent ticks the meter.

ServiceNow's CEO Bill McDermott used the company's May 2026 Knowledge conference to publicly reframe the entire business around AI orchestration rather than software features. McDermott told Fortune: "We don't live in the SaaS neighborhood." His keynote was titled "Welcome to Agentic Business." Whether that framing will age well, I genuinely don't know, but it's a significant pivot in positioning for a company that's been selling software features for decades.

GitHub Copilot already moved to usage-based billing from 1 June 2026. We wrote about this in our piece on Microsoft's dynamic workflows play. The pattern is consistent: usage, outcomes, resolutions. Not seats.

Zendesk moved in the same direction. They added per-resolution billing as an additional layer (approximately $2 per automated resolution), billed on top of their existing per-seat plans rather than replacing them. The framing circulating in commentary is striking: "The user on the other side of the screen is no longer human." To Human and To Agent is replacing To B and To C as the axis businesses actually care about.

Tomasz Tunguz, a venture analyst at Theory Ventures who tracks SaaS economics closely, has been arguing for months that the model is breaking. His framing: the subsidy era is over. The era he means is the one where per-seat pricing generated such extraordinary margins that vendors could absorb all kinds of cost inefficiencies and still be wildly profitable. The era he's referring to is the one where per-seat pricing generated such extraordinary margins that vendors could absorb all kinds of cost inefficiencies and still be wildly profitable. That cushion is compressing.

The scale of what's happening is worth sitting with. The iShares Software ETF (IGV), which tracks the sector broadly, is down roughly 23% year-to-date. HubSpot, Atlassian, and Figma are each down somewhere between 70% and 80% from their 52-week highs. Bulloak Capital put the total sector market cap losses at an estimated $2 trillion. That's not a correction. That's a renegotiation.

For context on what uncapped AI spending looks like at the enterprise level:

Related Article9 min read

How do you accidentally spend $500M on AI?

An unnamed enterprise company spent $500 million on Claude AI in a single month because nobody configured a spending limit. Uber burned its 2026 AI...

Read full article

Why Per-Seat Pricing Was Always Going to Break

Per-seat pricing is a brilliant model when every seat has a human in it. It's predictable for both buyer and seller. It scales with headcount, which scales with revenue, which makes the maths work for everyone. Software companies have been operating on 80-90% gross margins partly because of how elegant this model is.

AI agents are seat-shaped. They use software. They need access to systems. But they don't pay for licences. They don't count in headcount. And they can work continuously in ways that humans cannot, which is the bit that makes the economics genuinely uncomfortable.

Jason Lemkin, who founded SaaStr and has been writing about SaaS economics longer than most people in the industry, has been making a version of this point for months. The argument, paraphrased: if AI agents can do the work of 100 sales reps, you don't need 100 Salesforce seats. You need some seats and some agent quota. The market agreed with him in May 2026, quite publicly.

Here's the structural problem. Per-seat pricing assumed every seat had a human butt in it. That was a fair assumption for about 30 years. It stopped being a fair assumption sometime in late 2025, and the invoice just arrived.

The margin pressure is coming from two directions simultaneously. AI agents reduce the number of human seats a company needs. At the same time, running AI agents costs real compute money in ways that traditional software serving human users didn't. Software companies are watching their two most reliable economic assumptions erode at the same time: headcount-driven seat revenue down, compute costs up. Hence Tunguz's line about the subsidy era being over.

A post that circulated widely in late May captured the investor logic precisely:

Tunca Üçer

@tuncaucer

The SaaSpocalypse is over. But the lesson isn't "software is back", they say.

HubSpot: +23% revenue. 83% gross margins. Down 46% YTD.

The market isn't pricing our quarter. It's pricing whether AI agents replace our product or extend it.

saastr.com/the-saaspocaly…

2

31 May 2026

The market isn't pricing what HubSpot earned this quarter. It's pricing the question: in three years, when AI handles 60% of the customer interactions HubSpot currently bills for, what does the per-seat model look like then? The answer the market appears to be settling on is: smaller. Hence the valuation.

Most SaaS vendors haven't had to confront this question in their renewal conversations yet.

They will.

What's Replacing It

The good news (if you're looking for some) is that the industry isn't replacing a working model with nothing. Three distinct pricing approaches are emerging, and while it's not yet clear which one wins, they're all more defensible than "seats times price per month forever."

Per-resolution and outcome-based pricing is what HubSpot and Zendesk are doing. You pay when something gets done. For customer service and support workflows, this is relatively clean. There's a defined end state: the issue is resolved or it isn't. The risk for vendors is that AI gets very good at resolving things quickly, which compresses both the revenue per interaction and the total volume they can bill for. The risk for buyers is the opposite: opaque outcome-counting, where "resolved" gets defined in ways that maximise vendor revenue.

Per-action and consumption pricing is Salesforce's approach. $2 per autonomous action, billed separately from your subscription. This is more granular and more measurable, but it introduces something most enterprise procurement teams haven't had to manage before: unpredictable AI-driven cost spikes. If your AI agent is unusually active in a given month, your Salesforce bill goes up in ways that didn't happen under the old model. Finance teams will enjoy explaining that variance to their CFOs.

Platform and orchestration pricing is ServiceNow's direction. Rather than charging for features or outcomes, you sell the runtime that runs agents. You're the infrastructure layer, not the application. This is a higher-risk, higher-reward bet: if every enterprise ends up running AI agents through a platform layer, the orchestration layer is enormously valuable. If everyone runs their own, it's less so.

All three of these will coexist for years. The SaaS industry is having the same pricing identity crisis that mobile had when unlimited data plans appeared. Someone has to absorb the cost of AI usage. Right now, they're all trying to pass it to someone else in slightly different configurations.

What's clear is that "X seats at $Y per month" as the only pricing option a vendor offers is ending. If your major SaaS vendors haven't started talking to you about this yet, the conversation is coming.

What Australian Businesses Should Do

I want to be practical here rather than just alarming. Because the actual implication for most Australian businesses isn't "the software you use is about to explode in price." It's more nuanced than that.

If you're a SaaS buyer, which most businesses now are in some form:

First, audit your seats before your next renewal. I know this sounds obvious. We did it at Webcoda after I started writing this article and found three tools with seats that haven't been logged into in six weeks. I'm part of this problem. We all are. The per-seat model has been so comfortable and so predictable that nobody scrutinises it the way they'd scrutinise any other recurring cost of that size.

Second, identify which of your current workflows could be handled by AI agents. Not "which of my staff should I cut." That's a different conversation. The question is: which tasks are repetitive, structured, and outcome-definable enough that an AI agent could handle them? Because if the answer is "quite a few," then the number of seats you actually need in two years is different from the number you need today, and you should know that before you sign another three-year deal.

Third, start asking your vendors the question they don't particularly want to answer: "What's your plan for outcome-based pricing?" If they look blank, or they give you a very polished non-answer, that tells you something about how prepared they are for this shift. The vendors who've thought seriously about it will have a coherent answer. The ones who haven't will change the subject.

Fourth, watch the next 12 months of pricing announcements from your biggest vendors. HubSpot and Salesforce are first movers. ServiceNow and Zendesk are close behind. Others will follow because they have to. The businesses that understand the new pricing models before they're forced onto them will negotiate better deals than the ones who discover them at renewal time.

If your business has any SaaS-adjacent revenue (and agencies like Webcoda sit squarely in this category), the same structural question arrives at your door from the other direction. Are you charging per seat? Per project? Per outcome? The market is retraining buyers to think in results, and that changes procurement conversations eventually. We've been thinking about this at Webcoda. We don't have a clean answer yet, which is at least more honest than pretending we do.

Our piece on Microsoft cancelling Claude Code for its own engineers is the buyer-side version of the same story. When Microsoft decides a seat is no longer worth its cost because an alternative delivers better outcomes, the vendor loses that seat. The mechanism is the same, just viewed from the other direction.

Related Article9 min read

Microsoft told its engineers Claude Code was too expensive. Then started selling Anthropic's AI at a 65% markup.

Microsoft cancelled Claude Code licences for thousands of its own engineers because the tool was too popular and cost too much. Meanwhile, the...

Read full article

The broader AI cost context connects directly. The compute costs driving vendor margin compression are the same ones creating budget pressure on the buyer side. The $2 trillion in sector losses isn't disconnected from the AI investment boom. They're the same story told from two vantage points.

Related Article11 min read

You laid off your engineers to save money. Then spent it all on AI.

Companies cut hundreds of thousands of engineers in 2023-2025. AI productivity would cover the gap. Then the invoices started arriving. Uber burned...

Read full article

The Actual Level of Alarm Required

Per-seat pricing isn't dead today. The vast majority of your software bills this month will still arrive as "X seats at $Y." The transition is happening at different speeds in different categories, and "AI-replaceable workflows" varies enormously by industry, company size, and how ambitious you want to be about implementation.

But the direction is clear. And HubSpot's 19% single-day drop is important precisely because it wasn't driven by anything going wrong. Revenue is up. Margins are exceptional. The business is functioning. The market is discounting it because it looked at the pricing model change and decided the old assumptions don't hold.

The SaaSpocalypse sounds dramatic. The actual event is more boring: you'll pay differently for software. That's it. But "differently" just cost the sector somewhere north of $2 trillion in market capitalisation, and HubSpot's investors a fifth of their money in a single trading session. Boring is relative.

If you're renewing a significant SaaS contract this year and haven't had this conversation, it's worth having before you sign. We're happy to help you think through which questions to ask. Get in touch.

---

Key Takeaways

• HubSpot lost roughly 19% of its market value in a single trading session in May 2026 after announcing "per-resolution" pricing for its Breeze AI agents. The announcement wasn't bad news about the business. It was an admission about the model.
• The broader SaaS sector has shed an estimated $2 trillion in market capitalisation. The iShares Software ETF (IGV) is down roughly 23% year-to-date. HubSpot, Atlassian, and Figma are each down 70-80% from their 52-week highs.
• Three new pricing models are emerging: per-resolution/outcome (HubSpot, Zendesk), per-conversation consumption (Salesforce Agentforce at ~$2/conversation, with Agentic Work Units as the measurement standard), and platform/orchestration layer (ServiceNow).
• The structural problem: per-seat pricing assumes a human behind every seat. AI agents use seats but don't pay licences. The maths stops working when you replace enough of the humans.
• Australian businesses should audit their SaaS seats before their next renewal, identify AI-replaceable workflows, and start asking vendors about their outcome-based pricing plans before those conversations are forced.
• Per-seat pricing isn't dead today. But it's being renegotiated in real time, and the businesses that understand the new models before they're imposed will negotiate better deals than those who discover them at renewal time.

---