HubSpot Is Making Record Money. Its Stock Is Down 60%. Here's Why.

HubSpot is making record money. Its stock is down 60% from last year.

The two facts have been true at the same time since February 2026. Revenue grew 20% in the latest quarter, up to $912 million. Operating margins expanded to 20.3%. The company turned profitable on a GAAP basis for the first time in its history, with $43 million in net income where there had been a loss a year earlier. Free cash flow is running above $650 million annually.

And the stock has been sold from about $525 down to the $260s, erasing roughly $17 billion in market value.

The stated reason is AI. Not in the abstract sense, but a specific mechanism: AI agents can do the work of human employees inside CRM systems, which means fewer employees need software seats, which means the per-seat pricing model that has run enterprise software for 30 years stops working. Between January and February 2026, roughly $2 trillion in software market cap evaporated. Atlassian, Salesforce, ServiceNow — all of them fell as investors priced in the possibility that their core revenue drivers were about to disappear.

Most of the writing about this is right about the direction and wrong about what it means for HubSpot.

The per-seat model is being disrupted. That's not debatable. An AI agent that resolves customer support tickets doesn't need a license. An agent that runs outbound prospecting doesn't occupy a seat. The equation that powered SaaS growth for a generation — more employees, more seats, more recurring revenue — has a crack in it that's getting wider every quarter.

But HubSpot is not a per-seat company anymore. It stopped being one in April 2026, when it changed how it charges for AI agents. The Customer Agent went from $1 per conversation to $0.50 per resolved conversation. The Prospecting Agent went from a flat monthly fee to $1 per lead recommended for outreach. The company's own language: "You pay when it works, full stop".

This is not a cosmetic change. It's the difference between being the company that sells seats to the humans AI replaces, and being the company that gets paid when AI does the work.

Here's where the story gets less clean. The pricing pivot was a rational response, but it created a new problem. Sales cycles lengthened. Deals require C-suite and board approval now instead of departmental sign-off. Customer additions dropped from an expected 9,000 to 10,000 per quarter to an actual 7,000, and the next quarter guidance is only 5,000 to 6,000. Revenue growth decelerated from 23% in the first quarter of 2026 to 20% in the second, with constant-currency growth at 17%.

The stock fell 20% in a single day on August 5th when these numbers came out. The worst single-day loss in 12 years.

So you have a company that is simultaneously more profitable than ever and growing more slowly than the market expected. The two facts are not contradictory. They're the same fact viewed from different angles. HubSpot is transitioning from selling software licenses to selling outcomes. That transition makes the company more valuable in the long run if the AI works — you're now paid on value delivered, not time spent at a keyboard. But it also makes growth harder to predict and slower to accelerate.

The question that matters for investors is whether the pricing change is a genuine adaptation or a concession. The evidence goes both ways.

On the adaptation side: Over 55% of HubSpot's Pro-tier customers now use AI agents or the Breeze Assistant. Monthly agentic actions tripled year over year. Data Agent adoption grew 80% quarter over quarter. Credit consumption rose 67% in one quarter. These are not the numbers of a company losing its customer base. They're the numbers of a company where the product is being used more intensively, just in a way that doesn't map cleanly to the old per-seat metrics.

On the concession side: Average subscription revenue per customer grew only 4% reported, 2% in constant currency. Net revenue retention slipped to 102%, flat guidance for the rest of 2026. The company is growing by adding customers, not by extracting more value from existing ones. And the customer addition rate is decelerating, not accelerating.

The more interesting question is what happens when AI agents get really good at their jobs. HubSpot's customer agents resolve 65% of conversations. Gartner projects 80% autonomous resolution by 2029, probably faster. At 80% resolution, a volume of 5,000 monthly conversations costs a customer $2,000 per month. At 90%, it's $2,250. As accuracy approaches universal, per-resolution pricing starts looking the same as per-conversation pricing. The pricing distinction that matters today may not matter in two years.

I suspect the real test isn't whether AI agents replace human seats at HubSpot. It's whether HubSpot can be the platform those agents run on, rather than being replaced by them. The company has 306,000 customers, $3.45 billion in annual recurring revenue, and a data moat built over 17 years of CRM interaction history. An AI agent needs context to be useful, and that context lives in the CRM. HubSpot's position is that it's the operating system for these agents, not the UI that gets bypassed.

What the valuation tells you is that the market hasn't decided which story it believes yet. At $13 billion, HubSpot trades at 3.8 times trailing revenue. Its free cash flow yield is about 5%. These aren't growth stock multiples. They're what you see when investors have written off the acceleration story but haven't written off the business. The old multiples were priced around 20% growth reaccelerating. The current ones price in 20% growth continuing at a flat rate.

There's a gap between those two valuations that could go either way. If the outcome-based pricing model gains traction and customer growth reaccelerates — which would require agents to become a visible revenue engine rather than a usage metric — the market could revisit the higher end. If customer additions keep declining and net revenue retention stays flat, the current valuation may prove generous for a company whose core growth engine is slowing.

The thing to watch is not whether AI is coming. It's already here. Watch whether HubSpot's AI credit consumption keeps growing faster than its seat-based subscription revenue. That ratio, tracked quarterly, will tell you whether the pricing pivot is working. Watch whether average revenue per customer starts accelerating, which would signal that customers are buying AI outcomes in addition to traditional seats, not instead of them. And watch whether the 5,000 to 6,000 quarterly customer additions guidance holds or drops further.

The stock's price action over the past year tells you the market has already done one repricing. The question is whether this is the bottom of that repricing or whether there's another leg to go. You won't find the answer in the AI narrative. You'll find it in whether the company's own metrics show the new pricing model absorbing the disruption or merely delaying it.