- Nerd Out on Business
- Posts
- The company that invented inbound is running out of inbound.
The company that invented inbound is running out of inbound.
Brian Halligan returned as CEO last week to refound HubSpot around AI agents. Here’s the bet.
Today, I'm digging into HubSpot (HUBS) again.
They sell CRM software that SMB’s use to run their marketing, sales, and customer support. I covered them once before, and a lot has happened since.
Last Thursday, co-founder Brian Halligan came back as CEO, replacing Yamini Rangan. His first post on X said he and Dharmesh Shah are "completely re-founding the company" and that HubSpot is now "the Customer Harness Company."
This is also the fourth company in a few weeks I've covered whose growth revolved around SEO (Google traffic), after Chegg, Tripadvisor, and NerdWallet. HubSpot built a $3.6B software business on it, and literally wrote the book on how to do it. They coined the phrase and built their brand around the term “inbound marketing”.
A few things that stood out to me:
New customer growth got cut in half in two quarters. HubSpot added ~10,000 net new customers every quarter for five straight quarters. Then ~7,000 in Q2 2026, 5,258 in Q3, and a Q4 guide below 5,000. Management blames longer sales cycles, price-sensitive small businesses, and weaker inbound demand as AI changes how buyers find software.
Revenue is the last number to show it. Q3 revenue still grew 17%. But the Q4 guide implies 12%, net revenue retention slid from 105% to 101% in three quarters, and management expects net new ARR to decline this year. Trends like this are powerful and hard to combat.
What is a "Harness"?. A harness is the scaffolding around a model (tools, memory, permissions) that lets it do real work (eg. Claude Code/Lovable/Bolt/Replit/etc) but the concept doesn't need to just apply to software development. Halligan's bet is that agents like Claude and ChatGPT will do more of the marketing, selling, and support, and they'll need a trusted place to read and write customer data. HubSpot wants to be that layer instead of the screen a salesperson logs into.
Their pricing conundrum. If agents do the work, fewer people log in, and HubSpot still charges mostly by seat. Its new seats-plus-credits pilot drove 2x to 3x more usage, but at lower starting prices.
When I first wrote about HubSpot, I said per-seat pricing was under pressure and that changing how you make money as a public company is incredibly hard. Now it's happening in real time.
Halligan told Bloomberg the company may have focused too much on short-term financial targets and not enough on building great products.
When I was working at John Deere, I sat in a memorable meeting where one of the product leaders emphasized that there's a lot of company effort around new business models. He hammered home the point that our job as product managers was to first and foremost build great products for our customers. That's a lesson that should stick with any company.
Here's another lesson for all of us: revenue is the last metric to tell ua something is wrong. We need to focus on where customers are coming from, where the leads are coming from, and how customers are using our product. Because by the time revenue slows, the problem can two or three quarters old.
TL;DR
HubSpot sells subscription CRM software to small and mid-sized businesses; subscriptions are ~98% of revenue.
Net new customers fell from 10,752 in Q1 2026 to 5,258 in Q3, Q4 revenue growth is guided to ~12%, and net new ARR is expected to decline in 2026.
The causes: a weaker inbound engine as AI changes software discovery, small-business price sensitivity, and HubSpot's own shift from seats to credits and outcome pricing.
GAAP operating margin turned positive (5.5% TTM), gross margin is slipping as AI costs rise, and the stock is down 41% over the last year and 74% over five years.
Entrepreneur takeaway: your best acquisition channel is never permanent, and revenue is a lagging indicator.
The 30,000-Foot View
HubSpot sells a CRM plus add-on "Hubs" (Marketing, Sales, Service, Content, Commerce, Data) that share one customer record. Customers start on free tools or a cheap Starter plan, add seats and contacts as they grow, then upgrade tiers and add Hubs. Gross churn is steady in the high 80s (customer dollar retention), which is normal with this many small customers.
For nearly 20 years, the front end of that machine was inbound marketing: free content that ranked on Google, free tools, and small businesses finding HubSpot on their own. That made customer acquisition cheaper than almost anyone in SaaS. Now Google answers more questions on the results page and buyers ask ChatGPT which CRM to use. By mid-2025, 90% of leads already came from sources other than the blog.
Pricing is changing too. In 2026, HubSpot moved two AI agents to outcome pricing ($0.50 per resolved support conversation, $1 per qualified lead), let customers swap seats for AI credits mid-contract, and piloted a single seat plus usage credits in place of separate Hubs. That pilot expanded from the Nordics and Benelux to all new EMEA customers in Q4.
"Customer Harness Company" names where that leads. Shah describes a "headless" HubSpot whose capabilities are consumed by AI agents through MCP connectors. HubSpot would get paid for context, data, and permissions, whether or not a human ever logs in.
Revenue mix (Q3 2026):
Subscription: $930.2M, ~98%
Professional services and other: $17.3M, ~2%
Key Stats
Market cap: ~$10.6B
TTM revenue: ~$3.59B
TTM gross margin: ~83.0%
1Y total return: -41.4%
Customers: 311,704 (up 12%)
Net revenue retention: 101% (Q3 2026)
Employees: ~9,400 before the October layoff [VERIFY]
Industry: Application software, CRM
Company History
2005: Brian Halligan and Dharmesh Shah, who met at MIT Sloan, found HubSpot in Cambridge, Massachusetts.
2009: They publish Inbound Marketing, naming the category HubSpot sold into for the next 15 years.
2014: IPO on the NYSE and launch of a free CRM, which put HubSpot into hundreds of thousands of small businesses.
2018: Halligan introduces "the Flywheel" at INBOUND as the replacement for the sales funnel.
2021: Rangan becomes CEO; Halligan moves to executive chairman.
2025: Repays its convertible notes, buys back $500M of stock, and posts its first full year of GAAP operating profit ($11.4M on $3.13B of revenue).
August 2026: Q2 net adds of ~7,000 miss the 9,000 to 10,000 expected. Shah launches YouSpot, an AI-native CRM for solo operators at $10/month with no seats.
October 2026: Cuts nearly 660 roles (~7%) and reorganizes product teams around customer outcomes.
November 12, 2026: Halligan returns as CEO to lead a "refounding." Karen Burgess (from Adobe) is named the next CFO.
Show Me the Money
Standout financial features:
Growth stepped down every quarter this year: 23% in Q1, 20% in Q2, 17% in Q3, and ~12% implied for Q4. The Q4 guide ($945M to $947M) is roughly flat with Q3.
GAAP operating margin went from (9.3%) in 2023 to 5.5% TTM. Stock comp is still ~$504M TTM, more than 2.5x GAAP operating profit.
Gross margin slipped from 85.0% in 2024 to 83.0% TTM. In Q3, cost of subscription grew 28% against 18% subscription growth, partly from AI usage. AI is a real cost line now.
~$974M of buybacks in the first nine months of 2026 ($537M in Q2 alone) cut net cash by $530M in nine months. Diluted shares fell from ~53.2M to ~49.5M.
Financial Data
Metric | FY2023 | FY2024 | FY2025 | TTM (Sep 2026) |
|---|---|---|---|---|
Revenue | $2.17B | $2.63B | $3.13B | $3.59B |
Gross Profit | $1.83B | $2.23B | $2.62B | $2.98B |
Gross Margin | 84.4% | 85.0% | 83.8% | 83.0% |
Ops Profit | ($201M) | ($68M) | $11M | $197M |
Ops Margin | -9.3% | -2.6% | 0.4% | 5.5% |
CapEx | $100M | $127M | $181M | $227M |
Net Debt | ($1.26B) | ($1.77B) | ($1.84B) | ($1.31B) |
CapEx includes capitalized software. Net Debt excludes leases and nets all cash and investments, same method as my first HubSpot issue.
The New Customer Math
Quarter | Net New Customers | Total Customers |
|---|---|---|
Q3 2025 | 10,898 | 278,880 |
Q4 2025 | 9,826 | 288,706 |
Q1 2026 | 10,752 | 299,458 |
Q2 2026 | 6,988 | 306,446 |
Q3 2026 | 5,258 | 311,704 |
Q4 2026 (guide) | Below 5,000 |
Customers aren't leaving (gross churn is unchanged). They're buying fewer seats and upgrading less. And starting in Q3, the count includes $10/month YouSpot subscribers, so the 5,258 is flattered.
Stock Performance
Period | Total Return | Annualized |
|---|---|---|
3 months | +0.5% | |
1 year | -41.4% | |
5 years | -74.2% | -23.7% |
10 years | +275.8% | +14.2% |
Company | 1Y Total Return |
|---|---|
HubSpot (HUBS) | -41.4% |
Salesforce (CRM) | -17.2% |
Freshworks (FRSH) | +9.9% |
The 10-year number still looks great, but anyone who bought in the last five years is down roughly three-quarters, and HubSpot lagged both its big-company and SMB-focused rivals this year.
The N.O.O.B. Nine - Competitive Powers
The Nerd Out on Business Nine is made up of Hamilton Helmer's famous "7 Powers" of competitive advantage (Scale Economies, Network Economies, Counter-Positioning, Switching Costs, Branding, Cornered Resource, and Process Power) combined with two of my own (Data Flywheel and Distribution Advantage).
Power | Score | Rationale |
|---|---|---|
Branding | 4/5 | Still the default SMB CRM and the company that coined "inbound." |
Data Flywheel | 4/5 | The customer record is the context agents need; CRM objects created through LLM connectors grew 157% sequentially in Q3. |
Process Power | 3/5 | Pricing, trial, and sales changes stretched buying cycles, and a layoff plus a CEO and CFO change in six weeks means the operating model is being rebuilt. |
Scale Economies | 3.5/5 | GAAP margins are expanding on a fixed cost base, but AI costs are growing faster than revenue. |
Switching Costs | 4/5 | Gross churn held steady in the high 80s even as customers trimmed seats. |
Cornered Resource | 3/5 | Founders back in charge and a big partner ecosystem, but nothing exclusive. |
Network Economies | 3/5 | 1,800+ integrations create pull, but agents can connect to any CRM with an MCP server. |
Counter-Positioning | 2.5/5 | HubSpot is now the incumbent getting undercut by AI-native CRMs; YouSpot is it trying to undercut itself. |
Distribution Advantage | 3.5/5 | The content engine that earned a 5 last time is weakening; LLM connectors are an early replacement. |
Average Score: 3.4/5 - A moat shifting from distribution to data, with the refounding betting that the customer record becomes the thing agents can't work without.
Memorable Marketing
HubSpot's best marketing has always been naming a new game before selling the product: describe how buyers are changing, name it, and position HubSpot as the tool for the new game.
Notable tactics:
Inbound Marketing (2006 to 2009): A blog, free tools like Website Grader, and a book that named the idea of buyers finding you through content. The category name sold the product for a decade.
Free CRM (2014): A real CRM at $0 that made HubSpot the system of record before small businesses ever talked to sales.
The Flywheel (2018): Reframed customers as a source of growth after the sale, and gave Service Hub a reason to exist.
Owned media (2021 onward): Bought The Hustle and built a podcast network once search dependence looked risky.
HubSpot inside the AI tools (2025 onward): Connectors for ChatGPT and Claude. 20,000+ customers used the ChatGPT connector in its first quarter.
"The Customer Relationship Harness" (November 2026): The new tagline drew mostly negative reactions. Former HubSpot product marketer Marcus Andrews pointed out that Inbound and the Flywheel got mocked too.
Tactical takeaways:
A category name gives your sales team a story and buyers a vocabulary that points back to you.
Start your second acquisition channel while the first still works. HubSpot started buying media years before inbound cracked.
Show up inside the tools your buyers are moving to. Being a connector in ChatGPT is the new SEO.
AI Uses & Opportunities
Current exposure:
AI is the main headwind at the top of the funnel and a seat headwind once customers sign up.
AI is also a cost of revenue now, which is why gross margin is down two points from 2024.
The agents are getting real use: Customer Agent resolved 2M+ support issues in Q3, agent trials rose 64% sequentially, and Prospecting Agent users generated 76% more qualified leads than non-users, per HubSpot.
Future opportunities:
Headless HubSpot: charge per agent action through MCP and become the permissions and audit layer for customer data.
Extend outcome pricing to booked meetings, renewals saved, or pipeline created.
Sell answer-engine tools to its 311K customers, who all have the same inbound problem HubSpot does.
Let YouSpot be the new free CRM, even if it cannibalizes Starter.
Bumps in the Road
New customer growth halved in two quarters, and Jefferies expects 2027 guidance to come in materially below the current 11% consensus.
Moving from seats to credits shrinks starting contracts. It only works if usage grows faster than seats shrink.
New CEO, new CFO in February, new President of Product, and a layoff that runs through Q1 2027. Refounding a public company with quarterly guidance is hard.
The harness could get commoditized from above if Claude, ChatGPT, or Salesforce keep their own customer context.
Less cash for the rebuild after ~$1.47B of buybacks.
Your Swipe File
Your best acquisition channel often isn't permanent. HubSpot's inbound engine was the best in SaaS and AI search still dented it. Plan for unforeseen marketing disruptions via diversification.
Try to align your pricing with the value that customers receive from or how they use the product/service. Per-seat pricing never made the most sense, and now it's creating huge issues for many SaaS companies.
Revenue grew 17% in a quarter when net new customers fell 52% year over year. Revenue is a lagging indicator. Make sure your analytics and strategy are also focused on leading indicators.
I've been a bit inconsistent with these newsletters, but I'm going to pump up the frequency. My goal is three per week for the foreseeable future. We'll see if I can deliver.
If I don't, email me and tell me to step it up!
Nick