From Shubhi K | Product & Market Analysis
HubSpot AI Spend Caps Put a Ceiling on the Bill, Not a Forecast on It
On this page
HubSpot's new AI spend caps do one thing well. Once you buy extra credits, your account gets a default ceiling 50% above your total monthly credits, so the worst month now has a known price. HubSpot AI spend caps do not tell you what next month will cost. They change what a bad month looks like: the bill stops growing, and your agents stop working.
Key takeaways
- The default cap is narrower than the headline. HubSpot's own billing documentation sets the 50% account-level limit automatically only after an account buys extra credit capacity packs. Accounts on included credits alone already stopped at their allowance.
- The headroom is always one third of the ceiling. A limit of 150% of monthly credits means the overage slice is 50 of 150 units, whatever the account size. On HubSpot's own example of 8,000 monthly credits, the ceiling is 12,000 and the most you can overspend is $40.
- Hitting the cap turns a cost problem into a coverage problem. HubSpot says credit-using features pause until the next reset date. For a support team running Customer Agent, that means the agent is off for the rest of the month, not that the invoice is capped and work continues.
- HubSpot gave up a revenue ratchet to sell predictability. Before 16 September 2026, accounts that bought extra credits defaulted to auto-upgrade, which added a capacity pack for the rest of the commitment term. Pay-as-you-go is now the default, and caps sit on top of it.
Do HubSpot's spend caps fix AI billing predictability? Partly. The caps bound your worst month at 150% of your monthly credits and let you set per-feature limits, so no single team can drain the shared pool. They do not predict usage. When demand exceeds the cap, credit-using features pause until the next reset, so the risk moves from your invoice to your service levels.
What HubSpot changed on 15 October 2026
HubSpot announced three controls in one post on 15 October: default spend caps, self-service swaps from seats to credits, and two ways to pay for credits. The framing came from Jon Dick, HubSpot's Chief Customer Officer: "Nobody should have to miss out on getting the most out of AI because they don't know what the bill will be".
A seat has a fixed price. A credit depends on how much work an agent does, and HubSpot's own announcement concedes this is "hard to predict". The question is whether the new controls solve that, or only contain it.
The default spend cap
HubSpot's announcement says it "turned spend controls on by default" and sets an account-level credit limit "up to 50% above a customer's monthly credits". The customer can change it at any time.
The billing documentation is more precise. It says the limit is set automatically after you purchase additional HubSpot Credits capacity packs. An account with no packs already had a hard stop: when included credits run out, usage-based features pause until the reset date or until the account buys more.
Feature-level limits and alerts
Admins can set a credit limit per tool, and once a tool hits it, further usage for that tool is blocked until the limit resets or rises. Feature limits cannot exceed the account limit. Action-level limits exist only for buyer intent and are off by default.
Alerts go to billing contacts and Super Admins at 75%, 85% and 90% of credits used, and again when the limit is exceeded. They arrive as in-app banners and email. HubSpot also extended the same logic to marketing contacts: it says it will no longer upgrade a business into a bigger plan "just because its usage is growing".
Seat-to-credit swaps
Customers can now swap seats for credits, or for another seat type, without waiting for renewal. HubSpot's stated reason is blunt: "The usual result is paying for seats nobody uses while credits run low". The announcement does not publish eligibility rules, a conversion rate, or whether a swap can be reversed.
Credit packs and pay-as-you-go
The third control is payment. Customers can prepay credit packs at a flat monthly rate, at the published price of $10 per 1,000 credits, or pay as they go for usage beyond monthly credits. Pay-as-you-go overage bills monthly in arrears, in increments of 10 credits, on a separate invoice.
| Control | What it changes for a buyer | What it leaves unsolved |
|---|---|---|
| Account-level cap, 150% default | Puts a known price on the worst month once you buy packs | Says nothing about the expected month, and pauses work at the limit. |
| Feature-level limits | Turns one shared pool into per-team budgets | Off until an admin sets them; one team can still starve at its own limit. |
| Seat-to-credit swaps | Moves budget from idle seats to agent work mid-contract | No published rate, eligibility or reversal terms. |
| Pay-as-you-go default | Removes the auto-upgrade ratchet for new credit purchases | Accounts that bought packs before 16 September keep auto-upgrade unless changed. |
HubSpot's 2026 pricing timeline, dated
The caps make more sense read as the sixth step in a sequence, not as a standalone feature.
14 April. Customer Agent moved from $1.00 per conversation to $0.50 per resolved conversation, and Prospecting Agent moved to $1.00 per lead recommended for outreach. Dick's line then was "You pay when it works, full stop." The rate card is covered in the analysis of the Breeze credits repricing.
7 May. HubSpot reported Q1 revenue of $881.0 million, up 23%, and guided full-year revenue to $3.700 billion to $3.708 billion. The stock fell anyway. Tickeron reported it down nearly 19% after hours from a $243.74 close, and about 24% lower in premarket trading on 8 May, citing visibility concerns about outcome pricing among other reasons. I could not verify the 8 May closing move.
5 and 6 August. Q2 revenue was $911.7 million, up 20%, but net customer additions came in at 7,000 against 9,000 to 10,000 expected. Full-year guidance moved to $3.678 billion to $3.686 billion. Shares fell 19% the next session. Chief executive Yamini Rangan said customers "want proof of value before they commit and predictability in what it costs."
16 September. Pay-as-you-go became the default overage mode for credit purchases made on or after that date. Earlier purchasers stayed on auto-upgrade by default.
1 October. Partner agencies report that new EMEA customers moved to a seats-and-credits model, buying one edition plus seats instead of separate hubs. That date rests on agency reporting, not a HubSpot document, so treat it as directional.
15 October. Default caps, self-service swaps and the two payment options.
The sequence tells you what the caps are for. HubSpot cut unit prices in April and investors punished it twice. Its chief executive then said buyers wanted predictability. Five weeks later, the company changed its overage default, and four weeks after that it shipped caps. That is a vendor answering a stated buyer objection, fast.
How the HubSpot AI spend cap works in numbers
The cap formula is simple. The default limit is 150% of total monthly credits, meaning included credits plus purchased packs. Everything above total monthly credits and below the limit is overage, billed at the pay-as-you-go rate of $0.010 per credit.
Worked at HubSpot's own example
HubSpot's documentation uses an account with 3,000 included credits and 5,000 purchased. Total monthly credits are 8,000, so the default limit is 12,000. The maximum overage is 4,000 credits, which is $40.
In Customer Agent terms, at 50 credits per resolution, that headroom is 80 extra resolved conversations a month. A support team with a busy week after a product release could burn through it in a day, and then the agent pauses.
Worked at a larger account
Now take an account that buys 100,000 credits a month in packs, which costs $1,000 at the published pack price, on top of 3,000 included. Total monthly credits are 103,000 and the default limit is 154,500. The headroom is 51,500 credits, or $515.
These volumes are my illustration, not HubSpot data. The proportions are not. In both accounts, the headroom is exactly one third of the ceiling, because 50 is one third of 150. Your worst month is always your normal month plus half again, and no more.
This is the real gain. Before caps, an account on auto-upgrade that exceeded its limit got the next capacity pack added for the rest of its commitment term. A bad month could raise every remaining month. Under pay-as-you-go with a cap, a bad month costs at most half again and resets.
Why a spend cap is not a forecast
A forecast tells you what you will spend next month. A cap tells you the most you can spend. HubSpot has shipped the second one.
The forecasting problem has not moved. Credit usage still depends on inbound volume, how many workflows fire, how many leads the Prospecting Agent recommends, and how often a colleague tests a new agent. Building that prediction from your own consumption logs is covered in the comparison of AI spend forecasting tools, and nothing in the 15 October release replaces it.
The cap moves the risk from cost to coverage
When an account reaches its limit, HubSpot's documentation says "features that use HubSpot Credits will pause until the next reset date." The bill stops. So does the work.
For a marketing workflow, a pause is an inconvenience. For a Customer Agent answering live chat, it is an outage your customers see. If your cap is hit on day 20, the agent is off for 10 days. The pages I read do not say what a paused Customer Agent does with an inbound conversation, whether it routes to a human queue or simply goes silent. That is the first thing I would test, in a sandbox, before relying on the cap.
Caps convert an unbounded cost risk into a bounded cost risk plus a capacity risk. That is a good trade for most teams. It is not the same thing as predictability, and a buyer who treats it as such will learn the difference on the busiest day of the quarter.
The most useful control in the release is the least promoted one. Feature-level limits let you split the shared credit pool into budgets per team. Until now, marketing's workflow tests and service's resolutions drew from one balance. Per-feature limits fix that directly. I would set them before touching the account cap.
Seat-to-credit swaps: what to ask before you move budget
The swap is the control with the most commercial weight and the least published detail. HubSpot says customers can swap seats for credits "themselves without waiting for their contract to renew." It does not say how many credits a seat buys.
That missing number is the whole deal. A seat is a durable asset for the rest of your term. Credits, by HubSpot's own terms, expire at the end of each month and do not roll over. Swapping a seat for credits converts something you keep into something that evaporates if unused.
That is not a reason to refuse the swap. Paying for a seat nobody logs into is pure waste. The broader trend of agents replacing seats is traced in the analysis of seat compression across SaaS pricing, and the swap is HubSpot accepting that trend inside the contract rather than at renewal.
Before you swap, get three answers in writing. First, the conversion rate from one seat of each type into credits. Second, whether the swap can be reversed later in the term if you hire. Third, whether swapped credits count toward the total monthly credits that set your 150% cap, which would raise the ceiling as well as the base. I could not find any of the three in HubSpot's published material. Other terms to pin down are in the guide to AI contract clauses for finance teams.
What the spend caps cost HubSpot
Capping customers' spending costs HubSpot three things.
The auto-upgrade ratchet. Under auto-upgrade, one month over the limit added a capacity pack for the rest of the commitment term, and downgrades were possible only at the end of the term. That was expansion revenue without a sales conversation. Making pay-as-you-go the default on 16 September, then capping it, turns a permanent step up into a one-month overage at most.
Plan-level upgrades on growth. HubSpot now says it "won't upgrade a business into a bigger plan just because its usage is growing," including contact lists. That removes another automatic expansion path, at a time when net revenue retention had already slipped to 102%.
Committed seat revenue. Seats are the most predictable revenue a SaaS company books. Every seat swapped into credits moves HubSpot's own revenue onto a less predictable meter. HubSpot now has its customers' forecasting problem on its own income statement.
Set against that, the full-year revenue guide moved only from a $3.704 billion midpoint in May to $3.682 billion in August, a cut of about 0.6%. The 19% share fall was not about revenue this year. It was about how many new customers would commit to a variable bill. If a ceiling persuades even a fraction of hesitant buyers to activate agents, the overage HubSpot forgoes is small by comparison. The same tension between hybrid pricing and retention is measured in the analysis of the hybrid pricing NRR gap.
How HubSpot's controls compare with Intercom and Salesforce
Other vendors offer usage limits too. The useful comparison is what happens at the limit.
| Vendor | Alerts | Hard stop | What happens at the stop |
|---|---|---|---|
| HubSpot credits | 75%, 85%, 90% and on exceeding | Yes, default 150% once packs are bought | Credit-using features pause until the reset date. |
| Intercom Fin | 50%, 80% and 100% of a threshold you set | Yes, optional hard limit on outcomes | New conversations go to configured handover steps; live ones hand over on the next reply. |
| Salesforce Agentforce | Configurable threshold alerts in Digital Wallet | Not described on the Digital Wallet page | Not stated on that page. |
HubSpot and Intercom rows come from their help documentation. The Salesforce row reflects only its public Digital Wallet page, which describes monitoring and alerts. Salesforce contract terms may say more.
Intercom's design answers the question HubSpot's documentation leaves open. A Fin hard limit routes new conversations straight to your handover steps, so customers reach a human rather than silence. It also says plainly that a limit cannot be set below the contracted amount. On what a customer sees when the cap bites, Intercom's documentation is ahead of HubSpot's today.
HubSpot's advantage is the default. Intercom's hard limit is something you configure. HubSpot's account limit is on as soon as you buy packs. For most buyers, who never open billing settings, the default is the only control that exists.
Where this argument is weakest
Four places, and the first is the one I am least sure of.
I have not observed a paused Customer Agent. My claim that a cap creates a service outage rests on HubSpot's general statement that credit-using features pause. HubSpot may route paused agent conversations to a human inbox, which would make the coverage risk much smaller than I describe. Test it before you accept my framing.
Second, the announcement gives no rollout dates, regions or eligibility. The billing documentation predates it, last updated on 16 September. Some accounts may see different defaults, and older contracts may behave differently. Third, I attribute both share falls partly to pricing. HubSpot's chief executive and analysts named pricing, but guidance, conversion rates and downgrades all moved together, and no source isolates the pricing effect.
The case that a ceiling is all most buyers need
The strongest counter-argument is psychological. Most buyers do not want a forecast. They want to stop being afraid of the bill. A ceiling of half again, with alerts at 75%, removes the fear of the unbounded invoice, which is the thing that stops people switching agents on at all.
On that view, the cap does not need to be hit to work. It works the moment a finance lead reads it and approves the pilot. HubSpot's own data point supports this: the complaint Rangan relayed was about predictability of cost, not accuracy of a model. I agree with most of that. I still think a buyer running a customer-facing agent should know what the pause looks like before trusting it.
Frequently asked questions
What is the default HubSpot AI spend cap?
HubSpot sets an account-level credit limit up to 50% above your total monthly credits, meaning included credits plus purchased packs. Its billing documentation says this limit is set automatically after you buy additional credit capacity packs. On HubSpot's own example of 3,000 included and 5,000 purchased credits, the default limit is 12,000 credits. You can raise or lower it at any time.
What happens when you hit your HubSpot credit limit?
HubSpot's documentation says features that use HubSpot Credits pause until the next reset date when the account reaches its spend limit. Accounts without purchased packs pause when included credits run out. Feature-level limits block usage for that one tool until the limit resets or is increased. Billing contacts and Super Admins receive alerts at 75%, 85% and 90% of credits used, and again when the limit is exceeded.
Can I swap HubSpot seats for credits mid-contract?
Yes. Since 15 October 2026, HubSpot says customers can swap seats for credits, or for another seat type, themselves without waiting for renewal. HubSpot has not published eligibility rules, a conversion rate or reversal terms. Remember that credits expire monthly and do not roll over, while a seat lasts the term. Get the conversion rate and the reversal terms in writing before you swap.
Do HubSpot credits roll over to the next month?
No. HubSpot's billing documentation says unused credits expire at the end of each usage period and do not roll over. That applies to included credits and to purchased capacity packs. It matters for seat-to-credit swaps in particular, because a swapped seat becomes a monthly allowance that is lost if unused, rather than an asset you keep for the rest of the contract term.
Is HubSpot pay-as-you-go or auto-upgrade by default?
It depends on when you bought extra credits. HubSpot's documentation says accounts that bought additional credits before 16 September 2026 default to auto-upgrade, which adds the next capacity pack for the rest of the commitment term. Accounts that bought on or after that date default to pay-as-you-go, billed monthly in arrears at $0.010 per credit. Check which setting your account carries.
Do AI spend caps make AI billing predictable?
They make the worst case predictable, not the expected case. A cap bounds a month at a known maximum, which removes the risk of an unbounded invoice. It does not tell you what next month will cost, because usage still depends on volume and workflows. When demand exceeds the cap, work pauses instead, so predictable cost can come at the price of unpredictable coverage.
Where to start: set three numbers before month end
Open account and billing settings this week and check one thing first: whether your account is on auto-upgrade or pay-as-you-go. If you bought packs before 16 September, it is probably auto-upgrade, and that single setting matters more than the new cap.
Then set feature-level limits for each team that uses credits, sized to its share of your budget, so no workflow test can drain the support agent's allowance. Last, run a sandbox test of what Customer Agent does at its limit. If the answer is silence, set your cap above your busiest month and rely on the 75% alert, not the ceiling. The question to put to every agent vendor, HubSpot included, is the one in the guide to usage-based pricing concessions: what does a bad month cost, and what stops when you hit it?
Related on this site
Credits are confusing by design across vendors. See how the units differ in the explainer on AI credits and tokens, and how another platform hedged its seats in the analysis of Workday Flex Credits.
References
- HubSpot, HubSpot gives customers more control of their AI spend with new spend caps, seat-to-credit swaps, and flex payment options, 15 October 2026. Used for the three controls, the 50% default and both Jon Dick quotes on predictability.
- HubSpot Knowledge Base, Understand HubSpot credits and billing, last updated 16 September 2026. Used for the limit mechanics, the 12,000 credit example, pause behaviour, alert thresholds, the auto-upgrade and pay-as-you-go defaults, and non-rollover.
- HubSpot, HubSpot Credits, 2026. Used for the $0.010 per credit price and the $10 per 1,000 credit pack.
- HubSpot, Customer Agent and Prospecting Agent: now you pay when the task is complete, April 2026. Used for the 14 April effective date and the old and new agent prices.
- HubSpot, Q1 2026 results, Form 8-K Exhibit 99.1, 7 May 2026. Used for Q1 revenue and the May full-year guidance.
- Tickeron, Why is HubSpot stock down 24% today?, 8 May 2026. Used for the after-hours and premarket moves after Q1. Secondary market commentary.
- The Motley Fool, HubSpot just cut its own customer growth forecast, 8 August 2026. Used for Q2 revenue, customer additions, guidance, the 19% share fall and the Rangan quote.
- Fin help centre, Manage Fin usage alerts and limits, 6 August 2026. Used for Intercom's alert thresholds and hard limit behaviour.
The weakest part of this source base: the Q1 and Q2 share moves come from market commentary rather than exchange closing data, and the May figures are after-hours and premarket readings. HubSpot's 15 October announcement gives no rollout dates or eligibility, so default behaviour is taken from billing documentation that predates it. The 1 October EMEA change rests on partner agency reporting. The larger-account example uses volumes chosen by the author.
Related reading