Hybrid Pricing Hit 95% of AI Agent Deals in 2026. Its Committed Floor Is Where the Renewal Fight Starts.

Ask a SaaS founder in 2024 whether their pricing was usage-based or subscription-based and you'd get a clean answer. Ask the same question about an AI product in 2026 and the honest answer is almost always both — a subscription-shaped floor with a usage-shaped ceiling, sold as one deal. That structure has a name inside billing teams — a committed-spend floor — and it has quietly become the default way AI-era software gets priced. It's also created a renewal problem that doesn't look like ordinary churn, because the account in question never technically stops paying.

Orb's number is specific to AI agent companies, but the pattern it describes isn't contained to that category. A committed-spend floor is the same mechanism whether it's wrapped around agent tokens, API calls, storage, or compute — a customer prepays or commits to a minimum in exchange for a lower unit rate, and the vendor gets revenue predictability without giving up the upside of a customer who scales past that minimum. It's a genuinely good deal for both sides when usage is stable or growing. The part nobody built a clean process for is what happens when usage does the opposite.

What a committed-spend floor actually locks in

Orb's report puts real numbers on how thoroughly hybrid pricing has replaced the cleaner "pick one model" pitch decks from a few years ago. Usage-based pricing as a component climbed to 91.3% of AI agent companies in 2026, up from 83.3% the year before. A subscription or committed base still shows up in 71.3% of deals — and of those, 94.7% pair it with a usage-based layer rather than running the subscription alone. Outcome-based pricing, the model a lot of AI pricing commentary in 2024 predicted would take over, sits at just 3.8%, down slightly from 4.5%.

Read those numbers together and the picture is clear: a pure usage meter with no floor underneath it is now the exception, not the norm. Vendors converged on the committed floor for a reason that has nothing to do with fairness — it converts a volatile, hard-to-forecast usage line into something closer to guaranteed revenue, while still letting them capture upside from an account that grows. Buyers accepted it because the discount attached to committing upfront is usually real, and because an unbounded, unpredictable bill is worse for their own budgeting than a fixed floor. Neither side was thinking hard about what the floor does to an account whose usage moves in the other direction, because in a market growing as fast as AI adoption was through 2025, that direction barely came up.

The clause that only runs one way

A committed-spend floor contract almost always includes a true-up provision — language that bills the customer automatically the moment usage crosses above the committed amount. It is far rarer for the same contract to include a true-down provision, a contractual right for the customer to reduce that commitment mid-term when usage drops. An analysis published on GaaS in June 2026 put the asymmetry plainly: true-up clauses "let the vendor bill you when you exceed your committed seat count," but "what you almost never find is a true-down clause, a contractual right to reduce your committed quantity mid-term." On a committed-spend floor specifically, the same piece described the effect directly — "the floor is contractual and product-agnostic. Even if every individual line item deflates, you still owe the minimum."

That's the mechanism worth sitting with. A customer whose usage shrinks under a per-seat or per-license model at least sees the shrinkage reflected somewhere — fewer seats, a smaller renewal quote, a number a procurement team can point to. A customer under a committed floor sees none of that. Their bill doesn't move. As the GaaS piece put it, "the danger isn't a smaller bill, it's a frozen one" — usage can fall by half while the invoice stays exactly where it was, which means the customer's effective cost per unit of value is quietly doubling every cycle, with nothing on the invoice itself signaling that anything has changed.

Why this doesn't show up as churn until it suddenly does

We've written before about zero-usage churn — the way a metered account can drift toward abandonment with no cancel event to log, because a shrinking invoice never forces a decision the way a fixed subscription charge does. A committed-spend floor breaks that pattern in a way that sounds like it should help, but often makes things worse. The invoice doesn't shrink at all, so there's no early warning signal for anyone watching billing data, and the customer isn't quietly drifting away unnoticed — they know exactly what they're being charged, they know their usage has dropped, and they're accumulating a specific, growing grievance about it with nowhere to raise it until the contract comes up for renewal.

That's the difference worth understanding: a pure usage account that shrinks toward zero churns silently. A committed-floor account that shrinks toward zero churns loudly, all at once, at a single predictable moment — the renewal date — because that's the only point in the contract where the floor itself is ever back on the table. Customer success teams who track engagement and login activity will often see this coming from the product side well before finance sees it, since the usage decline itself is usually visible internally even though the invoice hides it from the customer's own dashboard. The gap between those two views is exactly where the frustration builds.

2026 is when the first real renewals hit

Timing makes this worse in 2026 specifically. Bessemer Venture Partners' AI pricing research frames the moment directly: a wave of AI deals signed in 2025, while adoption enthusiasm and budget scrutiny were both unusually loose, are now reaching their first renewal, and "pricing will need to reflect actual value, not merely potential or promise." A lot of those first-year committed floors were set generously on the vendor's side specifically to land the deal — a low floor, a steep prepay discount, sometimes a floor sized around a pilot's best-case usage rather than its likely steady state. None of that got tested until the renewal date actually arrived, because a floor that's too low to matter never triggers a true-up and never gets revisited mid-term.

There's a second pressure working in the buyer's favor by the time that renewal conversation happens. Common Paper's 2026 SaaS Contract Benchmark Report, drawn from 16,140 signed agreements across 2,223 companies over the trailing twelve months, found that automatic fee-increase clauses in SaaS contracts nearly halved — from 23% of agreements in 2024 down to 13% in 2026. Vendors are already giving ground on the price-escalator clauses that used to renew quietly in the buyer's disfavor. A committed floor that hasn't moved despite a customer's usage falling is now one of the last remaining places in the contract where the vendor is still holding a line the buyer has learned, from every other clause in the same document, that vendors are willing to negotiate.

What to actually check before your next renewal cycle

• Pull committed floor versus trailing actual usage for every account, at least one quarter before renewal. The accounts sitting well under their floor are exactly the ones where the renewal conversation is going to be adversarial if you show up with the old number unchanged.
• Separate "floor renewed flat despite lower usage" from a normal renewal in your reporting. This is a contraction event even though nothing about the account technically churned, in the same way a downgrade is a revenue loss that never touches your cancellation flow.
• Bring account owners the usage-versus-floor gap before the renewal call, not during it. The same principle applies here that we've written about for negotiated contract renewals generally — a rep who doesn't know the gap exists is negotiating from a worse position than a customer who's been staring at their own invoice for a year.
• Consider offering a true-down option as a retention lever, not a concession you're forced into. A floor reduction tied to a longer renewal term or a small rate increase elsewhere often keeps more total revenue than holding a floor that triggers a full non-renewal instead.

None of this means committed-spend floors are a bad pricing structure — they're not, and the growth numbers behind hybrid pricing's rise say the model is working for most accounts most of the time. It's a reason to treat the floor as a number that needs re-checking on a schedule, the same way you'd re-check a seat count, instead of a term you set once at signing and never revisit until a customer brings it up first. If your product runs any committed usage floor and a customer's real usage has already told you the renewal is going to be a fight, that's exactly the kind of context a cancellation or downgrade page from CancelFlow needs before the customer ever gets there — the honest move for an account already sitting well under its floor is a proactive right-sizing conversation, not a renewal notice that reads like nothing changed. Our dollar retention rate calculator is a fast way to see what a handful of flat-floor renewals against declining usage are actually doing to your net revenue retention once you pull them out of the blended number.

Frequently asked questions

Stop losing subscribers today

One script tag. One function call. A live cancellation flow in under 10 minutes.