The average enterprise ran 305 SaaS applications in 2026, almost exactly the same number it ran the year before. The portfolio shrank by 0.07%, a rounding error. Yet average annual SaaS spend climbed 8% to $55.7 million, according to Zylo’s 2026 SaaS Management Index, which tracks more than 40 million licenses and $100 billion in software, AI, and cloud spend. Those two numbers do not fit the story most finance teams tell about software cost. The bill went up while the app count stood still. That means the thing driving your SaaS spend in 2026 is no longer the number of tools you own. It is the pricing model underneath the tools you already have.
The sprawl era quietly ended
For most of the last decade, SaaS cost control was a counting exercise. Find the duplicate project trackers. Reclaim the 200 unused Salesforce seats. Kill the app three people logged into last quarter. That playbook worked because sprawl was real and growing, and because the waste sat in plain sight once you had a discovery tool pointed at it.
The counting is mostly done. Zylo’s data shows portfolios have flattened, with the average holding steady around 305 applications and the median at 240, which the report frames as “stabilization rather than continued sprawl.” License utilization rose to 54% from 47% the prior year, and average annual license waste fell to $19.8 million from $20.9 million, per Zylo’s SaaS statistics. Teams got better at the seat game. Underneath the stable app count sits real churn, roughly a third of the portfolio turning over in a year, so the stability is a balance of tools added and tools cut, not a frozen estate.
Here is the problem. If utilization is up, waste is down, and the app count is flat, the old playbook says the bill should be flat too. It rose 8% instead. The savings you earned by reclaiming seats got swallowed by something the seat-counting tools were never built to see.
The money moved into the pricing model
What changed is how vendors charge, not how many vendors you have. Three shifts show up in the 2026 data, and all three bypass the seat count.
First, renewals became the main cost event, and the main place prices move. Renewals account for 87% of total software spend, the average organization processes 211 of them a year, and 79% of IT leaders reported a price increase at renewal, according to Zylo. A renewal is not a formality. It is the moment a vendor repackages the same product at a higher number, and at 211 per year, most of them clear without a negotiation because nobody has time to fight 211 battles.
Second, consumption and AI feature charges now land outside the subscription you budgeted. Seventy-eight percent of IT leaders reported unexpected charges tied to AI or consumption, and 61% halted a project because of unplanned SaaS expenses, per Zylo’s index. AI-native app spend grew 108% year over year, and at the largest companies it grew 393% in a single year. This is the outcome-based and usage-based billing shift arriving inside the SaaS stack: a vendor adds an AI assistant metered per action, and the line item appears after you have already approved the renewal.
Third, the seat is losing its grip as the unit of pricing. Pure per-seat pricing fell from 21% to 15% of products in twelve months, while roughly 42% of products now offer a usage-based option, up from 27% in 2023, per SaaS pricing tracking data. The dominant enterprise shape is now hybrid: a fixed seat license for access, plus a metered layer that prices the AI-driven work on top. You can hold your seat count perfectly flat and still watch the metered layer compound.
Why your SaaS dashboard can’t see it
Most SaaS management tooling was built to answer “which apps, how many seats, who logged in.” That is a license question, and license questions have clean answers that map to a contract. Consumption does not live in the contract. It lives in each vendor’s own usage meter, denominated in whatever unit that vendor picked: messages, resolutions, tokens, API calls, documents processed. None of those roll up to a seat, and none of them show up in a seat-reclamation report.
The visibility gap moved accordingly. It used to be “we don’t know which apps employees signed up for.” It is now “we don’t know how much each app is metering.” Zylo found 60% of IT leaders lack sufficient visibility into AI tool usage, and 77% discovered AI tools running without IT awareness. That is the same blind spot behind shadow AI spend, seen from the SaaS side instead of the cloud side. When the FinOps Foundation reports that 98% of organizations now manage AI spend, up from 63% two years earlier, this is a large part of what they are managing: metered charges buried inside tools finance already thinks it has under contract.
Microsoft made the pattern concrete when it turned on Copilot’s usage meter by default. A seat-based product you already pay for sprouts a consumption meter, and a budget alert on that meter warns you without stopping the spend. Multiply that across a 305-app portfolio where 92% of vendors have shipped or plan to ship AI features, and the metered layer stops being an exception. It becomes the growth engine of the bill.
What the 2026 renewal cycle actually requires
If 87% of spend flows through renewals and the price increases hide there, the renewal calendar is the control surface. Three moves matter more than another seat audit.
Model the metered layer as its own line, with its own owner and its own cap. Treat consumption and AI charges the way a cloud team treats variable compute: forecast it, tag it to a team, alert on it, and name someone accountable when it runs hot. A flat subscription and a metered layer are two different financial instruments stapled to one invoice, and only one of them behaves.
Bring usage data to the renewal, not just seat counts. The question shifted from “are people using it” to “what did this tool actually produce,” which is the same cost-per-outcome discipline finance now applies to AI. A vendor asking for a 12% increase on a tool whose metered usage you can prove is flat has a weak hand, if you show up with the numbers.
Negotiate the price increase as the default, not the exception. When 79% of renewals carry an increase, the increase is the market, and benchmark data is the counterweight. Running a fractional COO engagement through Ops Harmony, I watched a mid-market client approve 40-plus renewals in a quarter on autopilot, each a single-digit bump that looked trivial in isolation; stacked across the portfolio they outweighed every seat the team had spent the quarter reclaiming. The seats were visible and small. The renewals were invisible and large.
The counting era of SaaS cost control is over because the counting got easy and the tools got good. The bill kept rising anyway, which tells you the leak moved. In 2026 it is not in how many tools you bought. It is in how each of them decided to charge you after you signed.