The Sub-5% Club: The Terminal State of Software

For twenty years, the slow lane in B2B wasn’t great. But it was an OK place to be. Growth faded, but the annuity was forever. High NRR, high margins, sticky contracts, structured data that was stuck in an app and couldn’t leave, predictable cash. You stopped being a growth stock and become a bond with a little upside. That was the deal, and it was a good one. Private Equity bought up many names here, like clockwork.

That deal is muchly dead. A whole cohort of public software companies is now growing under 5%, and the market has stopped pricing them as annuities. Dropbox, Zoom, DocuSign, PagerDuty: all stuck in single digits or worse, all trading like the cash they throw off might not be durable. A terminal state. Not slow growth. The end of growth, with a question mark over whether what’s left is even safe.

The annuity was built on seats, and AI eats seats, makes the data far more portable to the competition, and has raised the bar dramatically in almost every category.

What sub-5% actually looks like right now

Real names, real numbers, most recent quarter.

Dropbox and PagerDuty are the cleanest examples, and they sit at the bottom for different reasons.

Dropbox grew 0.8% last quarter. Strip out a product they’re winding down and it was 2%. Total ARR grew 0.3%, and management’s own full-year guidance calls for revenue todeclinesomewhere between 0.4% and 0.9%. A $2.5B revenue business with 80% gross margins, throwing off over a billion in free cash flow, and the forward number has a minus sign in front of it.PagerDuty is arguably further gone. Revenue grew 1.0%, ARR is dead flat at $496M, and net dollar retention has fallen to 97%. That last number is the one that matters most, because below 100 means the existing base is shrinking and new logos are just filling the hole. Dropbox is flat with the cash still intact. PagerDuty is flat with retention already cracking. Both are the terminal state, one a step ahead of the other.Zoom is sitting right on the line. 5.5% in the most recent quarter, 4.4% for the full prior fiscal year. Online revenue, the self-serve base, grew 2.8%. Enterprise net dollar expansion is 99%. When NDR drops below 100, your existing customers are spending less with you each year, and you are covering it with new logos.DocuSign at 8.7% looks healthier until you see the slope: a 15% five-year average grinding down toward a high-single-digit guide. The direction matters more than the level.

Under 5% is a different category, not just a slower one

At sub-5% growth, the math of a software valuation changes character.

Roughly 85% to 95% of the enterprise value in a software DCF lives in the terminal value. When you are growing 30%, nobody is debating the terminal value, the near-term growth carries the story. When you are growing 4%, the terminal value is the story. The market stops asking how fast you are growing and asks one question only: is this annuity durable?

For two decades the answer was an automatic yes. Switching costs, 10 to 20 years of workflow lock-in, contractual recurring revenue. You paid a premium to the S&P 500 for those economics because they were more reliable than almost anything else you could own.

That premium is gone. As of early 2026, software trades at roughly 22.7x forward earnings, below the S&P 500 for the first time ever. IGV, the software ETF, is down around 30% from its September 2025 peak. About $2 trillion in market cap, erased. This is not 2008, where the underlying economics stayed intact and software bounced back fast. The market is not saying software is temporarily expensive. It is saying it is no longer sure the annuity is an annuity.

The mechanism: seats were the annuity, and agents eat seats

Follow the seats.

The seat-based model assumes headcount grows, or at least holds. You land, you expand, you add seats as the customer adds people. Net revenue retention above 110% was the engine of the entire era.

AI agents attack that engine directly. If an agent does the work of ten support reps, the customer does not need ten seats. They need one human and an agent. Expansion stops being a tailwind and becomes a headwind. The model does not just slow, it can reverse. Dropbox guiding to a revenue decline is what reversal looks like in slow motion.

And the budget is physically moving. Anthropic crossed $19B in annualized run rate this year, up from $9B at the end of 2025. Roughly 75% of new hyperscaler infrastructure spend in 2026, over $450 billion, is targeting AI. That money has a source. It used to buy CRM seats, ITSM modules, storage tiers. CIOs have finite budgets. Every dollar going to agents is a dollar not expanding a seat contract.

The bifurcation is the real story

It is not all down. The cohort is splitting in two.

Infrastructure for AI is re-accelerating. Cloudflare guided to 28-29%. Snowflake printed 30% product growth. Twilio just put up 20%, its fastest in over three years, on AI voice workloads. DigitalOcean is guiding to 19-23%. These companies get paid more when AI usage goes up, because agents consume compute, storage, bandwidth, and data.

Applications built on seats are the ones drifting toward terminal state. HubSpot’s customer count growth tells the story by itself: 21%, then 19%, 18%, 17%, 16% over five straight quarters. Still a great company. But the direction is one way.

There is no longer one category called software. There is infrastructure that AI feeds, and there are applications that AI feeds on.

There is no more graceful decline in B2B software. And “getting profitable” is just not enough, in almost every case

First, net revenue retention is the whole ballgame now, and theSeat-driven expansion is exposed. Usage-driven and outcome-driven expansion can survive and even re-accelerate, because when AI usage rises your revenue rises with it. If your expansion math depends on your customer hiring more people, you have a structural problem, not a sales problem.sourceof your expansion is what gets repriced.Second, sub-10% growth is not a resting state anymore. The old playbook said you could coast into a mature annuity and milk it for a decade. That coast assumed the moat held passively. It does not. A flat business with an AI gun pointed at its expansion motion is a declining business that has not declined yet.Third, the cash-return endgame is real, but it is a different game. Dropbox trades around 6x forward free cash flow with roughly a 17% free cash flow yield, buying back stock aggressively. Zoom is generating $1.7B in free cash flow and just added another billion to its buyback. These are not growth stocks. They are cash-return vehicles. That can be a perfectly good investment if the annuity holds. It is a value trap if AI turns flat into decline. The market is currently betting trap, which is the only reason a 17% FCF yield exists at all.Fourth, PE isn’t buying the slow growers anymore. At least not very often. PE has almost fully retreated from buying slower growth B2B / SaaS startups.