--- slug: vibe-revenue type: antipattern summary: "Reading an AI startup's fast-growing run-rate as durable demand when much of it is trial budget that has not yet become software the customer depends on." created: 2026-05-29 updated: 2026-06-06 related: ai-wrapper-trap: relation: related note: "A thin wrapper is one common source of the churned, non-durable revenue this trap mistakes for durable demand; both are AI-era failures hiding behind real early traction." data-moat: relation: prevented-by note: "The proprietary-data advantage that makes a product hard to leave is the most reliable way experimental revenue becomes durable." defensibility: relation: related note: "Durable revenue and durable advantage are the same question asked from two sides: revenue stays when the product cannot easily be left." product-market-fit: relation: violates note: "Experimental adoption that does not retain is not the durable, repeated demand product-market fit describes; the trap reads a trial as a fit." false-positive-trap: relation: contrasts-with note: "Both mistake a misleading early signal for proof; the false positive is a narrow segment read as a broad market, while vibe revenue is trial spend read as durable demand." unit-economics: relation: related note: "Unit economics built on lifetime-value assumptions break when the revenue churns at the rate experimental AI adoption does, so the metrics flatter a business that is not there." cac-ltv-ratio: relation: related note: "Lifetime value is computed from a retention rate; when the retention is experimental rather than durable, a healthy-looking CAC/LTV ratio is measuring a customer who is already leaving." burn-multiple: relation: related note: "A burn multiple looks efficient when net new ARR is growing fast, even when that ARR is trial budget that will not renew, so the trap can make a fragile company read as a capital-efficient one." --- # Vibe Revenue > **Antipattern** > > A recurring trap that causes harm — learn to recognize and escape it. *Treating an AI startup's fast-climbing run-rate as durable demand when much of it is experimental budget: money customers are spending to try the tool, not to depend on it.* > **📝 Where the name comes from** > > "Vibe revenue" is a 2025 coinage, popularized by CNBC in November of that year, describing companies valued on "noise and vibe revenue" rather than durable demand. It rhymes deliberately with "vibe coding," the practice of building software by prompting rather than writing it: both name a thing that feels real and produces a real-looking artifact, while leaving open whether anything operationally necessary sits underneath. The seed-stage version of the same concern got a drier name a year earlier: "quality of revenue," the diligence question of whether a dollar of recognized revenue will still be there next year. A startup books real money. Enterprises sign real contracts, the recognized revenue is genuine, and the run-rate chart rises faster than traditional SaaS usually did. The number is not fake. The trap is what the number is taken to mean. For much of the 2025 AI cohort, a large share of that revenue is *experimental*: budget a customer allocated to learn whether an AI tool works before committing a core process to it. That revenue looks like the early arc of a durable software business and gets priced like one. It doesn't behave like the recurring revenue the valuation assumes. The question that now gates AI-startup diligence is no longer "how fast is revenue growing?" It is "how much of this revenue will renew when the experiment ends?" ## Symptoms The trap is hard to see because the headline metric, revenue growing fast, is the one founders and investors trust most. The tell isn't weak revenue. It's a gap between the run-rate and the durability underneath it. Watch for these together: - **Annual recurring revenue (ARR) is quoted as a run-rate, multiplied up from a single strong month.** The pitch says "$5M ARR" on the back of a $400K month that was itself half new logos. Run-rate ARR annualizes a moment; it tells you the speed of acquisition, not the durability of what was acquired. - **Gross revenue retention is far below the software norm.** Gross revenue retention (GRR) measures how much of last year's revenue you keep before any expansion. It is the floor under the business. Durable software holds GRR in the 85-90%+ range. Across the 2025 AI cohort, many companies run materially below that, with monthly logo churn at roughly double the 5-7% that healthy software historically saw. - **The largest customers are running pilots, not deployments.** The accounts driving the curve are in a trial, a proof-of-concept, or a single team's discretionary budget, not embedded in a workflow the company has reorganized around. The contract is annual; the commitment is a quarter. - **Net revenue retention is propped up by a few expanding accounts.** Net revenue retention (NRR) counts expansion as well as churn, so a handful of accounts doubling their seats can hide a wide base quietly leaving. A strong NRR with a weak GRR is a base that is churning under a few winners. - **A surprising share of revenue traces back to other AI startups.** When the customer list is heavy with companies that are themselves spending venture or cloud-credit money to try AI, the revenue is partly circular: it lasts exactly as long as the funding cycle that is paying for it. A single one of these is survivable. The cluster, in a company raising on its growth rate, is the trap. ## Why It Happens Vibe revenue is not a con. It is the predictable result of a genuinely new buying behavior meeting a metric vocabulary built for a different one. Enterprises are buying AI to learn, not yet to depend. The 2025 enterprise posture was experimentation at scale: most large companies ran many small AI trials at once, funding them out of innovation budgets to find out what worked before committing a core process to any of them. That produces a flood of real contracts whose purpose is evaluation. The vendor sees signed revenue; the customer sees an experiment with a renewal decision pending. Both are right, and only one of them has priced in the renewal risk. The metrics were built for software that does not churn like this. ARR, NRR, and lifetime value all assume that a customer who signs is a customer who stays, because for classic subscription software that assumption mostly held. Applied to experimental AI adoption, the same metrics flatter the business: a high run-rate, a respectable NRR carried by a few expanders, a [CAC/LTV ratio](cac-ltv-ratio.md) computed from a lifetime that assumes a retention the cohort is not delivering. The dashboard is not lying. It is answering the wrong question, because its definitions predate the buying behavior it is now measuring. Capital rewards the run-rate over the retention. Through 2025, the fastest-growing AI revenue raised the largest rounds at the highest multiples, so a founder is pulled to lead with the growth rate and an investor pattern-matching on "fastest to $10M ARR ever" is pulled to underwrite it. The slower, less flattering questions (what is GRR, how much of this is pilots, how much is circular) are the ones that get asked last, if the round is competitive enough to skip diligence. The story that the revenue is durable carries the valuation long before anyone has the cohort data to confirm it. ## The Harm The damage arrives on a delay, which is what makes it dangerous: the costs are committed during the good months and land during the bad ones. The company spends against revenue that won't renew. A run-rate read as durable sizes everything downstream: the hiring plan, the burn, the next raise. When the pilots conclude and a chunk of the base does not convert to dependence, the company is built for a revenue level it no longer has, and the [unit economics](unit-economics.md) that looked sound were computed on a lifetime the customers never delivered. The [burn multiple](burn-multiple.md) that read as efficient was efficient against ARR that evaporated. The next round reprices on retention, not growth. By 2026 the diligence question had moved from speed to durability, and a company that raised on run-rate faces a down round or a closed door when the GRR comes out. The "quality of revenue" frame that seed investors named in 2024 became the Series A and B gate: investors learned to discount experimental ARR, so the founder who maximized the headline number finds it counts for less exactly when the next dollar is hardest to raise. Revenue that was overvalued going in is correspondingly punished coming out. The macro version compounds the company-level one. When a large share of an entire cohort's revenue is experimental, and some of it is circular (AI startups paying AI startups, much of it on cloud credits and venture funding), a contraction in the funding cycle pulls demand out of the whole segment at once. Analysts named the risk of a "gross retention apocalypse": the moment experimental budgets are cut across the board and a class of companies discovers simultaneously that its recurring revenue was never recurring. A founder who mistook vibe revenue for durable revenue is most exposed precisely when the broader correction makes capital scarcest. ## The Way Out The exit is not to dismiss AI revenue as fake. Much of it is real, durable, and the leading edge of a genuine shift in enterprise spending: the 2025 data shows enterprises moving real budget to AI applications and rewarding the ones that prove out. The discipline is to separate the durable share from the experimental share, to report and price the business on the durable share, and to spend the experimental window converting trials into dependence before it closes. Measure gross retention, and lead with it. Report GRR alongside NRR and run-rate, and treat it as the honest floor of the business. A founder who knows their gross retention knows how much of their revenue is real next year; a founder who quotes only run-rate and NRR is, often without meaning to, hiding the churn under the expanders. The same number is the cheapest diligence an investor can run: ask for GRR by cohort, and the experimental revenue separates itself from the durable revenue on the page. Convert the trial into a deployment the customer cannot easily leave. The experimental window is an opportunity, not a verdict. Use it to embed: accumulate a [data moat](data-moat.md) from the customer's own usage, integrate into the systems their work already runs through, and become the place a process happens rather than a tool a team is evaluating beside three others. Revenue retains when leaving is expensive, which is the same property that makes a product [defensible](defensibility.md). The pilot that becomes operating infrastructure is durable revenue; the pilot that stays a pilot was always vibe revenue, whatever the contract said. > **💡 Tip** > > Split your revenue into two lines: customers who would feel real operational pain if you shut off tomorrow, and customers who are still deciding whether they need you. Manage and report against the first line, and size the burn and the raise to it. The gap between the two lines is the revenue you don't actually have yet. The work for the quarter is to move accounts from the second line to the first before their pilot budget renews. If the honest split shows almost everything sitting in the experimental line with no path to dependence, the disciplined move is the same one the wrapper trap demands: treat the business as smaller and more fragile than the run-rate implies, raise and spend accordingly, and either build the durable asset that earns the renewal or accept the real size of what is there. ## How It Plays Out A mid-stage AI startup sells an enterprise assistant and crosses $8M run-rate ARR in under a year, faster than traditional SaaS usually reached that mark. The customer list is a roster of recognizable logos. What the run-rate hides is the shape of the contracts: most are one-year deals funded from innovation budgets, sold to a single team that wanted to find out whether the tool delivered. Twelve months on, the renewal conversations split. The accounts that wired the assistant into a daily workflow, where the team's prompts, corrections, and context now live inside the product, renew and expand. The accounts that ran it as a side experiment, where the assistant sat beside the existing tools rather than replacing them, conclude the lift was marginal and let the contract lapse. Gross retention comes in well under the software norm. The company raised and hired against $8M; it renews around half of it, and the Series A it sized for the headline number reprices on the cohort data instead. The quieter version never reaches a renewal cycle. A founder watches net revenue retention hold above 100% and reads it as proof the base is sticky, not noticing that a few fast-expanding accounts are masking a wide bottom of pilots quietly churning out. Some of those pilots are other early-stage AI companies spending their own seed money to try the tool: revenue that lasts exactly as long as the customer's runway. When the broader funding climate tightens, the experimental budgets that were the demand get cut first across the whole segment at once, and the run-rate that looked like a business turns out to have been a measurement of how much venture capital was flowing through the category that year. ## Sources - CNBC's November 2025 reporting popularized the term "vibe revenue" for AI companies valued on noise and run-rate rather than durable demand, and documented founders' own worry about a bubble; read as a marker of when the concept entered the working vocabulary. - The "quality of revenue" frame emerged from seed-stage venture writing in late 2024 as the diligence question of whether recognized AI revenue would still exist a year out; it is the drier, earlier name for the same concern. - ChartMogul's SaaS retention research documents the gap between durable-software retention and the materially higher churn observed across the 2025 AI cohort, the empirical basis for the "gross retention" warning. - Commentary tying the dynamic to circular financing, with AI startups funded by venture and cloud credits spending on other AI startups, frames the macro "gross retention apocalypse" risk in which experimental budgets contract across a segment at once. Treat the segment-level claims as directional, since the underlying behavior is still moving. --- - [Next: Data Moat](data-moat.md) - [Previous: The AI Wrapper Trap](ai-wrapper-trap.md)