The AI Pivot — Can Credits Break the Seat Ceiling Without Cracking the Margin?

The AI Pivot — Can Credits Break the Seat Ceiling Without Cracking the Margin?

The whole report has circled one unanswered question. Chapter 2 named the seat ceiling — the risk that AI does the work a customer would once have hired (and seated) a person to do, so monday.com's per-seat revenue stops expanding even as usage climbs. Chapter 5 priced a scenario table whose every line above the bear case depended on AI monetization, then admitted it could not resolve it. This chapter resolves what the evidence allows: the seats-plus-credits pivot is a single lever aimed at two problems at once — escaping the seat ceiling and paying for the compute that AI consumes. The cost side of that bet is now certain and visible in the filings; the revenue side is, by management's own words, unmodeled. That asymmetry — a guaranteed margin headwind in 2026 against an uncertain revenue payoff later — is the swing factor the price has been arguing about.

How monday.com now charges: a hybrid, not a switch

monday.com did not abandon seats. It bolted a usage meter onto them. For new customers from 2026, billing runs on two vectors of expansion — seats, plus AI credits that accrue with consumption [1]. Existing customers are not forced across; they are offered an opt-in migration management expects to run over the next couple of years, with enterprise accounts handed complimentary AI packages to seed adoption [2].

Underneath, the platform splits AI into two pricing logics. Foundational, embedded capability — the "AI Blocks" woven into existing workflows — stays seat-based, because customers told the company they value the predictability of per-product-unit pricing and want to consume that capability inside their subscription [3]. The heavier, model-driven work — agents, the Vibe app builder, compute-intensive outputs — is metered through credits [4]. The company laid the full map out at its September 2025 Investor Day.

Source: monday.com Investor Day 2025, "A hybrid monetization model" [5]; credit vs seat-based split confirmed on the Q4 FY2025 call [6].

The logic of the credit vector is precisely the seat-ceiling escape Chapter 2 was waiting for. As management put it, when an AI agent "takes on more work across organizations, revenue expands naturally without requiring additional seat purchases" [7]. The seat was always a proxy for a human doing work; once the work is done by software, the company needs a new unit to bill — and consumption is that unit.

The certain cost: the first crack in a 90% gross margin

For a decade, monday.com's gross margin was the quiet luxury of the model — revenue arrived as software, and serving it cost almost nothing. The FY2025 20-F now rewrites that line. Cost of revenue, it discloses, includes "hosting and cloud infrastructure fees (including costs associated with AI compute, model usage and data processing)" — and the company states plainly that "as adoption of our AI-powered offerings increases, we expect associated infrastructure and compute costs to increase. As a result, we expect our gross margin to decline modestly in the mid-term" [8].

Management has put a number on "modestly." At the September 2025 Investor Day it guided that "incremental AI-related compute costs will moderate gross margins to mid-80%" [9]. On the Q4 FY2025 call the CFO was blunter: "mid-80s to high 80s, and we used to have 90%" [10]. The first installment already shows: Q1 FY2026 non-GAAP gross margin printed 89%, down from 90% a year earlier, explicitly attributed to AI [11].

FY2021–FY2025 non-GAAP gross margin as reported; FY2026E plotted at ~85.5% to represent the guided "mid-to-high 80s" band. Sources: Investor Day 2025 [12]; Q4 FY2025 transcript [13]; FY2025 20-F [14].

The dollars are not trivial. In FY2025, cost of revenue was $133.1 million on $1.232 billion of revenue — up 28%, a hair faster than revenue's 27%, but the AI bite had barely begun (hosting costs rose just $1.9 million that year) [15]. The guided shift is larger. A move from 90% to the mid-80s is roughly four-to-five points of gross margin; on a FY2026 revenue base near $1.5 billion, that is on the order of $60–75 million of incremental compute cost a year — a sum comparable to the entire treasury-interest line that Chapter 4 showed flatters net income. This is not a rounding error in the cash story; it is a genuine, recurring claim on the P&L that did not exist two years ago.

Why credits are the margin defense, not just the growth lever

The elegance of the consumption meter is that it attacks the cost it creates. An analyst asked the question directly on the Q1 FY2026 call: does adding usage-based elements "more effectively match your revenue with your costs tied to AI so that we could see better gross margin preservation over time?" [16]. That is the design intent. Seat-priced AI is a fixed-revenue product with a variable, usage-driven cost — the worst possible shape for a margin. Credit-priced AI moves the cost and the revenue onto the same axis: the customer who runs the expensive workload is the customer who pays for it.

Whether monday.com can hold that line depends on customer tolerance for a meter, and here the early signal is mixed. The credit table carries "a lot of parameters depending on what type of usage there is," and as one analyst noted, "even at $0.01 per credit… that could balloon for customers" [17]. Management's answer is governance: dashboards that let administrators see who is spending credits, on what, so usage can be scoped and budgeted rather than feared. The competitive overhang Chapter 3 documented — that Microsoft, Google and the suite giants can bundle AI into prices monday.com cannot match — bears directly here: a meter only holds if the metered capability is differentiated enough that customers will not simply take the bundled-for-free alternative. monday.com is betting its Work-OS context and data graph make its AI worth paying a meter for. That bet is unproven.

The unproven half: does revenue actually re-accelerate?

Set against the certain cost is a revenue contribution that is real but small, and a forward path management explicitly refuses to quantify. In Q1 FY2026, AI drove approximately 10% of net new ARR — encouraging, and the first quarter the company disclosed the figure [18]. But the company is scrupulous about what that counts: "when we say AI contribution, we mean direct contribution, not contribution made possible by AI," and that figure runs almost entirely on the existing offering — the embedded AI Blocks, Sidekick, Vibe — because agents had been live for about a week and "have not yet meaningfully contributed" [19].

Sources: Q1 FY2026 transcript — AI net new ARR [20], gross margin [21]; new-products ARR share, Q3 FY2025 transcript [22].

The adoption telemetry is genuinely fast. Vibe, the AI app builder, became the fastest product in monday.com's history to pass $1 million of ARR, and the platform's AI Blocks powered more than 77 million actions in the quarter while Sidekick processed over half a million user messages [23]. By Q3 FY2025 customers had built more than 60,000 apps on Vibe since its July launch, and new products as a whole had already crossed 10% of total ARR ahead of schedule [24].

But adoption is not yet revenue, and management will not pretend otherwise. On the most important question — what the seats-plus-credits model does to the growth rate — the company provided no quantitative projection, saying it needs additional quarters of adoption data before it can update guidance [25]. The CFO was candid that the company still does not "know how to model and expect revenue coming from agents and token-based usage." That honesty is to its credit. It also means the single most consequential variable in the bull case is, for now, a hope rather than a forecast.

What it means for the thesis

Place the two halves side by side and the shape of the bet is clear.

Source: synthesized from the FY2025 20-F and FY2025–FY2026 earnings calls cited throughout this chapter.

This is what makes the security so asymmetric in time, not just in payoff. The margin cost lands first and on a schedule — FY2026, in the P&L, already visible at 89%. The revenue reward, if it comes, arrives later and on no schedule management is willing to commit to. A bear can point to a real, dated erosion of the company's best financial attribute and a monetization story that is still 90% adoption metrics. A bull can answer that monday.com has built exactly the pricing architecture the AI era demands — usage-aligned, margin-defending, seat-ceiling-proof — and is selling it into a base already pulling 10% of new ARR from AI with agents not yet counted.

The pivot deepens the through-line rather than settling it. It confirms the company is neither passive nor impaired: it has read the threat correctly and re-engineered its pricing around it. But it converts Chapter 5's open question into a measurable one. The reader no longer has to ask whether AI matters to the case; they have to watch two numbers move in opposite directions — gross margin down, AI's share of net new ARR up — and judge, quarter by quarter, which one is winning. That, finally, is a question the next few earnings calls can answer.