The quote, and the room it was said in
Speaking at the RetailClub AI Festival in Huntington Beach, California in September, ThredUp co-founder and CEO James Reinhart told Retail Dive he was surprised by how quickly the resale company's AI usage costs grew, and that he simply did not care. Token usage rose faster than he had modeled going in. His response was to keep pushing employee AI use rather than restrict it, reasoning that the eventual benefits will exceed the upfront spend.
That is a specific, falsifiable claim, not a vague enthusiasm statement, and it is worth treating it that way rather than filing it under the usual executive AI boosterism. Reinhart built his comments around a slide deck he presented at the event, one he said was built using Claude and reviewed by his personal assistant, which is itself a small, concrete example of the kind of AI assisted work he wants to see scale across the rest of the company.
What the numbers actually say
ThredUp's second quarter 2026 results give some texture to the shrug. Revenue grew about 17 percent year over year to 90.8 million dollars, with growth in both active buyers and orders, a healthy top line trend for a resale platform. But operating expenses rose more than 17 percent to about 78.6 million dollars in the same period, and net loss grew 15 percent to 5.9 million dollars rather than narrowing.
Revenue and expenses grew at nearly identical rates that quarter, and the loss widened rather than shrank. None of that proves AI spend specifically is the problem, since ThredUp has other cost lines, but it does mean the company has not yet shown the productivity payoff Reinhart is promising investors. The bet he described in Huntington Beach is still a bet, not a result, and the Q2 numbers are the most recent evidence available on how it is actually performing.
The case for not caring
Reinhart's reasoning on the Q2 earnings call was that AI should eventually cut costs by slowing the pace of headcount growth and by making existing teams more productive, which is a conventional and defensible argument. If AI spend substitutes for hiring that would otherwise happen, the near term token bill is simply a different line item replacing a future payroll line, and comparing this quarter's AI cost against this quarter's headcount savings understates the real tradeoff.
The problem with that argument is timing. Headcount avoidance is a benefit that shows up gradually, often invisibly, as roles that simply never get created, while AI usage costs show up immediately and visibly on the income statement. A leader willing to carry that mismatch for several quarters, the way Reinhart describes doing, needs either a long runway or a board with real patience for a cost curve that has not bent down yet by his own account.
The forecasting admission
The most useful part of Reinhart's comments, for other technology leaders, sits past the bravado about not caring. It is his admission that ThredUp still needs to improve how it forecasts AI costs now that it is a public company. That is a specific, actionable gap: a leadership team that has encouraged broad AI adoption faster than its own finance function can currently model the resulting spend, and said so to a room full of industry peers rather than only to its own board.
This is an extremely common position for mid sized public companies right now, and Reinhart deserves some credit for saying it plainly rather than pretending the forecasting problem does not exist. It also means ThredUp's current AI cost trajectory is running ahead of its own planning process, a different and more uncomfortable claim than simply saying costs grew faster than expected once and then stabilized at a new, predictable level the finance team can now model with confidence.
Why this resonates beyond ThredUp
Reinhart's stance sits in deliberate contrast to the more cautious tone most public company executives take when asked about AI spend on earnings calls, where the standard answer leans heavily on efficiency framing and avoids admitting surprise. Saying out loud that costs outran your own model, in front of an industry audience, and that you are fine with it anyway, is a riskier public stance than most CFOs would sign off on without a fight.
That rhetorical risk is exactly why the comments traveled. It is a cleaner, more honest version of a conversation happening inside nearly every retail technology organization right now: AI usage is scaling faster than anyone's cost model anticipated, and leadership has to decide in real time whether to slow adoption to protect the forecast or keep pushing adoption and fix the forecast later. ThredUp chose the second path, in public, which is unusual enough to be worth tracking.
What to take into your next budget cycle
If you are building next year's AI budget, Reinhart's experience argues for modeling a wider error band on usage cost than you probably have today, rather than assuming last quarter's token spend predicts next quarter's. Token usage tied to broad, unrestricted employee adoption does not scale linearly in ways finance teams can easily forecast from historical trend lines alone, and a single new use case spreading virally inside a company can move the bill by a wide margin within one quarter.
The more durable lesson is about sequencing the conversation with your own board or investors. Reinhart got ahead of the surprise by naming it publicly before an analyst or journalist found the gap in a filing. If your AI spend is going to outrun your forecast this year, and for most organizations currently scaling adoption it probably will, disclose the surprise and your reasoning on your own timeline rather than waiting for someone else to notice the variance first.
