Roblox's AI Creation Tools Can't Fix a Bookings Problem—Yet

Roblox left its annual developer conference in San Jose this week with a burst of announcements that read like a pure growth story: type a prompt and the new AI "Build" app assembles a working game, an offline mode for playing on a plane or in a subway car, and engine upgrades meant to finally open the platform to 2D and puzzle titles. Any one of these would anchor a headline in a normal week. What makes them hard to read is the number sitting beside the stock: Roblox has lost roughly two-thirds of its value over the past year, the market's verdict on a bookings slowdown that began before any of these tools shipped.

The conference and the selloff are two halves of the same question, and it is worth separating the near-term problem from the long-term answer before deciding what any of this means.

The growth number that misleads

Roblox reported fiscal Q2 with a headline that seems to contradict the stock price entirely: revenue rose 36% year over year to about $1.5 billion. The figure that matters for current demand is a different one the company reports only a step back from revenue — bookings, which grew 8% to $1.6 billion, at the low end of guidance and below internal targets, a deceleration from the 42% growth a year earlier.

The gap between the two is not a mixing of apples and oranges; it is the mechanism of how the business is paid. Bookings are cash collected when players buy the platform's virtual currency. Revenue is that cash recognized over time, as the currency is actually spent inside games. So revenue this quarter is largely a delayed echo of what players paid in earlier periods, while bookings is the live reading of what they are paying now. The 36% "growth" is partly the echo running hot; the 8% is the reality of the present.

That present has been slowing since management deliberately changed the recommendation algorithm months ago to favor long-term retention over immediate spending. The shift pushed younger players away from the high-spend viral hits of 2025 toward newer and "evergreen" games that earn less revenue per hour of play, and the pain concentrated in U.S. users under 13. Tougher age verification made onboarding harder on top of it. The result showed up directly in the guidance: the company forecast third-quarter bookings of $1.58 billion to $1.65 billion against roughly $1.77 billion expected on the street — a decline of 14% to 18% year over year.

On the conference call the CFO said the monetization weakness was "likely to continue", and the company gave no full-year figure and would not confirm that 2026 is a growth year at all. That is a demand-side event, not a product problem. Users are not leaving — daily actives rose 10% to 123 million and engagement hours climbed to 29 billion in the quarter. The people are there; they are spending less per hour because the algorithm and the safety changes are steering the platform away from the games that monetized best.

What the conference actually changes

This is where the developer conference fits. Roblox's moat has never been any single game — it is the pipeline of creators who make the content that fills those 29 billion hours. The company's stated economics run on that flywheel: cheaper and easier creation produces more and better experiences, which pulls in more players, which pays creators, which funds more creation. RDC was a coordinated push to widen both ends of that flywheel at once, and the evidence on the supply side is real.

Creators earned about $1.7 billion in the twelve months through June — up about 50% — and the numbers have the signature of the pipeline widening rather than merely paying the same people more. The AI Build app, still in an early alpha in New Zealand, has already produced roughly 9,000 published games, and 71% of the creators who used it had never opened Roblox Studio before. That is the two figures carrying the entire conference: generative creation is pulling in people who would not have built anything on the old tools. New distribution follows the same logic — a web player that lets anyone click a link and play in the browser by year's end, tools to ship games as standalone apps on phones and consoles, and an upcoming real-money wallet and debit-style card meant to pay creators faster. Each is mechanical: more surfaces to reach players, more and faster money flow to creators.

What separates this conference from the earnings problem is timing. The hostility is in the near-term meter; the cure is on the long clock. Almost everything announced lands in late 2026 or across 2027 — the offline mode is not expected in full until mid-2027, the web player by the end of this year, the wallet later this year and the card in 2027. None of it is in the financials yet, and by the company's own math the tools exist to grow bookings back to a 20%-plus trajectory, not to fix next quarter.

The judgment

So the honest reading is that the market is not wrong about the near term, and the conference does not yet refute it. The stock's 66% drawdown is a bookings repricing, and a supply-and-distribution push that begins shipping in late 2026 does not reverse a demand-side slowdown that management concedes will persist this quarter. The gap in valuation — around $44 per share against a 52-week high above $140 — already prices in a lot of that deceleration, which is the argument that the risk is partly spent.

But conviction on the long thesis should not be borrowed from the product roadmap; it should follow evidence that the demand meter is turning back up. The single number to watch is bookings, not the revenue headline, and the earliest legitimate signal is whether third-quarter bookings land anywhere near the high end of guidance and whether the company is willing, by year's end, to call 2027 a growth year. The creator pipeline underneath all of it — DAUs up, hours up, creator earnings up 50%, an AI tool pulling first-time creators in — is intact and arguably improving. The platform's engine is fine. The gauge on the dashboard that currently reads the economy's fuel is bookings, and until that needle moves, the conference is a reason to keep watching, not yet a reason to change the allocation.