Roblox just priced trust and safety. The number is the cost of waiting.

I came up through player support, so I have lost this argument more times than I can count.

You know the version. Someone senior asks what the safety team is actually worth, and you start talking about sentiment, about players feeling protected, about the incidents that did not happen. And you watch it hit the floor, because “the bad thing did not occur” has never once cleared a budget review. Support gets filed as a cost center. Safety gets filed as a cost center. And a cost center is a thing you are expected to shrink.

The reason we keep losing is not that we argue badly. It is that we have never had a number.

Last week Roblox handed us one. Sort of. And I think most of the coverage read it exactly backwards.

What actually happened

Roblox reported its second quarter on July 30. Revenue came in at $1.469 billion, up 36 percent, against a street expectation closer to $1.59 billion. Bookings grew 8 percent to $1.557 billion, at the bottom of their own guidance. They lost $183 million on the quarter.

Then came the part that moved the stock. They guided third-quarter bookings to a 14 to 18 percent year-over-year decline, the first guided drop in the company’s history, and they withdrew full-year guidance entirely. The next session the stock had its worst day on record.

And management said, out loud, why. Algorithm updates and stricter age verification and safety measures had depressed near-term onboarding, engagement, and monetization. On the call they got more specific: the monetization hit landed primarily in the United States under-13 cohort.

That is the thing worth sitting with. A public company just attributed a guidance withdrawal to child safety work and said so in plain language. We have spent years insisting you cannot put a price on trust and safety. Roblox went ahead and published one.

The number is real. It is also not the number people think it is.

Here is where I part ways with the takes.

Roblox is not showing us what safety costs. It is showing us what deferred safety costs.

They retrofitted mandatory age verification onto a platform with more than 130 million daily users, after it had already scaled, under regulatory pressure and active litigation. Mandatory enforcement started in Australia, New Zealand, and the Netherlands, then went global in January. That is the most expensive order of operations available to anyone. Every user already on the platform had to be re-verified, and some meaningful share of them decided it was not worth the friction.

Daily actives have gone from roughly 152 million in the third quarter of 2025 to 144, then 132, and now 123 million. That curve bends almost exactly when the age checks landed. Age-check penetration is at 57 percent globally, around 70 percent in the US and UK, and near 80 percent in Australia.

Notice what that means. The cost is a toll on an installed base, and it is being paid once. A studio that verifies at signup, before it has 130 million people to go back and ask, never pays that bill at all. Roblox is not the argument against investing in safety. It is the argument against waiting to.

What got better while the headline got worse

The other thing the coverage skipped is that the composition of the business improved while the top line disappointed.

Monthly unique payers rose 15 percent, to 27 million. The US 18 to 34 cohort is growing daily actives and hours at roughly 40 percent year over year. The monetization that fell was concentrated in under-13s, which is precisely what you would expect when a platform stops over-monetizing children.

So the honest read is not that safety broke the business. It is that safety changed who the business is made of, and the market priced the transition rather than the destination. Revenue still grew 36 percent. There are still 123 million people showing up every day. What got punished was a forecast.

If your instinct is that a smaller, older, verified, paying audience is a worse business than a larger unverified one, I would ask you to hold that thought until the regulators reach your jurisdiction.

We still cannot actually price it, and that should bother you more

Now the part that genuinely frustrates me.

Roblox reports infrastructure and trust and safety as a single combined line item. It came in at $363 million for the quarter, up 39 percent year over year, from $261 million a year ago. Data centers and child safety, added together, in one number.

Think about what that means. This is the company that has spent more visibly on safety than anyone in games, that took a guidance withdrawal for it, that has been dragged in front of regulators over it. And in its own financial reporting, you still cannot separate what it spends keeping kids safe from what it spends keeping servers on.

We finally got the most expensive natural experiment our industry has ever run, and we still do not have a unit price.

That is not a Roblox failure. It is a category failure. As long as safety lives inside an infrastructure line, it will keep being managed like infrastructure: something you provision as cheaply as possible and only think about when it breaks.

What I would do on Monday

If you run a live game and you are going to have this conversation with your finance team, do not walk in with sentiment. Walk in with these:

• Verified-cohort economics. Compare retention and spend for verified versus unverified users over the same window. If the verified cohort retains better, you are no longer defending a cost, you are describing an acquisition filter.
• Your own deferral clock. Estimate what re-verifying your current base costs you at today’s size, then at your projected size in eighteen months. That delta is the number. It is the only version of this argument that gets more persuasive the longer someone stalls.
• Regulatory exposure by market. List the jurisdictions you operate in and when each one’s rules bite. Roblox did not choose January. Australia did.
• Unbundle your own line item. If safety is buried in your infrastructure budget, pull it out. You cannot prove the return on a number you cannot see, and nobody is going to do it for you.

None of that gives you a clean ROI. I am not going to pretend it does. But it moves the conversation from “trust us, this matters” to “here is what it costs us to wait,” and in my experience that is the only version finance has ever actually listened to.

The lesson of this quarter is not that safety is expensive. It is that Roblox paid retail because it showed up late, and it told everyone the price on the way out.