> if cash flow ever does stop increasing, they can always get back to sustainability by pumping the brakes on reinvestment spending
This is a point that's sometimes less obvious with cash flow games. It's possible to have positive cash flow even with negative unit economics, _even when no economy of scale can sufficiently improve those unit economics_ [0], so long as you have enough growth and a good cash flow situation.
That's one of the criticisms Uber has had over the years; are they capable of sustaining their apparent pre-reinvestment profits if they cut out that spending? It's potentially a bit different from the Amazon situation because most of the money is going straight into speculative bets, acquiring competitors, ads, ride subsidies, and other activities designed to lock in the market, and it's unclear if that will give them a meaningful moat, as opposed to, e.g., capital investments in a fantastic, in-house distribution and shipping mechanism.
Can Roblox actually become sustainable by cutting spending somewhere?
[0] Imagine a product with -50% unit ROI. For every dollar in revenue you have two dollars in guaranteed costs. However, suppose the product is paid for fairly early relative to those costs (e.g., the business offers a steep discount on yearly subscriptions if you pay up-front, the costs are incurred linearly throughout the year as the subscription is used, and there's a till-the-start-of-next-month plus 30 days lag on billing for computing resources used). You haven't actually used enough resources to be in the red till 6 months after the subscription starts, and you're not actually on the hook for that last payment till 7 months have elapsed. If you're also able to hit a 2x annual growth rate in your paid subscriber count (not realistic for large companies, not uncommon for a few years with good product-market-fit in gaming or some SAAS products), you've paid for the year's losses before the year has ended and still have an extra month at the end where the money is sitting in your account. As your company doubles its subscribers, your coffers will continue to double as well, even if you have indefinitely negative unit economics.
In the real world you usually have smaller numbers being considered (smaller losses, less growth), allowing the game to go on for many more years.