Key Takeaways

  • Meta is doing a "180," moving from a period of unchecked AI token consumption to a budgeted, efficiency-focused "min-maxing" approach for its billions in AI spending, as revealed in an internal memo.
  • The concept of "min-maxing" applies beyond AI compute, pushing founders to seek the highest possible output for the lowest possible cost, much like a gamer optimizing resources for an advantage.
  • Despite Meta's public assurances, Gavin Baker suggests the market's skepticism about their ability to monetize massive GPU infrastructure internally could force them to rent out compute externally if stock performance demands it.
  • This strategic pivot reflects a broader business principle: move past simply maximizing or minimizing expenditure, and instead focus on optimizing resources for maximum return on investment.

Meta's Compute Pivot: From Free-For-All to ROI

Forget "token maxing." Meta, once seen as a free-spending giant in AI compute, is doing a full "180." According to John Coogan, a recent internal memo signals a dramatic shift from letting engineers freely consume AI tokens to a much tighter, budgeted "min-maxing" approach. Coogan explains the concept simply: "You shouldn't be token maxing or token mining. You should be token minmaxing. And the idea of minmaxing is when you're in a video game, you want to get the most resources for the lowest cost." It's about getting the biggest bang for your buck, not just burning through resources because they're available.

This isn't just about AI; it's a core business principle for managing billions in spending. Meta is moving past a phase where engineers might have freely grabbed compute. Now, it's about making allocation decisions, setting budgets, and demanding a clear return on investment. The implicit lesson for founders here is stark: unrestrained resource consumption, however exciting the project, eventually hits the wall of financial reality.

GPU Assets: The Next Monetization Play?

Beneath Meta's internal shift lies a deeper market tension. Gavin Baker points out that Meta's “EV to net PP&E multiple is in an interesting place.” This dry financial metric tells a story: the market is deeply skeptical about Meta's ability to effectively monetize its vast, expensive GPU assets internally. Imagine having a supercomputer in your garage, but no one believes you can make money with it.

Baker suggests that Meta's public stance – that there's “no chance that we're going to monetize GPUs externally” – might quickly change. He quips, “well, I don't know, check back in a few hours.” If market conditions sour or stock performance flags, Meta could rapidly pivot to become an external compute provider, essentially renting out their GPU power to other companies. This potential pivot underscores a critical lesson: every asset on your balance sheet is a potential revenue stream if internal utilization isn't driving sufficient value. Just as John Coogan noted with marketing, “the Super Bowl is literally for a marketer the Super Bowl is spend 5 million in one day and it works sometimes.” Big spending isn't inherently bad, but it must be purposeful, measurable, and yield a return.

What to Do With This

Pull your last three major resource allocation decisions – whether it's marketing spend, engineering hours on a specific feature, or cloud credits. For each, explicitly define what a "min-maxed" outcome would have looked like (e.g., "acquire 100 users for under $5 each"), then measure your actual results against that target. Implement a weekly review where you question whether every dollar spent or hour invested is truly maximizing output for minimum cost, making a specific case for each resource burn.