Key Takeaways

  • Meta is offering 'MetaMP compute,' selling raw AI processing power and models, a move that raises questions about the return on its billions in AI infrastructure investment.
  • Critics, like John Coogan, see this as a sign of low confidence in Meta's ability to create compelling, near-term AI-powered features for its own apps like Instagram, citing “generic LLM response[s]” in current products.
  • One perspective suggests Meta might be playing a long game, building immense compute capacity now and waiting for "extremely obvious" AI features to emerge within the next "year or two."
  • There's concern that Meta's exploration of prediction markets, despite a smaller potential profit pool, risks attracting intense global regulatory scrutiny, diverting focus from its core product challenges.
  • The pivot sparks debate over whether Meta is abandoning its "personal super intelligence" vision or simply diversifying its AI strategy.

The Disagreement: Build for Yourselves, or Sell to the Masses?

On one side, the conversation on TBPN reveals a deep confusion about Meta's latest AI play: pivoting to sell 'MetaMP compute,' raw AI processing power and models to external parties. John Coogan voiced the prevailing skepticism, saying, “It doesn't give you a lot of confidence that there's near-term products on the horizon for Meta that are going to be able to utilize that capacity themselves.” He highlighted the irony of Meta pouring billions into AI infrastructure while its own family of apps, like Instagram, still lacks compelling, AI-powered features. Coogan shared a telling anecdote: his interaction with a Meta LLM yielded a “very very generic LLM response,” one that disturbingly referenced “a blog post from a social media management SAS company.” It's an internal product challenge that makes the external compute sale feel, as Jordi Hays put it, "deeply confusing."

The concern here is that Meta's massive AI investment isn't solving its own product needs, making the ROI of selling compute questionable. The concern extends beyond product utility to potential regulatory landmines. Hays questioned the value of Meta entering areas like prediction markets. “Is the potential profit pool risk the attention that worth the risk of all the attention you're going to get from lawmakers globally by integrating betting into the product that is already under attack on like a million different fronts?” he asked. It frames the compute sale and diversification as a potential distraction, pulling Meta's attention into smaller, riskier profit pools rather than doubling down on its core social product and user experience.

However, another view suggests Meta might be playing a much longer, more patient game. Tyler argued that Meta, with its immense resources, “can actually wait until like the the features that they need to implement are like extremely obvious, right?” The idea is that the company is building out its foundational compute power, anticipating a future where truly "wow" AI features for apps like Instagram will become undeniable. "But maybe you know in a year or two there's going to be some feature that like wow everyone is really wants this in Instagram by then you know they'll have all the compute necessary," Tyler theorized. This perspective positions Meta's current move not as a sign of weakness or confusion, but as strategic foresight—a patient accumulation of resources for an inevitable AI-powered future, even if the exact killer apps aren't clear yet.

Who's Right (and When They're Wrong)

Both arguments hold water, but for different contexts. The skeptical view (Coogan and Hays) correctly highlights the immediate product gap and the risks of diversification. It's tough to justify selling spare capacity when your core products aren't yet showing the fruit of your own AI investments. For smaller, leaner startups, this immediate product-market fit is everything. Burning runway on infrastructure without clear product utility is a death sentence. Furthermore, the regulatory drag of venturing into sensitive areas like prediction markets, as Hays points out, is a real and present danger for any company, especially one already under scrutiny.

However, Tyler's patient approach might apply to a company with Meta's unique scale and cash reserves. When you have billions to pour into R&D and a user base in the billions, you can afford to build ahead of the curve, even if the precise product applications aren't manifest today. Meta isn't a startup; it's an empire. Its bets are different. They can invest for years in foundational technology, knowing that when a breakthrough does come, they'll be uniquely positioned to deploy it at scale. The risk for them is less about immediate survival and more about missing a generational shift if they don't prepare the ground.

What to Do With This

For ambitious founders, Meta's dilemma offers a sharp lesson in strategic resource allocation. Don't fall into the trap of building massive, speculative infrastructure without a crystal-clear product vision. Unless you have Meta's billions and market dominance, your runway demands that every large investment in underlying technology directly supports a near-term, compelling product. Before committing significant capital to infrastructure, ask: what specific, desirable feature will this enable in the next 6-12 months that users will pay for or engage with? If the answer is "we'll see," rethink your strategy.