The AI economy has sprinted ahead, but some of the people building it now say the system underneath is starting to buckle.

At the Milken Global Conference in Beverly Hills, five figures spanning the AI supply chain told TechCrunch that the pressure points have moved well beyond hype cycles and product launches. Their discussion stretched from chip shortages to orbital data centers, but the throughline stayed the same: demand for AI keeps rising faster than the physical and technical infrastructure required to support it.

The message from across the stack was blunt: AI growth now depends as much on physical limits and system design as on software breakthroughs.

The warning matters because it comes from people who touch different parts of the market, not from a single company defending its own position. Reports indicate the concerns center on both near-term bottlenecks and deeper structural questions. Chips remain scarce, infrastructure appears strained, and some participants suggested the underlying architecture of modern AI systems may itself need a rethink if the industry wants to keep scaling efficiently.

Key Facts

  • Five AI supply chain figures discussed industry pressure points at the Milken Global Conference.
  • The conversation highlighted chip shortages and infrastructure constraints.
  • Speakers also raised the idea of orbital data centers as part of future capacity planning.
  • Some suggested the current architecture behind AI may be fundamentally flawed.

That mix of immediate scarcity and long-range uncertainty cuts to the heart of the current AI race. Investors and executives have treated compute capacity as the fuel for the next wave of products, but fuel only matters if the engine can handle it. If reports from this conversation hold, the industry faces a harder reality: scaling AI may require not just more hardware, but a different blueprint for how the whole system works.

What happens next will shape more than tech company roadmaps. If chip supply, energy demands, and data center capacity continue to tighten, costs could rise and deployment could slow across the market. And if the architecture itself needs to change, the next chapter of AI may belong not to the companies that moved first, but to those that rebuild the foundation fastest.