tech
May 7, 2026
Five architects of the AI economy explain where the wheels are coming off
Earlier this week, five people who touch every layer of the AI supply chain sat down at the Milken Global Conference in Beverly Hills, where they talked with TechCrunch about everything from chip shortages to orbital data centers to the possibility that the whole architecture that undergirds the tech is wrong.
TL;DR
- The AI boom faces physical limitations, with chip manufacturing being a major supply constraint for the next 2-5 years.
- Demand for AI infrastructure is immense, evidenced by Google Cloud's rapidly growing backlog.
- Gathering real-world data remains a bottleneck for training physical AI systems, even with advanced simulations.
- Energy consumption is a critical concern, prompting exploration of solutions like space-based data centers.
- Google's integrated AI stack, from custom chips to models, offers significant energy efficiency advantages.
- Alternative AI architectures, like energy-based models, are being developed to better mimic human brain function and process information faster.
- AI agents are evolving into 'digital workers' that require careful control and granular permissions to ensure security and trust.
- Physical AI systems have geopolitical implications, raising concerns about national sovereignty and control over technology operating within borders.
- China's AI progress is noted but limited by its inability to manufacture advanced semiconductors, a constraint addressed by EUV lithography.
- AI is expected to boost creativity and help solve large-scale problems, while also changing the nature of entry-level jobs and filling labor shortages in physical industries.
Continue reading the original article