Five AI infrastructure leaders at Milken Conference discuss chip shortages, energy constraints, and alternative AI architectures

At the Milken Global Conference in Beverly Hills, CEOs and founders from ASML, Google Cloud, Applied Intuition, Perplexity, and Logical Intelligence discussed supply chain bottlenecks. ASML's Christophe Fouquet stated the market will remain supply limited for the next two, three, maybe five years despite accelerating chip manufacturing. Google Cloud's Francis deSouza reported the company's backlog nearly doubled in a single quarter, from $250 billion to $460 billion, driven by AI infrastructure demand.

Google is exploring data centers in space to address energy constraints, accessing more abundant energy in orbit. DeSouza noted that space's vacuum eliminates convection cooling, requiring radiation as the sole heat-shedding method. Google's co-engineering strategy—custom TPU chips integrated with its full AI stack—achieves greater energy efficiency (flops per watt) than off-the-shelf components.

Logical Intelligence, led by quantum physicist Eve Bodnia, is building energy-based models (EBMs)—AI systems that learn underlying rules in data rather than predicting the next token in a sequence. Her largest model runs 200 million parameters and claims to run thousands of times faster than leading large language models, and updates knowledge as data changes without full retraining.

Perplexity's Dimitry Shevelenko described the company's evolution into a 'digital worker' product. Perplexity Computer uses agents that present a plan and request approval before taking actions on a user's behalf. Enterprise administrators can specify tool and connector access levels (read-only or read-write) and granular permissions to maintain security control.