AI Industry Leaders Warn of Growing Infrastructure and Supply Bottlenecks

What happened

Leading figures across the AI supply chain gathered at the Milken Institute Global Conference in Beverly Hills to discuss the mounting physical constraints on the industry. Executives from ASML, Google Cloud, Applied Intuition, and Perplexity, alongside researchers from Logical Intelligence, warned that the architecture undergirding the current AI economy is facing severe bottlenecks in hardware manufacturing, power availability, and data acquisition.

What's new in this update

Google Cloud COO Francis deSouza disclosed a massive surge in demand that is outpacing the company's ability to deliver. Google Cloud's backlog—representing committed but undelivered revenue—nearly doubled in a single quarter, growing from $250 billion to $460 billion. To mitigate terrestrial energy constraints, deSouza confirmed that Google is exploring orbital data centers in space as a serious infrastructure alternative.

Key details

ASML CEO Christophe Fouquet, whose company holds a monopoly on the lithography machines required for modern chip production, stated that hyperscalers will likely not receive all the chips they are paying for over the next several years. For physical AI applications, such as autonomous defense and mining equipment, Applied Intuition CEO Qasar Younis noted that synthetic simulations cannot yet replace the 'long tail' of data gathered by machines in the physical world, creating a data-driven bottleneck that silicon cannot solve alone.

Background and context

The AI industry is currently dominated by massive investments from 'hyperscalers' like Google, Microsoft, Amazon, and Meta. While much of the focus has been on software breakthroughs, the underlying physical layer relies on a fragile supply chain. ASML's EUV lithography machines are the sole source for advanced chip manufacturing, and the energy demands of massive GPU clusters have begun to strain municipal power grids, prompting the search for unconventional energy solutions.

What to watch next

The industry is increasingly looking toward fundamental architectural changes to bypass current limits. Eve Bodnia's Logical Intelligence, which recently appointed Meta’s former chief AI scientist Yann LeCun as founding chair of its technical research board, is working to challenge the foundational tech most of the industry takes for granted. Additionally, the timeline for orbital data centers and the 2-to-5-year chip supply window will likely dictate the speed of AI expansion through the end of the decade.