Executives and researchers representing multiple layers of the artificial intelligence industry said growing demand for AI infrastructure is running into major supply and operational constraints, with shortages affecting chips, cloud capacity, and real-world training data.
The discussion took place earlier this week during the Milken Institute Global Conference in Beverly Hills, where leaders from semiconductor manufacturing, cloud computing, AI software, autonomous systems, and quantum computing discussed the current limits facing the industry.
Participants included Christophe Fouquet, Francis deSouza, Qasar Younis, Dmitry Shevelenko, and Eve Bodnia.
The panel covered topics ranging from semiconductor shortages and AI infrastructure investment to questions about whether current AI architectures are fundamentally sustainable.
Chip Manufacturing Remains Supply Constrained
Fouquet said demand for advanced semiconductors continues to exceed available manufacturing capacity despite rapid expansion efforts across the industry.
ASML produces the extreme ultraviolet lithography systems used to manufacture advanced chips, giving the Dutch company a central role in global semiconductor production.
Fouquet described what he called a “huge acceleration of chips manufacturing,” but said supply shortages are likely to continue for several years.
“I have a strong belief that for the next two, three, maybe five years, the market will be supply limited,” he said during the discussion.
The comments suggest major cloud companies including Google, Microsoft, Amazon, and Meta Platforms may continue facing difficulty securing enough chips to meet AI demand.
Google Cloud Backlog Expands Rapidly
DeSouza said demand for AI infrastructure and cloud computing services continues growing at a rapid pace.
He noted that Google Cloud generated more than $20 billion in revenue during the most recent quarter, representing 63% growth.
According to deSouza, the company’s backlog of contracted but undelivered revenue also expanded sharply, increasing from $250 billion to $460 billion within a single quarter.
“The demand is real,” deSouza said.
The figures highlight the growing pressure on infrastructure providers as companies increase spending on AI systems and cloud-based computing services.
Physical AI Faces Real-World Data Limits
Younis said companies building AI systems for physical environments face different constraints that cannot be solved solely through additional computing power.
Applied Intuition develops autonomy systems used in vehicles, drones, mining equipment, and defense technologies.
According to Younis, one of the largest bottlenecks involves collecting sufficient real-world operational data.
“You have to find it from the real world,” he said, referring to the data needed to train physical AI systems safely and reliably.
He added that synthetic simulations still cannot fully replace real-world testing and observation for autonomous systems operating in unpredictable environments.
“There will be a long time before you can fully train models that run on the physical world synthetically,” Younis said.
Featured image credits: IO Health
For more stories like it, click the +Follow button at the top of this page to follow us.