\n<\/p>\n
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 this editor about everything from chip shortages to orbital data centers to the possibility that the whole architecture that undergirds the tech is wrong.<\/p>\n
On stage with TechCrunch: Christophe Fouquet, CEO of ASML, the Dutch company that holds a monopoly on the extreme ultraviolet lithography machines without which modern chips would not exist; Francis deSouza, COO of Google Cloud, who is overseeing one of the biggest infrastructure bets in corporate history; Qasar Younis, co-founder and CEO of Applied Intuition, a $15 billion physical AI company that started in simulation and has since moved into defense; Dimitry Shevelenko, the chief business officer of Perplexity, the AI-native search-to-agents company; and Eve Bodnia, a quantum physicist who left academia to challenge the foundational architecture most of the AI industry takes for granted at her startup, Logical Intelligence. (Meta\u2019s former chief AI scientist, Yan LeCun, signed on as founding chair of its technical research board earlier this year.)<\/p>\n
Here\u2019s what the five had to say:<\/p>\n
The bottlenecks are real<\/strong><\/p>\n The AI boom is running into hard physical limits, and the constraints begin further down the stack than many may realize. Fouquet was the first to say it, describing a \u201chuge acceleration of chips manufacturing,\u201d while expressing his \u201cstrong belief\u201d that despite all that effort, \u201cfor the next two, three, maybe five years, the market will be supply limited,\u201d meaning the hyperscalers \u2014 Google, Microsoft, Amazon, Meta \u2014 aren\u2019t going to get all the chips they\u2019re paying for, full stop. <\/p>\n DeSouza highlighted how big \u2014 and how fast growing \u2014 an issue this is, reminding the audience that Google Cloud\u2019s revenue crossed $20 billion last quarter, growing 63%, while its backlog \u2014 the committed but not yet delivered revenue \u2014 nearly doubled in a single quarter, from $250 billion to $460 billion. \u201cThe demand is real,\u201d he said with impressive calm.<\/p>\n For Younis, the constraint comes primarily from elsewhere. Applied Intuition builds autonomy systems for cars, trucks, drones, mining equipment and defense vehicles, and his bottleneck isn\u2019t silicon \u2014 it\u2019s the data that one can only gather by sending machines into the real world and watching what happens. \u201cYou have to find it from the real world,\u201d he said, and no amount of synthetic simulation fully closes that gap. \u201cThere will be a long time before you can fully train models that run on the physical world synthetically.\u201d<\/p>\n Techcrunch event<\/p>\n \n\t\t\t\t\t\t\t\t\tSan Francisco, CA<\/span> The energy problem is also real<\/strong><\/p>\n If chips are the first bottleneck, energy is the one looming behind it. DeSouza confirmed that Google is exploring data centers in space as a serious response to energy constraints. \u201cYou get access to more abundant energy,\u201d he noted. Of course, even in orbit, it isn\u2019t simple. DeSouza observed space is a vacuum, so eliminates convection, leaving radiation as the only way to shed heat into the surrounding environment (a much slower and harder-to-engineer process than the air and liquid cooling systems that data centers rely on today). But the company is still treating it as a legitimate path.<\/p>\n The deeper argument de Souza made, somewhat unsurprisingly, was about efficiency through integration. Google\u2019s strategy of co-engineering its full AI stack \u2014 from custom TPU chips through to models and agents \u2014 pays dividends in flops per watt (more computation per unit of energy) that a company buying off-the-shelf components simply can\u2019t replicate, he suggested. \u201cRunning Gemini on TPUs is much more energy efficient than any other configuration,\u201d because chip designers know what\u2019s coming in the model before it ships, he said. <\/p>\n Fouquet\u2019s made a similar point later in the discussion. \u201cNothing can be priceless,\u201d he said. The industry is in an strange moment right now, investing extraordinary amounts of capital, driven by strategic necessity. But more compute means more energy, and more energy has a price.<\/p>\n A different kind of intelligence<\/strong><\/p>\n While the rest of the industry debates scale, architecture, and inference efficiency within the large language model paradigm, Bodnia is building something very different.<\/p>\n Her company, Logical Intelligence, is built on so-called energy-based models (EBMs), a class of AI that doesn\u2019t predict the next token in a sequence but instead attempts to understand the rules underlying data, in a way she argues is closer to how the human brain actually works. \u201cLanguage is a user interface between my brain and yours,\u201d she said. \u201cThe reasoning itself is not attached to any language.\u201d <\/p>\n Her largest model runs to 200 million parameters \u2014 compared to the hundreds of billions in leading LLMs \u2014 and she claims it runs thousands of times faster. More importantly, it\u2019s designed to update its knowledge as data changes, rather than requiring retraining from scratch.<\/p>\n For chip design, robotics and other domains where a system needs to grasp physical rules rather than linguistic patterns, she argues EBMs are the more natural fit. \u201cWhen you drive a car, you\u2019re not searching for patterns in any language. You look around you, understand the rules about the world around you, and make a decision.\u201d It\u2019s an interesting argument and one that\u2019s likely to attract more attention in the coming months, given the AI field is beginning to ask whether scale alone is sufficient.<\/p>\n Agents, guardrails, and trust<\/strong><\/p>\n Shevelenko spent much of the conversation explaining how Perplexity has evolved from a search product into something it now calls a \u201cdigital worker.\u201d Perplexity Computer, its newest offering, is designed not as a tool a knowledge worker uses, but as a staff that a knowledge worker directs. \u201cEvery day you wake up and you have a hundred staff on your team,\u201d he said of the opportunity. \u201cWhat are you going to do to make the most of it?\u201d<\/p>\n It\u2019s a compelling pitch; it also raises obvious questions about control, so I asked them. His answer was: granularity. Enterprise administrators can specify not just which connectors and tools an agent can access, but whether those permissions are read-only or read-write \u2014 a distinction that matters enormously when agents are acting inside corporate systems. When Comet, Perplexity\u2019s computer-use agent, takes actions on a user\u2019s behalf, it presents a plan and asks for approval first. Some users find the friction annoying, Shevelenko said, but he said heconsiders it essential, particularly after joining the board of Lazard, where said he has found himself unexpectedly sympathetic to the conservative instincts of a CISO protecting a 180-year-old brand built entirely on client trust. \u201cGranularity is the bedrock of good security hygiene,\u201d he said.<\/p>\n Sovereignty, not just safety<\/strong><\/p>\n Younis offered what may have been the panel\u2019s most geopolitically charged observation, which is that physical AI and national sovereignty are entangled in ways that purely digital AI never was.<\/p>\n The internet initially spread as American technology and faced pushback only at the application layer \u2014 the Ubers and DoorDashes \u2014 when offline consequences became visible. Physical AI is different. Autonomous vehicles, defense drones, mining equipment, agricultural machines \u2014 these manifest in the real world in ways governments can\u2019t ignore, raising questions about safety, data collection, and who ultimately controls systems that operate inside a nation\u2019s borders. \u201cAlmost consistently, every country is saying: we don\u2019t want this intelligence in a physical form in our borders, controlled by another country.\u201d Fewer nations, he told the crowd, can currently field a robotaxi than possess nuclear weapons.<\/p>\n Fouquet framed it a little differently. China\u2019s AI progress is real \u2014 DeepSeek\u2019s release earlier this year sent something close to a panic through parts of the industry \u2014 but that progress is constrained below the model layer. Without access to EUV lithography, Chinese chipmakers cannot manufacture the most advanced semiconductors, and models built on older hardware operate at a compounding disadvantage no matter how good the software gets. \u201cToday, in the United States, you have the data, you have the computing access, you have the chips, you have the talent. China does a very good job on the top of the stack, but is lacking some elements below,\u201d Fouquet said.<\/p>\n The generation question<\/strong><\/p>\n Near the end of our panel, someone in the audience asked the obvious uncomfortable question: is all of this going to impact the next generation\u2019s capacity for critical thinking?<\/p>\n The answers were optimistic, as you\u2019d expect from people who\u2019ve staked their careers on this technology. DeSouza immediately pointed to the scale of problems that more powerful tools might finally let humanity address. Think neurological diseases whose biological mechanisms we don\u2019t yet understand, greenhouse gas removal, and grid infrastructure that has been deferred for decades. \u201cThis should unleash us to the next level of creativity,\u201d he said.<\/p>\n Shevelenko made a more pragmatic point: the entry-level job may be disappearing, but the ability to launch something independently has never been more accessible. \u201c[For] anybody who has Perplexity Computer . . . the constraint is your own curiosity and agency.\u201d<\/p>\n Younis drew the sharpest distinction between knowledge work and physical labor. He pointed to the fact that the average American farmer is 58 years old and that labor shortages in mining, long-haul trucking, and agriculture are chronic and growing \u2014 not because wages are too low, but because people don\u2019t want those jobs. In those domains, physical AI isn\u2019t displacing willing workers. It\u2019s filling a void that already exists and looks only to deepen from here.<\/p>\n<\/div>\n When you purchase through links in our articles, we may earn a small commission<\/a>. This doesn\u2019t affect our editorial independence.<\/em><\/p>\n
\n\t\t\t\t\t\t\t\t\t\t\t\t\t|<\/span>
\n\t\t\t\t\t\t\t\t\t\t\t\t\tOctober 13-15, 2026<\/span>\n\t\t\t\t\t\t\t<\/p>\n<\/p><\/div>\n<\/p><\/div>\n<\/div>\n