The October 19th, Afternoon Workshops: 1 PM - 5 PM

The Massive Global Silicon Build: The Emergence of Super Computing & Unlimited Compute

Every Intelligence Factory begins with the same bottleneck: silicon. Chip fabrication capacity — not compute demand, not capital, not even energy — is increasingly the constraint determining which nations and companies can actually build at the scale the Intelligence Economy requires. The race to expand fabrication capacity has become as strategically important as the factories it supplies, with billions committed to new fabs, packaging facilities, and materials supply chains across the United States, Asia and the Middle East. Whoever controls capacity at the silicon layer effectively controls the pace of everyone building above it.

This is no longer a chip industry story. It is a global industrial policy story — one where semiconductor strategy, national security and economic competitiveness have become inseparable. This session examines the strategies now being deployed to secure global capacity: who's building fabs, who's funding them, and who risks being left dependent on someone else's silicon.

Taiwan still makes the chips the world depends on — TSMC alone produces the majority of the planet's most advanced semiconductors. The Netherlands' ASML makes something even scarcer: the extreme ultraviolet lithography machines required to manufacture them, a chokepoint technology no other company on Earth has replicated. China, cut off from that supply chain by export controls, is racing to build domestic capacity from scratch. The United States, after decades of offshoring, is now underwriting a new fabrication ecosystem across Arizona, Ohio and beyond — while SpaceX/X.ai has claimed it will outpace them all. This is no longer a chip industry story. It is the defining industrial competition of the decade — measured in fabs built, machines secured, and nations willing to spend whatever it takes not to be left dependent on someone else's silicon.

This session examines who's actually winning the race to execute.

It begins as sand — ordinary silica, indistinguishable from a beach. What happens next is closer to alchemy than industry: melted, purified to a standard beyond anything nature produces on its own, grown into a single flawless crystal, sliced thinner than a hair, and etched with patterns smaller than a virus using light bent by machines only one company on Earth knows how to build. What comes out the other end doesn't just compute. It thinks, in some approximation of the word — and increasingly, it reasons. That transformation, sand into something that can move markets, model proteins, or write a sentence with intention, used to take a generation. It is now a race among labs, toolmakers, and nations to compress it into a handful of years. The countries treating this as a single unbroken chain — materials, process, packaging, architecture, cooling, power, software — rather than separate departments working in isolation, are the ones pulling ahead. Superintelligence is not a software surprise arriving from nowhere. It is the end of a longer spell, one that starts in a sand pit and does not stop until it reasons. This session traces that entire chain — and asks who is casting it fastest.

Every era looks like magic until someone builds the machine that makes it ordinary. The GPU was that machine for this decade. It will not be the last one. There is a version of this story where silicon is not mined but summoned — pulled from stardust that has been falling to Earth since before there were eyes to watch it fall, shaped by hands that learned, generation by generation, to bend light itself into instruction. The GPU was only ever a way station, a clever machine built for a narrower magic than the one now being attempted. What comes after it is not simply faster. It is architected differently, the way a new instrument is built when the old one can no longer carry the music being written for it. Somewhere between a physicist's chalkboard and a fabrication floor lit like a cathedral, the next compute era is being drawn by people who still believe, the way children believe, that the impossible is only unbuilt. This is not evolution. It is invention at the edge of ingenuity, where genius stops explaining itself and simply arrives. This session follows the architects chasing what comes after the GPU — into the room where the next machine is still being dreamed before it is drawn.

The die is now a diplomatic instrument. Trade ministries, defense ministries, and foreign ministries sit on the same file because a shipment of tools or a license on a server can move a balance of power faster than a treaty. Taiwan remains the center of the leading edge, which means every alliance is also an insurance policy. The United States, Japan, Korea, and the Netherlands are writing industrial policy as foreign policy: where a fab may be built, which tool may leave, who may buy the advanced package. China is answering with a national compute grid and domestic silicon, and with outbound rules on what a weight file and a talent base may do. Export control has already jumped from the chip to the rack, the cloud hour, and the remote login. A limited partner in one capital can make a campus in another uninsurable. This session puts semiconductors where they now belong: at the table with ambassadors, not only with process engineers. Which alliances can guarantee supply when a strait closes. Which licenses are really sanctions. Which sovereigns will accept a factory they do not fully own. Diplomacy that still treats chips as a commerce footnote will arrive after the capacity has already chosen a flag.

Every fab that gets built starts as three separate conversations that eventually have to become one. Scientists need years of runway and capital patient enough to fund research with no guaranteed return. Governments need a policy framework that de-risks investment enough to justify it, and the political will to sustain that commitment across election cycles. Investors need a return that makes sense on a project measured in decades, not quarters. The fabs actually breaking ground right now are the ones where all three converged — a national lab's research, a government's subsidy and permitting, and a fund willing to underwrite the gap between announcement and first wafer. Everywhere those three fail to align, ambition stalls in press releases. This session brings together the people who actually close these deals: the scientists, the policymakers, and the capital allocators who turn a fabrication roadmap into a building with machines running inside it.