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

The Hyperscaler & Infrastructure Build - A State by State Strategy: Energy - Compute - Investment

Every great economic era begins with the decision to build. The Intelligence Economy presents that opportunity at extraordinary scale. Across the country and around the world, new centers of compute can become catalysts for energy innovation, advanced infrastructure, scientific research, industrial development and entirely new regional economies.

But transformation requires more than technology. It requires the alignment of ambition and resources—states prepared to lead, energy providers prepared to expand, hyperscalers prepared to build, and investors prepared to deploy capital against a future still being created. The opportunity extends far beyond the construction of Data Centers. Intelligence infrastructure can reshape where companies invest, where scientists and engineers gather, where new industries emerge and where the next generation chooses to build its future. Capital deployment therefore becomes an act of economic imagination: the ability to recognize what does not yet exist—and assemble the partnerships, resources and conviction necessary to build it.

Capital investment does not simply follow the model or polite request from a hyperscaler or technology entrepreneur. The massive investment needed to secure the Intelligence Factory follows the permit and the megawatt. The states that lock both first will own the next decade of training capacity. Everyone else will rent it. Texas is already sitting on multi-gigawatt campuses and an ERCOT queue measured in hundreds of gigawatts. Ohio has two Intelligence Factory permits in ninety days. Arizona and Northern Virginia are fighting over the last interconnect. Utah wrote SB 132 so a 100 megawatt user can generate behind the meter and not put the household on the bill. Virginia taxes the kilowatt-hour. Arizona paused new exemptions. New York paused 50 megawatt permits. Local boards from Loudoun to Box Elder can still kill a site the capitol thought it had won. This session is for the people who write the check and the people who sign the CUP. Where is power actually deliverable in 2027 and 2028. Which statutes survive the next election. Which states will take a hyperscaler’s own gas plant, nuclear restart, or storage island instead of forcing the load onto a crowded grid. The future does not rest in a term sheet. It rests in the jurisdiction that can say yes to power and yes to the factory in the same year.

The emergence of the Intelligence Factory, it's demand on resources, the implication of its place in the greater society has provided a challenge and odd circumstance that the universe of government and state agencies, the funding giants from Wall Street and the hyperscapers could never have anticipated. The challenge is almost cinematic and beyond the scope of any prior "Industrial Expansion." No Intelligence Factory closes with one signature. It closes when four chairs are occupied at the same time: the state that can permit, the hyperscaler that will take the load, the generator that can deliver firm power, and the capital that will live with a 20-year offtake. That alliance is already being built in pieces. Labs are signing ratepayer pledges and bringing their own gas, nuclear restarts, and storage so households do not carry the campus. Utah used SB 132 to put a large-load customer and Rocky Mountain Power on one agreement that keeps existing ratepayers whole. DOE is putting federal land, private equity, and utilities on the same site. Meta, Microsoft, Amazon, and Google have turned themselves into offtakers of last resort for reactors that the merchant market would not restart. Infrastructure funds will not fund a gigawatt that lacks a sovereign or state partner on the other side of the table. This session is the working table, not a panel of speeches. Who brings the interconnect. Who carries construction risk. Who steps in if the lab misses a training cycle. The states and firms that learn to sit together will allocate the decade. The ones that keep negotiating in separate rooms will watch the capacity go elsewhere.

The Intelligence Factory is closer to an MegaWatt Dream Machine than it is to any power plant constructed in a prior era. An Intelligence Factory is a power plant that has been designed to train and reinforce models. If the electrons are late, the chips sit dark. For this decade that plant still lives on the ground: gas that can be built on a factory clock, nuclear restarts that can be contracted, new reactors that can be financed for the 2030s, batteries that buy hours, and the transformers and laterals that now decide COD more than the GPU allocation. The next increment will not all sit on this grid. Orbital compute, satellite power, and the Earth-Moon corridor are already in the capital stack as infrastructure, not as science fiction. Launch cadence, continuous sunlight, thermal rejection in vacuum, and who underwrites a gigawatt that never touches a state PUC are the same feeding problem at a different altitude. Hyperscalers writing 20-year offtakes for dead reactors will be the same names writing the first off-Earth offtakes. This room prices steel, copper, propellant, and a stranded gigawatt if the campus or the constellation never reaches nameplate. The factory only exists if the machine is fed, on the ground first, then above it.

Will this era be remembered for its "Economic and Technological Visionaries," for its "Robber Barons," for its "Wild Speculators," or the "Sovereign Funds of the Global Giants?" The factory, the fuel, the chips, and the money no longer live in one country, and the alliances that bind them will decide who trains at scale. Japan is on the term sheet for American campuses. European capital wants gigafactories and sovereign cloud on this side of the Atlantic. Gulf and Canadian funds are hunting firm power and interconnect. London is carving Growth Zones and charging a fortune to stay in the queue. Brussels is enforcing model rules while it tries to buy capacity it does not have. Beijing is building a national compute grid on domestic silicon and weighing outbound controls on weights. Washington is running export control, inbound review, and federal land as a single industrial policy. A limited partner in one jurisdiction can poison a site in another. This session is the map of who may sit at the table. Which sovereigns can own a piece of a Texas or Ohio factory. Which offtakes can cross a border. Which alliances survive a change in license policy or a war scare. Capital that still thinks in domestic project finance will miss the decade. The horizon is already international. The question is who writes the alliance before the capacity is spoken for.

Compute has left the capex budget and entered the capital markets. Jensen Huang said the quiet part on CNBC: the chip is an investable asset. NVIDIA then put six houses on it. Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR are standing up independent platforms to mobilize more than $500 billion of third-party capital so an AI factory can be financed the way a pipeline or a tower once was. That is the new structure. Not a hyperscaler writing a check for a building, but dedicated pools that underwrite NVIDIA compute as a productive asset: lowest token cost, long useful life, fungible across labs and clouds, usage-linked revenue, CUDA extending the economic life of the iron. The offtaker is no longer only one model company. It is a global book of countries, enterprises, and AI clouds that need scarce capacity and will pay for it over years. This session is for the people who have to turn that announcement into a close. How you tranche a multi-gigawatt factory when the collateral is a cluster and a software stack. Who owns the residual when the offtake rolls. What a pension or a sovereign is actually buying when they buy compute. The paper that gets written in the next 24 months will decide who gets an AI factory and who rents one.