Two paths

Dual-purpose an existing 100 MWdc PV farm.

The offering

Existing PV farm

Phase 0 is 50 MW, Phase 1 is 58 MW, then 153 MW, then 175 MW

One retasked plant, two yields: power the grid will not take, and 24/7 cooling the compute requires.

Turn daytime solar equipment into a 24/7 cooling asset, then drop in modular AI. Screened to 100 MWdc+ plants. Phase 1 is 58 MW. Phase 2 is 153 MW. Phase 3 is 175 MW, full off-grid.

100 MWdc · +103 MW · 153 MW firm

Same land

Later phases

153 MW, then 175 MW full off-grid

Phase 2 stays on the grid, adds engines and the heat train, and reaches 153 MW. Phase 3 is full off-grid at 175 MW of firm IT.

153 MW then 175 MW · 500 acres · same acreage

How to thrive in an era of superintelligence and capable humanoid robotics.

In that era, the existing grid connection is the scarce asset.

Superintelligence can make a single invention cheap. Humanoid robots can make labor cheap. Neither can make a grid connection.

1An older solar plant already has the land and the grid equipment. A new interconnection often takes five to seven years. The existing interconnection is the head start. It does not by itself authorize nighttime imports or firm compute. Those still need utility and regulatory approval.

2Humanoid robots can lower the cost of farm labor. Cheaper labor deflates the price of food, so land kept for farming earns less. That is the land a hall developer will try to take. An AI hall on farmland is a not-in-my-backyard (NIMBY) public fight over arable land use. Same·Point starts on a plant already used for power, so it does not take an acre out of crops.

3Superintelligence can design around one mechanical claim, such as a drive. A design-around does not produce a full solution that includes electrical capacity, a permit, or power every hour. Two pending U.S. applications, 282 claims as filed, seek protection of the path from the digital twin and site model through cooling-first design, EPC execution, and operations.

4Same·Point repurposes that plant. Cooling comes first, so less electricity is spent on cooling and more of the capacity can serve compute. The named case prices 99.9% firm IT, meaning uptime, at $3 million per MW-year. That figure is a planning illustration, not a contracted price. A megawatt freed from cooling is worth the firm-IT rate only if it can be sold as firm IT. Otherwise it is worth only the power bill it saves.

5Lenders pay for machines that have already run, so the first plants use proved equipment. The new cooling is shown there before later plants are financed on it.

Same footprint. Existing grid infrastructure. More firm compute.

Claim counts are as filed and may change in prosecution. These are pending applications, not issued patents.

The screen

How a site qualifies.

The working set is existing interconnected PV. Screen, not a funnel: fiber, offtake, soils, and overnight import still have to clear.

Existing PV farm

Existing interconnected PV
Cooling-first conversion screen
No new solar land
50 MW IT per 100 MWdc — named case

Later phases

Existing AI load and interconnect
Adjacent land — ideally with PV already installed

Where

Screened U.S. plants.

Ground-mounted U.S. utility-scale PV that already exports. Same cooling-first set as the partner screen: 474 plants, 43.5 GWdc. Qualifying plants, not signed contracts.

Screened plants: 474
Nameplate: 43.5 GWdc
Screen: Cooling-first · September 2026

Phase 1 AI this set can host

~21.8 GW

~$65B / year at $3M per MW-year of 99.9% firm IT. Named case: 50 MW IT fast-track per 100 MWdc PV. Qualifying plants, not signed contracts.

Plants per state01–34–1011–2526+Each dot is a plant

1CA116 · 9.7 GWdc
2TX67 · 12.4 GWdc
3NC27 · 1.7 GWdc
4GA25 · 3.1 GWdc
5VA25 · 1.8 GWdc
6AZ23 · 1.8 GWdc
7OR21 · 0.5 GWdc
8SC20 · 1.2 GWdc
9UT18 · 1.8 GWdc
10FL16 · 1.4 GWdc

Same cooling-first set as Partners → Screen (September 2026). 474 plants, 43.5 GWdc. Public locations. Qualifying plants, not signed contracts.

Rest of world

Screened plants outside the United States.

Public registers where they exist. Size and interconnect, not a U.S. ITC vintage test. AI compute at 50 MW per 100 MWdc of screened aperture. Japan and Korea use a 50 MW plant floor. Not offtake.

Country	Screened plants	Nameplate	Phase 1 AI	Phase 1 firm value
Germany	—	33 GWdc	~16.5 GW	~$50B / yr
Spain	—	32.2 GWdc	~16.1 GW	~$48B / yr
Brazil	470	20 GWdc	~10 GW	~$30B / yr
Australia	102	14.3 GWdc	~7.2 GW	~$22B / yr
UAE and Saudi Arabia	~20	12 GWdc	~6 GW	~$18B / yr
Chile	—	8.5 GWdc	~4.3 GW	~$13B / yr
Japan	86	6.6 GWdc	~3.3 GW	~$10B / yr
France	425	5.6 GWdc	~2.8 GW	~$8B / yr
Italy	111	3.5 GWdc	~1.8 GW	~$5B / yr
South Korea	—	1.9 GWdc	~1 GW	~$3B / yr

United States

43.5 GWdc

474 plants · ~21.8 GW Phase 1 AI · $65B / yr

Rest of world

137.6 GWdc

1,214+ counted plants · ~69 GW Phase 1 AI · $207B / yr

Phase 1 AI this set can host

~91 GW

181.1 GWdc · $272B / yr at $3M / MW-year

Phase 1 only — 50 MW of design-target 99.9% firm AI per 100 MWdc of screened aperture. Later phases are options on the same land and are not in this total. Floors differ by register: 100 MWdc in the U.S. and the Gulf, 50 MW in Japan and Korea, 30 MW in Brazil and Australia, 10 MW in France, Italy, Spain, Chile, and Germany. Germany, Spain, Chile, and Korea have no public plant-level GIS in this table, so site counts are blank. Qualifying plants, not signed contracts.

Watch list: South Carolina, Colorado, Wisconsin, and Alabama — fewer large aged plants, still worth a look.

Nominate a site See the conversion case