Group · What we do
What we do
One vertically integrated, off-grid edge platform, answering all three parts of the same problem at once: the compute, the power it needs, and the link that reaches it.
The bottleneck
Three constraints meet in one place, and that place is the site.
Compute
AI and Web3 workloads want dense compute close to where the data is made. At the centralized core it is slow to reach and expensive to hold.
Power
Hyperscale sites face rising power costs and a dependency on ageing, congested grids — and an interconnection queue measured in years.
Connectivity
Rural and under-served markets do not have the terrestrial fibre backhaul a modern data centre is normally designed around.
The transition to an AI-driven economy cannot run on legacy grids and centralized fiber.
The case
Why, what, where, when and how.
- Why
- Compute demand is arriving faster than grids and fibre can be extended to meet it. Waiting in an interconnection queue is the default answer, and for an AI workload with a delivery date it is not an answer at all.
- What
- One campus that carries its own generation, its own backhaul and its own halls — specified together, so the power engineer and the network engineer are reading the same drawing.
- Where
- Four named US markets in the initial pipeline: South Oregon, Milwaukee, Pennsylvania and North Georgia. Every other geography on the map is an ambition, drawn differently and labelled as such.
- When
- A proof-of-concept build-out in South Oregon in the first quarter of 2027, with the next three centres behind it across 2027 and 2028.
- How
- Factory-built halls on a prepared pad, bridge generation while the ladder to geothermal and a fluoride-salt reactor is climbed, and a Ka-band gateway standing before the first trench is considered.
Ecosystem
Three categories of thing have to arrive on the same pad.
- Space segment
- Sites are built to host a Ka-band gateway, so backhaul is a low-earth-orbit path rather than a trench. The operator relationships behind that are listed, with their standing labelled, on the ecosystem page.
- Power generation
- LPG and hydrogen bridge generation for phase one, enhanced geothermal where the resource supports it, and a fluoride-salt reactor class on a later-decade horizon. Xalant holds no permit and has filed nothing.
- Compute infrastructure
- Liquid-ready halls and direct-to-chip cooling, specified around rack densities the fabricators can actually build. Fabricators are not named until they have agreed to be.
Who it is for
Who a campus like this is for.
AI training and inference operators who need capacity next to the demand rather than next to the grid
Web3 and high-performance workloads that price power before they price anything else
Landowners and site hosts with power, land or both, and no wish to become a data-centre developer
Local providers and municipalities who want the surplus capacity a campus creates to stay local
How it works
The four steps that make it one machine rather than three contracts.
Generate
Power is made on the site itself — bridge generation first, enhanced geothermal and a fluoride-salt reactor on the ladder behind it.
Process
That power is designed to feed edge data centers on the same pad, running high-density AI and Web3 workloads next to the demand.
Connect
Ground gateways carry the processed result back to the wider network — satellite first, so a site is connected before a trench is dug.
Yield
Surplus baseload power and spare gateway capacity are planned to be offered to the community around the site, any excess sold back to a grid.
The integration engineOne machine
Reference render
Talk to Xalant
Bring a use case. Xalant is designed to bring the site, the power, the backhaul and the compute as one contract.
Site and energy partners, enterprise and AI workloads, local providers and communities — in the four pipeline markets, or in a geography that is being targeted next.