Clichmont Builds, Doesn’t Rent, AI Compute
As demand for artificial intelligence compute surges, Clichmont is taking a contrarian approach: owning and controlling the physical infrastructure—data centers, power, cooling, and connectivity—rather than renting GPU capacity from hyperscalers. CEO Alexis Cathalifaud argues that infrastructure ownership provides control over the economics of compute, not just access to chips.
“GPU access gives you compute; infrastructure ownership gives you control over the economics of compute,” Cathalifaud said. “When you rent GPU capacity, you inherit someone else’s pricing, availability, power constraints, networking architecture, deployment schedule, and margins.”
The company’s thesis hinges on the idea that GPUs depreciate quickly, while power-ready data-center capacity—land, grid connections, substations, cooling, fiber, and permitted megawatts—remains valuable across multiple generations of accelerators. Securing 10,000 GPUs is one problem; securing the tens of megawatts of reliable electricity to operate them is another.
Why Megawatts Matter More Than Chips
Cathalifaud believes the real bottleneck for AI infrastructure is energy, not GPUs. “A GPU without reliable power is just expensive hardware sitting in a rack,” he said. “The real competition over the next decade won’t simply be for GPUs—it will be for megawatts.”
Clichmont evaluates sites based on power availability, time-to-power, grid reliability, cooling architecture, and scalability. The company operates a solar-powered facility in Alicante, Spain, and is building a new site in Bodø, Norway. Northern Norway offers efficient cooling and strong energy fundamentals; Alicante provides a different energy profile with solar integration.
“Chips can be shipped around the world. You can’t ship 100 megawatts,” Cathalifaud said. “The compute ultimately has to go where the energy is.”
Competing With CoreWeave, Crusoe, and Lambda
Clichmont is up against established players like CoreWeave, Crusoe, and Lambda, some of which are public or heading there. Cathalifaud doesn’t think their models are wrong—they proved AI compute is a massive market. But he believes the durable bottleneck is infrastructure, not GPUs.
“GPUs change every generation. The durable bottleneck is the infrastructure required to run them—power, land, cooling, and connectivity,” he said. “In a market where everyone is chasing chips, we’d rather own the place where the chips have to live.”
Rather than competing for GPU rental volume, Clichmont aims to be one of Europe’s most efficient independent AI infrastructure operators, serving enterprise AI, HPC, and private compute.
The $CLAI Token: Skepticism and Utility
Clichmont has a token, $CLAI, attached to its infrastructure business. Cathalifaud acknowledges skepticism. “A token shouldn’t exist just because a company operates in AI,” he said. “If $CLAI were simply a financing wrapper around our data centers, I wouldn’t consider that a compelling reason to create it.”
Instead, $CLAI is intended as a digital economic layer for on-chain participation, treasury activity, and community governance. However, Cathalifaud sets a high bar: “The physical infrastructure has to exist independently of the token, and the token has to demonstrate real utility independently of speculation.”
The test, he says, is whether the token does something useful, transparent, and measurable that couldn’t be accomplished with a normal database or conventional corporate structure.
The Hardest Part: Scaling Physical Infrastructure
For software veterans, the biggest surprise is that physical infrastructure doesn’t scale at software speed. “In software, if demand doubles, you can often provision more capacity quickly. In a data center, every additional megawatt has a physical dependency—grid capacity, transformers, switchgear, cooling, fiber, permits, construction, and hardware,” Cathalifaud explained.
One missing component can delay an entire deployment, and mistakes are expensive and difficult to reverse. “You can’t patch a badly designed 50-megawatt electrical system overnight,” he said. The real skill is sequencing capital, power, construction, and customer demand so they arrive at roughly the same moment.
The Biggest Risk: Capital Intensity and Timing
Cathalifaud identifies the biggest risk as capital intensity combined with timing. Building infrastructure means committing significant capital today against assumptions about future demand, power economics, and technology. A rental model offers flexibility; ownership is a long-duration decision.
“The biggest danger isn’t simply spending too much—it’s building the wrong capacity, in the wrong place, at the wrong time,” he said. Clichmont’s objective is to control strategic infrastructure while remaining flexible around technology, ensuring buildings, power, cooling, and connectivity survive multiple GPU generations.
Three-Year Vision: Efficiency Over Scale
Looking ahead three years, Cathalifaud doesn’t expect Clichmont to be the biggest player. “CoreWeave and Nebius have enormous scale and access to capital. Trying to replicate them would be the wrong strategy,” he said. Instead, he wants Clichmont recognized as one of Europe’s most efficient independent AI infrastructure operators, with real operating assets, secured power, and high-density GPU capacity.
“Our advantage has to come from being disciplined about where we build and what we own,” he added. While CoreWeave and Nebius build hyperscale AI clouds, Clichmont aims to occupy a different position: a focused owner-operator of compute-ready infrastructure in strategically selected markets.
Investors should watch Clichmont’s ability to secure power purchase agreements and grid connections for its Bodø and Alicante sites, as well as the development of $CLAI’s on-chain utility. The company’s next milestone—likely a capacity announcement or token utility update—will test whether its build-first thesis can deliver control over AI’s physical bottlenecks.
Comments are closed.