Another smart play by Nvidia, it joins forces with data center developer Cloverleaf.

Nvidia is taking a minority stake in Cloverleaf, which secures land and utility power for future data centers, while Cloverleaf will use Nvidia's DSX platform to plan power, cooling, networking, compute and facility design together much earlier.

i.e. Nvidia is is moving upstream and getting involved before a customer even builds the data center, increasing the chance those scarce powered sites are physically designed to accommodate Nvidia's full hardware and software stack rather than choosing the compute layer later.

GPUs can arrive long before grids, substations and data-center shells are ready, creating the risk that expensive accelerators sit idle while infrastructure catches up.

Cloverleaf operates at that bottleneck, working with utilities and developers to turn land with prospective electricity access into sites that data-center operators can actually build on.

By putting NVIDIA DSX into Cloverleaf’s site-design process, NVIDIA can influence choices that become expensive to reverse later, including power delivery, cooling and networking. A future operator could still install another accelerator, but that vendor would inherit infrastructure choices already made around NVIDIA’s stack. NVIDIA can therefore shape the building before it wins the hardware order, creating switching friction without formal exclusivity.

Cloverleaf puts NVIDIA closer to an input chipmakers cannot manufacture faster: grid-connected, buildable land with a credible path to utility power.

Cloverleaf has sold more than 7 GW of powered land and has a pipeline exceeding 10 GW, giving NVIDIA access to unusually scarce development inventory.

This news follows NVIDIA's $1.5B SB Energy investment and a deal worth up to $3B with Lancium.

The pattern is so clear, NVIDIA now explicitly calls land, power and shell a deployment constraint and says exceptional sites can host multiple generations of its computing systems.