OpenAI and Microsoft are embarking on an ambitious project to build a huge data center designed to house an artificial intelligence supercomputer that will potentially cost more than $115 billion, The Information reports. This US supercomputer, which OpenAI calls Stargate, is set to be the largest of several data center projects planned by the two companies over the next six years, with the hope of being operational by 2028.

The data center, which is expected to be funded by Microsoft, could end up being “100 times more expensive” than some of the largest data centers in existence. Stargate will be the Phase 5 system in a series of supercomputers that are being developed in stages, with Phase 4, a less expensive one, possibly being launched by 2026 in Mount Pleasant, Wisconsin. Stargate’s enormous energy needs, estimated at at least several gigawatts, are prompting companies to consider alternative energy sources, including nuclear power.

Implementing a data center of this scale will also be challenging, in part because the project requires many more GPUs in a single rack to increase the efficiency and performance of the chips than Microsoft has previously implemented. This also means that new cooling methods will need to be developed.

In addition, OpenAI and Microsoft will try to reduce dependence on NVIDIA during the project. According to sources, OpenAI wants to avoid using NVIDIA’s InfiniBand cables in Stargate, although Microsoft uses them in current projects. OpenAI wants to use standard Ethernet cables instead.

Because the Stargate project is in the early planning stages, many details remain undetermined, including its cost, which may be subject to revision. In addition, it is still unknown where this supercomputer will be located and whether it will be built in one data center or in several neighboring data centers.

OpenAI CEO Sam Altman will also be in talks with investors to raise funding for the initiative to increase global chip manufacturing capacity and expand AI capabilities.