WSS: Cryptocurrency infrastructure operators continue shifts to data centre sector as neocloud and hyperscale leasing shows no signs of slowing down
One of the recurring themes in the sector these days continues to be the transition of cryptocurrency firms to the data centre colocation and HPC sector. These organizations have the raw land and energy resources that are required for large-scale hyperscale and AI builds, and they are actively building out purpose-built facilities that are often greenfield developments, with the flexibility to scale and adjust for the rapidly changing requirements that are out there. We tracked some recent developments in this space over the past week. TeraWulf continues to lease capacity and disclosed updated leasing numbers, with over half a gigawatt of IT load committed to the likes of Google and Fluidstack. TeraWulf has also acquired some brownfield sites that it plans to build out more capacity on. The business has progressed quickly and TeraWulf is now reporting revenue from the data centre business, with an expected uptick in the next few quarters as customers begin to build into the capacity. Another one of these companies is Galaxy Digital, also publicly listed. Galaxy Digital is expanding capacity in west Texas and is on-boarding anchor customer CoreWeave. CoreWeave has committed to over 500MW of capacity.
Neocloud leasing, needless to say, is a steady driver of demand for the capacity being built by these cryptocurrency turned data centre operators. The neoclouds are growing based on the demand they are seeing from hyperscalers and burgeoning AI models. And in the past week, CoreWeave confirmed it would provide cloud infrastructure services for Anthropic. This is reported to be a multi-year and multi-million dollar deal. Anthropic is a new customer for CoreWeave as it continues to diversify its customer base and reduce exposure to a single large hyperscale client. While Anthropic will be running on NVIDIA chips with CoreWeave, it is actively building out a multi-silicon strategy. In recent weeks, Anthropic confirmed a new partnership with Google and Broadcom that will see Google work with Broadcom to produce TPUs that will end up running on Google Cloud, allowing Anthropic to consume directly as a cloud infrastructure service. AWS is also adding to its silicon options and Cerebras will deploy its inference-oriented chips on the AWS cloud, while doing the same with Oracle Cloud. Anthropic continues to view AWS as its primary cloud infrastructure provider and will be sure to consume Cerebras technology as well. And AWS continues to add GPU capacity, confirming recently that it has committed to procure 1m GPUs from NVIDIA.
The sector continues to see heathy demand and expansion activity is ongoing. But there are still concerns about demand being overstated and a possible bubble, but at the same time, there are various supply constraints that make building out capacity challenging. The conflict in the Middle East adds another significant wrinkle. War can obviously cause economic instability that impacts overall demand for infrastructure. But the conflict we are seeing right now also has significant implications around energy and supply chains, that if disrupted or causing significant cost increases to raw energy resources, could have a meaningful impact on demand. Furthermore, a lot of the capital coming into the AI world is from the Middle East and there is a reasonable expectation that things may slow down until there is less uncertainty around the region. We take a closer look in some recent commentary and evaluate some of the risks.
or