WSS: Bulls and bears debate as large deals are signed and neoclouds continue to expand both organically and through M&A
The sector continues to push forward and the past week saw developments that speak to how AI is going to drive hyperscale cloud growth forward, while underscoring how the two are deeply intertwined. The idea that AI is something new and different, separate and distinct from hyperscale cloud, has always been wrongheaded and is driving the bearish sentiments that continue to be unwarranted.
Showing how the scale and scope of AI infrastructure is set to drive hyperscale cloud growth, Oracle disclosed a huge new contract that should quash any doubts that are out there. Oracle Cloud has signed an undisclosed customer, reported widely to be OpenAI, that will spend $30b annually on Oracle Cloud services. This number is several times larger than the current ARR for Oracle Cloud, which still has a significant level of RPOs that will continue to add to its revenue base.
Meanwhile, elsewhere in the hyperscale world, Meta is building out infrastructure to support its AI initiatives and is reported to be raising the same number – $30b – to fund its data centre builds. Meta has already raised its CapEx targets and is looking to take things to another level.
Even as hypescalers continue to build and bring in business, bearish sentiments remain. Hedge fund manager Jim Chanos has warned of the sector’s impending doom more than once in the last few years (and yet seems to continue to be wrong). And there are others that believe a bubble is forming and ready to pop. The DeepSeek training model efficiencies realized earlier this year and the various hyperscale pullbacks caused plenty of concern and are pointed to by the skeptics. They also wrongly compare the current period to the dotcom bust over 20 years ago, which is kind of like comparing apples to oranges. But the fact is that these developments are misunderstood and actually reflect normal dynamics when it comes to infrastructure procurement. And when it comes to things like DeepSeek, these are actually positive developments that will help the sector develop going forward. We take a closer look at the latest bearish sentiment and provide some commentary.
Elsewhere in the sector, the GPU cloud space continues to expand aggressively and data centre colocation providers are benefitting from the demand. The latest deployment saw Nebius stand up infrastructure with Ark Data Centres in the UK. The scale, scope and cost of GPU-based cloud infrastructure has forced operating companies to become even more creative and CoreWeave has been exactly that, fronting the CapEx required to build data centres for colocation operators. Recognizing this is part of the operating puzzle, CoreWeave created a new C-level position dedicated to this task. CoreWeave remains busy and in the past week deployed Dell-based servers in its data centres with the latest NVIDIA chips. There was reporting that CoreWeave was again trying to acquire Core Scientific and that deal was completed early in the recent week. We’ll have more detailed commentary next week, but the acquisition looks to be a way for CoreWeave to bring significant data centre capacity into the fold, while alleviating leasing overhead and fronting CapEx. Most importantly, this gives CoreWeave a data centre construction capability and will have implications for its data centre leasing efforts and strategy.
Finally, the last week saw a number of developments come out of the LatAm region. Microsoft opened a new cloud infrastructure region in Chile and Google is building a subsea cable connecting Chile to Australia. The LatAm region remains in an early stage of development. Brazil is reported to be the next location for a ByteDance data centre and Omnia, a new hyperscale data centre operating platform targeting multiple LatAm markets, is being assembled and set to hit the market. All are in pursuit of the long-term growth opportunity.
or