WSS: Hyperscale growth acceleration building as hunt for capacity continues to intensify
The past week saw earnings season pick up pace and we took a closer look at the results coming from Meta. Meta continues to outlay CapEx into building data centres and raised the bottom end of its CapEx guidance, implying that it has more visibility and continuing to expect a significant increase this year. Among the hyperscalers, Meta is the one that does not have a raw cloud infrastructure business, but that is very possibly going to change soon. On the earnings call, Meta provided a few more details about how Meta Compute is evolving and there is credible reporting that Meta is working on a possible cloud computing deal with frontier lab Anthropic. In the hyperscale segment, Oracle and Google have already reported, and the past week also saw Amazon and Microsoft report. The results followed the pattern of hyperscale cloud growth acceleration, driven by AI demand that is moving through the enterprise space and manifesting in large AI and cloud infrastructure deals, which is driving top-line results. Amazon saw quarterly y/y revenue growth climb from 28% y/y to 37% y/y, while Microsoft saw quarterly growth for Azure reach 43% y/y. We will have more details next week, but the early returns are clear and cloud infrastructure is seeing meaningful growth acceleration.
Anthropic continues to drive infrastructure consumption across the ecosystem. As mentioned above, it is reported to be working with Meta, and recently signed a long-term lease with data centre operator TeraWulf. Anthropic also confirmed a new agreement with AMD that will see it deploy gigawatt levels of capacity housing both GPUs and CPUs provided by AMD. Anthropic’s growth requirements are spanning across deployment models – cloud and data centre – and is extending to multiple chip vendors.
Like NVIDIA, AMD is making a more concerted push to build out a service provider ecosystem, and in the past week cloud and neocloud providers Vultr and TensorWave confirmed plans to deploy some of AMD’s rackscale solutions built on its newest lines of Instinct GPUs. Microsoft also confirmed it would do the same. The moves speak to ongoing chip supplier diversification efforts, the need to tap into silicon capacity and the need for both GPUs and CPUs.
The race to build AI infrastructure, needless to say, also requires significant quantities of capital, and in the past week we saw Broadcom launch a financing investment vehicle in partnership with Apollo and Blackstone to help the growth requirements of frontier labs like OpenAI and Anthropic. Meanwhile, Blackstone launched an APAC-focused infrastructure investment platform with a focus on India, where it did a data centre deal just a few months back, acquiring Lumina CloudInfra.
In other developments, we saw Megaport partner up with Wasabi for cloud storage infrastructure. Wasabi will become a white label supplier of cloud storage to Megaport’s cloud infrastructure service. The growth in neocloud and webscale platforms is also accelerating alongside hyperscale, but these providers have less resources to develop technology in-house and often will make bets on third party vendors in areas where they have gaps. Cloud storage is an area that is seeing this kind of focus as we saw with the recent partnership between CoreWeave and Backblaze.
At the end of the month, we publish our monthly insights, analyzing some of the most notable developments and trends. We take an early look at earnings season, delve into execution patterns, Anthropic’s movements and the ins and outs of legacy migrations that are still happening. The sector continues to push into the future, but legacy environments are still running and making its way to the outsourced infrastructure world, with AI is acting as the latest accelerant, helping push things forward and supplying more upside growth potential into the ecosystem.
or