For the first time in almost three years, the datacenter denizens grew faster than AWS, resulting in a modest tick back in their favor on the closely watched Platformonomics Repatriation Index™.
The PRI™ is a quick and dirty heuristic (people give me grief that it is “not a valid index” of the entire datacenter industry ¯\_(ツ)_/¯) that looks for the “repatriation” of workloads out of the cloud.
The Index is the ratio of the combined revenues of Digital Realty Trust ($DLR) and Equinix ($EQIX), the two largest datacenter operators, to the revenue of AWS, the cloud bellwether. The Index premise is that if repatriation is actually happening, it should show up on the biggest datacenter operators’ revenue.
In the first quarter of 2023, DLR grew 19%, Equinix 15% and AWS 16%, resulting in the datacenter duo “taking share”. But before we all clear our calendars for a blowout “repatriation is a thing” celebration, it is worth digging a little deeper.
A few observations:
MAGNITUDE: This is a small move in the Index, a shift of 1.6%. Hardly a tidal wave. The defenders of the DIY datacenter are still down about 7% in the last three years. And I have restrained myself by limiting the cloud comparison to AWS, who are the largest but slowest growing of the hyperclouds. Azure grew by 27% and Google Cloud by 28% in the same quarter.
PRICING: Equinix raised prices Jan 1, while DLR is bragging about their “strongest renewal pricing quarter since the early days of the pandemic” and their inflation-indexed pricing “accelerators” (they also brag about landing leases from hyperclouds, which begets Inception-esque repatriation). Thus some of the datacenter operators’ growth comes from price increases as opposed to a hypothesized wave of repatriated workloads. AWS, by contrast, is not known for its price increases. Due to tighter economic conditions, AWS has emphasized they are aggressively helping customers manage cloud costs at the expense of their own near-term revenue.
TOTAL REVENUE: AWS did $21.35 billion in revenue in Q1 while DLR+EQIX did $3.34 billion. It is hard to see a repatriation trend here. Because of the disparity in size, even a small amount of cloud repatriation should have a big impact on the datacenter operators’ revenue (taking 10% of AWS’s revenue would bump DLR+EQIX revenue over 50%). Instead, we see slowing AWS growth without a concurrent big pop in datacenter revenue.
INCREMENTAL REVENUE: AWS added $2.9 billion in revenue while DLR+EQIX added $476 million. AWS’s incremental revenue has been slowing, but it is hard to see a repatriation trend here. In fact, it is hard to see the datacenter guys at all relative to AWS. Usually my joke about being indistinguishable from the x-axis refers to IBM and Oracle’s CAPEX, but DLR+EQIX exhibit similar x-axis hugging behavior here.
THE POSTER CHILD (NOTE THE SINGULAR): Meanwhile, repatriots are still fixated on 37Signals (and their 20 servers). The repatriation poster child still hasn’t quite completed their much advertised move out of cloud (you can’t spell repatriation without the letters P and R when it comes to 37Signals). If repatriation was a big trend, one would expect hundreds of examples, not just 37Signals (and droopy Dropbox). Meanwhile, the march away from on-prem continues with large companies moving to the cloud (one suspects they each have more than 20 servers).
NEW AI WORKLOADS ARE NOT REPATRIATION: A reminder that repatriation involves workloads in the cloud moving out of the cloud. Building a new AI training or inference system that doesn’t run in the cloud isn’t repatriation. (Note building such a system requires you to get your hands on scarce GPUs, where you must stand in line behind the hyperclouds).
We will continue to monitor the universe for signs of repatriation, but the near-term forecast is for the surf to remain calm.