Time to Restart the Clock
Is the data center industry concluding its collective multi-year freak-out? To be clear about what I mean by that - end-user buyers, operators, and real estate developers have come together in a panicky gestalt where data center projects - real, speculative, and fake, have been thrown at customers to try to solve the worst supply problem that the industry has ever seen, caused by a confluence of unexpected AI/ML demand, poor land banking, and unclear demand plans.?
At this year’s PTC event, we’re seeing somewhat of a pivot - we’ve made a shift from an unknown future customer demand plan meaning that customers would buy anything in sight to the hazy grey of T-24 months resulting in a rational reticence to purchase 2028 and further data center assets.?
What this means for the industry is that the age of demanding signed leases for campuses where a shovelful of dirt hasn’t moved is drawing to a close. Some degree of speculative development is again becoming necessary. Delivery timelines in the T-18 to 24 month timeline are becoming necessary in order to close many deals. That means things like zoning, power, network, site prep, and possibly tilt-up shells must be in delivery prior to lease execution.?
The question is whether the buyers - and the universe of them has been expanding far beyond the original four hyperscalers - are willing to pay for this acceleration. They are going to have to, whether they realize it or not. We’ve been living on artificially cheap financing and we haven’t had to order long lead-time equipment without a signed lease. That may not be the case now.
领英推荐
So, how do we make this work? General John Pershing famously said, “Infantry wins battles, logistics wins wars” - as goes a war, so goes the data center. A far more fungible supply chain where utility transformers, generators, switchgear, gas turbines, and other long lead time equipment must be site independent, purchased far in advance, and movable from site to site. This will help to compress the datacenter delivery timeline. Finding financial sponsors willing to speculatively fund the earliest phases of development, then exiting, is vital.? Isn’t everyone doing that? The short answer is that many of the top datacenter providers are. Some of them aren’t and seemingly none of the smaller guys.?
We’re also seeing specialized Supply Chain providers starting to emerge - Quantum Technology Systems and Integra Mission Critical have both leaned heavily into this concept. They cut across multiple disciples to help accelerate delivery while keeping costs steady.?
In any marketplace, there is always a balance between buyers and sellers. It has been a seller’s market for the last 18 months, since Machine Learning landed. It appears things are tilting back towards buyers as the market normalizes. This isn’t an unhealthy development for the market and it will reward sophisticated and experienced data center providers and developers over newer and less experienced market entrants.?
It seems like all of our clocks stopped in the Summer of 2023 as ChatGPT roared on the scene. The timeline is restarting now - be ready.
COO and Co-Founder, NYI | Co-Founder, Nomad Futurist | CEO, Critical Ventures | Podcast Host | Board Member | Digital Infrastructure Executive | 2023 iMasons IM100 Award Winner
1 个月Great insights as always, Daniel Golding and certainly reinforces what we’ve been discussing on Cool Vector Media! #enjoytheride
Well said Dan, it’s all about Supply Chain fungibility!
Comment #2: point #3/ Not sure about the market swinging back in favor of the buyers yet, only because some of the trends you highlight will actually reduce the amount of viable sites / operators, and therefore, limit supply. I do think a certain rationality has arisen from many parties in the ecosystem: buyers are getting more discerning about what projects / developers are viable and worth spending time on; developers – or the ones with experience and track record – are more discerning about which sites to pursue; and, utilities are altering their queue processes to separate the contenders from the pretenders by, for example, requiring cash up front or site control for a load study. There’s still an imbalance in supply and demand that won’t reach an equilibrium in the next few years, imo. Thanks as always for the thought leadership, Daniel Golding
Fantastic piece as always, Dan. I agree and would highlight / amplify a few of your points. 1/ tenants were never going to lease much further out that 24 months, because in my experience that's about as far out as they get the demand signals they use to build the internal business case to lease. So, agree with you: anything else was aberration that wasn't going to last. 2/ Again, agree with you that zoning, power, network, and some site prep should be in place before you even MARKET a site, much less expect to lease it. (continuing with an additional comment due to character limit)....
Spot on, sir!