The Five Hurdles To Clear In The Hyperscale Race
Experience has created great insight.
ESD designs and consults for three of the top five hyperscale clients worldwide, and four of the top five colocation/wholesale providers. With that said, this experience creates great insight as to the general requirements of today’s hyperscale tenants in a colocation world. While we would never divulge our clients’ secret sauces, there are similarities from one to another that need to comply to each hyperscale tenant.
Utility cost: So many of the large hyperscale companies look to reduce utility cost, and therefore look for lower utility rates which are often in tier II or tier III markets. Large scale hyperscale users often locate data centers in remote areas and negotiate rates based upon MW installations ranging from 40MW to over 200MW. Due to the volume, they are able to get a reduction of rates which collocation companies may not be able to negotiate.
This creates three problems for the colocation/wholesale suppliers. The first is remote local utility companies have probably not heard of the colocation/wholesale provider, and therefore may be hesitant in a sizeable reduction; whereas today’s top five hyperscale tenants are very well known, and the local authorities are happy to announce their presence in the region. Secondly, there is little to no exit strategy in these markets which makes building a speculative data center financially risky. Finally, there may be hefty reserve capacity costs due to stranding unused power. This will skew a pro forma model and would be difficult to finance.
Scalability: Hyperscale tenants need to be scalable quickly to support their business users. While a catcher block system is very scalable, it’s typically installed at 1.5- to 2 MW increments. Other hyperscale tenants can take down larger chunks of capacity ranging in the 6 to 12 MW increments. This type of installation is more fit for a distributed redundant system. As with most collocation operators, use of static switches (STS) is the preferred design.
Renewable energy: Four years ago, Greenpeace went after the hyperscale compute companies and graded each based upon carbon consumption. None of them liked it. Since then hyperscale companies have created programs to either purchase renewable credits, or build their own windfarms, etc. Google is by far in the lead in this area, while others have either developed a program or are actively investigating renewable energy.
While renewable energy may not be required by the hyperscale tenant, the option to have it may be the difference in winning the lease. Iron Mountain recently announced that all of their data centers are 100% renewable energy which may give them an edge against competitors. The larger scale wholesale providers may have a harder time making that statement, but often are buying renewable energy credits when available.
PUE: The hyperscale data center tenants/owners, have studied PUE inside and out for the last three years. Within each study, a total cost of ownership (TCO) has been conducted to evaluate operating cost, as well as operator/maintenance cost. While geographic region plays into effect, many of the hyperscale operators are moving toward direct evaporative solutions. The PUE’s range from 1.18 to 1.3, and the reduction in water usage is more and more enticing.
Speed to market: Hyperscale moves fast, and so do collocation/wholesale providers. The companies that are successful in the hyperscale space depend on supplier management chains leveraging bulk purchasing. The most effective method to get to speed-to-market is developing vendor purchasing programs (typically two vendors per long lead — generators, switchgear, UPS, etc.) to quickly deliver equipment for fast track construction. The successful collocation/wholesale providers create purchasing programs to inventory equipment thus reducing construction times. In many cases construction can be reduced from six months to three months, giving the colocation/wholesale provider a competitive advantage.
Another method of speed-to-market is to always stay out ahead of built construction of 3 MW. This calls for a speculative approach that may require inventory (built white space) to be constructed that creates a higher risk for the provider. High risk, high return.
WHAT TO EXPECT NEXT
As within any industry, higher volume means lower prices. While the local market may be $145 per KW, the hyperscale tenant will negotiate a lower than market price per KW. This usually means a 5% to 8% reduction in rent.
One of the areas that makes sense to support the hyperscale tenant is a change in approach to delivery. Each hyperscale tenant has different requirements, and approaching the project may result in a build-to-suit leaseback. This approach is very appetizing to the hyperscale tenant because in many cases they end up negotiating upgrades or additional features into the data center lease. A build-to-suit directly addresses their needs, and is easily negotiated.