Credit Suisse’s Sami Badri at CAPRE’s NYC Data Center Summit: When it Comes to CapEx, Keep an Eye on DRAM, Server Prices

CAPRE’s 7th Annual Greater New York Data Center Summit was held on April 5, 2018 at Westin New York at Times Square. Above: Sami Badri, Senior Analyst, Credit Suisse presents the keynote address: Cloud Enablers Boosting Enterprise Colocation Growth.

NEW YORK, NY — According to Sam Badri, experts have significantly underestimated data center capex spending. But once you pop the hood, this massive influx of capital makes a lot of sense. Badri is a Senior Analyst for Credit Suisse, and a frequent contributor at CapRE Data Center Summits. At CAPRE’s 2019 New York Data Center and Cloud Infrastructure Summit, Badri had a lot to say about capex – where the industry is, how it got here, and why. Check out a previous CAPRE Insider Report to learn more about the first two questions. In this report, we cover the third.

Sami Badri, Senior Analyst, CreditSuisse

“What drove the 2018 capex strength? Where did the money actual go?” asked Badri. “There are four real drivers that got us to 48%. One is that GDPR took effect in May of 2018, even though things are very straightforward, you’d be surprised at how many companies – especially hyperscalers – were positive in spend, just to make sure that they understood the rules of the game very clearly, and then deployed their topology accordingly.”

The second factor is that tax reform effect in January of 2018. “All of a sudden every corporation had lower tax rates and better accounting treatments toward their capex deployment,” he explained. “All of a sudden that’s a double tail wind….that was double the motivation for corporations to spend.”

Furthermore, there are two forces that flew a bit beneath the surface in 2018, according to Badri. “DRAM pricing is a component – for memory that goes into a lot of storage equipment…it basically goes into every single piece of equipment except for networking,” he began. “Some networking gear does have it, but most specifically, it’s servers and storage. If you think about a data center, about 10%-15% of a data center is networking gear, connectivity etc. Essentially 80%-85% of it is storage and server gear.”

“The hyperscale capex tidal wave that was being deployed, was essentially flowing into equipment that was consuming this specific product,” Badri gathered. “Or for this specific component. DRAM pricing doubled from Q3 2016 into Q2 2018. So what accentuated that hyperscale capex growth, had a lot to do with this component specifically.”

New Event SquareBadri then urged the audience to consider servers. “You had a step up in the actual price per server of about 33%. Which by the way is extremely-counter intuitive because equipment and components, over time, prices should go down, not up, on an average selling basis,” he shared. “The reason for that is scale and technological advancement. For something like hardware, it makes no sense that you’d have this kind of, I’d describe it as scalability of price and flexibility. So you have these two big drivers essentially funnel a huge amount of price inflation into products. Those products are acquired by capex buyers, and that flowed into data centers.”

Next, Badri moved onto hyperscalers and global distribution. “Hyperscalers were trying to distribute deployments globally – in multiple regions, multiple metros. What they’re trying to do is achieve is very robust availability zones by key metro areas,” he shared. “The problem with the old, centralized model is latency. You had latency problems, and on top of that, if you wanted to transmit from a core data center into a cell phone, say on the ground, there was a bit of a problem routing through the internet. If this was a big workload, it was going to cost you a lot of money in carrier broadband.”

And this led to the globally distributed Cloud model. “This in essence is exactly what AWS started building about ten years ago, and has been maturing this process today in North America. So what does this look for data centers?” he rhetorically asked. “A complete dispersion across multiple markets. This led to a complete step up in intensity in the first half of 2018, completely synchronized to the DRAM and server spike production.”

Finally, Badri ended the introduction to his remarks with a prediction about construction. “The construction coming online was about 7% at the end of 2018,” he indicated. “This is most likely going to dip down to about 6.5%, which is a very healthy level from where we stand today.”

For more from Badri at CAPRE’s New York City Data Center & Cloud Infrastructure Summit, check out a previous CAPRE Insider Report: “Sami Badri Explains Growth of Hypercale CapEx Spending at CAPRE’s NYC Data Center Summit.”