Edge data centres @ the tower?

As content must get closer to the end user, so must data centres. From today’s large scale data centres in tier one and tier two cities, could future use cases require thousands of micro data centres dispersed closer to the network edge? And could towercos host those data centres at the foot of cell sites? If so, do towercos want to lease space to data centre companies, or buy/operate their own data centres? DataBank is one of the latest acquisitions of visionary infrastructure investor Digital Bridge – we spoke to DataBank CEO and serial data centre entrepreneur Raul Martynek.

The Future Network: Thanks for speaking to us today Raul. Please would you introduce our readers to the data centre market, to its convergence with the tower market, and to the concept of edge data centres.

The concept of edge data centres has come out of the evolution of the data centre sector.

The first incarnations of data centres occurred during the bubble era, but when the bubble burst a lot of empty data centres went bankrupt.

From 2001-2004 the data centre sector re-invented itself to house critical IT infrastructure for rising “Web 2.0” companies and corporate enterprises.  Many enterprises had built and were struggling to manage their own data centres, but increasingly they realised that it was better to outsource to a multi-tenant data centre operator, lowering the enterprises’ cost base and enabling them to focuson their core business.

In the U.S., the data centre industry was initially concentrated in five big tier one markets: the New York Metropolitan areas, Ashburn for Washington DC, Chicago, Dallas, and San Francisco / Santa Clara, which have all been key internet interconnection points from the outset. The principle being that content has to live on servers and storage, thus data centres should be close to the nodes (“on ramps”) of the Internet.

So now we have the rise of the concept of the edge. The rise of video has started to break the Internet. Delivery of over the top (OTT) content is increasingly carried via the Internet instead of via coax cable. Consider an Internet user in Ohio, who hits a piece of content that physically resides in Washington and California. It takes time to deliver that content as a function of latency. Now we cannot deliver content only from the five tier one markets – data has to move closer to the end user.

The Future Network: Where does DataBank fit in this ecosystem?

I joined Digital Bridge in December 2015 after selling my previous data centre business.

We wanted to differentiate our entry as an edge data centre play. By early 2016 we had our investment thesis thought through and just needed an expandable asset to invest in. We saw that in DataBank, which we acquired in July 2016. DataBank started in Dallas, a tier one market, but has expanded into several non-tier one cities including Kansas City, Minneapolis, Salt Lake City, Cleveland, Pittsburgh, Atlanta and Baltimore.

At the heart of our vision is this concept of bringing content and computation closer to the end user. Consider the U.S. data centre strategy of a company like Apple with all the information they have to deliver to the iPhone; health, photos and everything else up in the iCloud. Companies like Apple, Uber and Facebook are increasingly deploying infrastructure closer to population centres – instead of serving St Louis from Chicago, serving them locally from St Louis.

The Future Network: And how do you see the convergence of edge data centres and towers?

Co-locating edge data centres at existing towers is the logical conclusion as the bits and bytes continue to get closer to the end user. The idea may be a little ahead of its time now, but we’re keen to be a first mover!

Let’s consider a city like Minneapolis, with its population of around 2mn. A company like Amazon may want to put local infrastructure within 50 miles of Minneapolis, which was previously served from another more distant location. There are hundreds, probably thousands of macro towers in Minneapolis – with a cell tower we could get content from 50 miles to within half a mile of the user. At the speed of light, the difference between content travelling half a mile versus 50 miles is less than half a millisecond – with today’s technology use cases, there isn’t sufficient incremental benefit. But with autonomous vehicles, robots and AI replicating human activity – and requiring close to real time computation – content will need to be half a mile, not 50 miles from the end user.

Like carriers don’t want to just be a dumb pipe, tower industry stakeholders are also starting to recognise they provision more than just vertical real estate. Towercos and their partners provision connectivity, and that presents an opportunity to further monetise our networks and the end users on our network.

We created a cross company team within Digital Bridge to track these converging sectors, to leverage existing customer relationships, and to facilitate end users taking advantage of our unique architecture for delivery of content and bytes.

Digital Bridge are assembling a unique portfolio of companies; Vertical Bridge, Mexico Tower Partners and Andean Tower Partners – our tower plays; DataBank and Vantage for data centres; and ExteNet Systems for small cells.

Is small cell a competitor to edge computing? As fibre-fed small cells become even more numerous, converging fibre becomes a good place to put data centres – with links to thousands of small cells.

The Future Network: Am I right in thinking there have been no deployments of edge data centres at towers yet?

We’re at ground zero in 2018 – there have been no edge production deployments yet. But companies like ourselves, Vapor IO (which has partnered with Crown Castle), and Edge Micro are getting a lot of press. We’re all positioning ourselves to be perceived as a good partner of whomever the end user of this edge architecture is going to be.

The Future Network: That begs the question, who will be the end user of this architecture?

Most cell towers are effectively monopolies in providing connectivity to a given locality of end users. And an edge data centre represents the fastest way to deliver access. So the killer apps will be whatever puts the greatest pressure on latency. Autonomous vehicles, for example, generate 3TB of data per day from hundreds of sensors sampling many thousands of times per second. And that data may have to be stored for six months to investigate accidents – it will be heavily regulated.

Video is another use case: ultra HD, augmented and virtual reality. If that were incorporated into a car to aid the driver in understanding their physical environment, it would create a huge data need in real time (<5 milliseconds).

In order to really kick-start the convergence of towers and edge data centres, there has to be an economic business case to support the 1,000-1,500 users typically served by each developed market cell tower.

Those business cases may also come from industrial automation: robotics, IoT services over large geographies, or manufacturing IoT localised in a given facility, using drones and robots and RFID tags on mobile vehicles… this all gives rise to significant complexity, requiring ubiquitous coverage.

Consider the example of LinkNYC, now part of Google, which has transformed tens of thousands of dysfunctional payphones on sidewalks into thousands of fiberised monolithic kiosks with armoured flat screen monitors generating advertising revenue, but with dozens of sensors in the base: noise detection that can hear disturbances, even gunshots; chemical analyses, waste management. That kind of approach makes sense to create an IoT in New York; edge data centres are an analogous way of achieving this in a less dense metropolis.

The Future Network: Do you think tower companies will want to lease space at cell sites to third parties to operate these micro data centres at the network edge, or do you think the towercos will want to own the data centres?

The towercos haven’t decided yet which model to follow.

Towercos already effectively operate edge data centres today – they build a cabinet with a controlled environment and lease to carriers to put their network equipment in there. Conceptually speaking, that’s exactly what a data centre is.

Existing data centre operators aren’t running to towercos to secure a huge volume of edge data centre sites. We’re busy continuing to build larger scale buildings, that’s what our customers want at the moment.

Hence you’re hearing about alliances between data centre and tower companies, but it’s like a set from a Western movie: they’re just facades with nothing behind them.

Nonetheless, it would be foolish for towercos not to investigate opportunities in edge data centres. It all comes down to where content is going, and the changing requirements of subscribers and enterprises.

The Future Network: Talk to us about the commonalities in the business model between data centres and towers.

The data centre business model shares a lot of commonality with the tower model.

That said, we must still educate and get investors comfortable with the differences in competitive dynamics and customer dynamics. For example U.S. towercos have four primary customers signing 10-15 year leases, while in the data centre business we have 1,200 customers with an average contract length of 48 months.

Both and data centres are capitally intensive up front – data centres costing around US$10mn per megawatt. We might spend half of that up front to get the shell up and running – often with no customers up front – and a data centre burns more maintenance than a tower. So a data centre starts with a negative NOI, typically breaks even at around 25% occupancy, with a contribution margin of 80% or better after breakeven. At 75% occupancy and above, the model becomes extremely attractive, much like the tower model with a tenancy ratio above two. Revenue growth in the data centre business might be characterised as more gradual and less regular. At scale both are very attractive business models.

While the valuation of towers boils down to forecasting the lease up rate, similarly the valuation of data centre assets is a function of historical and forecast future leasing on a projected KW per quarter basis.

Perhaps the biggest difference is that churn is typically higher in data centres, around 8% per annum, versus half that in towers. Most the churn in the data centre market is related to the dynamism of the Digital Economy – consolidations, bankruptcies et cetera.

The Future Network: Talk to us about the commonalities in the investor pool attracted to data centres and towers.

Once upon a time towers weren’t considered infrastructure – now the listed towercos are Real Estate Investment Trusts (REITs). The five publicly listed data centre operators in the U.S. are also REITs. It is not a big stretch for an investor in a towerco REIT to get into data centres, and vice versa.

This is my third year at Digital Bridge, and the interest of traditional real estate and infrastructure investors in data centres is at its highest level over those three years.

Crown Castle is now putting a significant percentage of their investment capital into fibre – which is a vehicle for both towers and data centres.

The Future Network: How would you characterise the M&A pipeline in data centres?

Extremely active over the last five years.

Compared to towers, the data centre ecosystem is highly fragmented. In the USA alone there are 250-350 data centre operators.

Consolidation will happen – it just started later because the sector is younger. For example there are four companies consolidated into DataBank. There are lots of data centre entrepreneurs and investors looking to make timely exits, and lots of appetite to consolidate.

The Future Network: How would you summarise the current state of the edge data centre market.

We’re not there yet – the current model of data centres is meeting current requirements, and it generates economies of scale. In a future ecosystem of dispersed edge data centres at thousands of locations, we’ll need feet on the street everywhere.

Standing back from the hype, there remain considerable barriers to the adoption of distributed micro data centre architecture, and we have yet to determine whether the continued expansion of traditional cloud computing and colocation could obviate the need for edge data centres at the tower.

---

Introducing Raul Martynek, CEO, DataBank

Raul Martynek is perhaps the ideal person to talk about opportunities at the convergence between the towers and data centres. He currently serves as CEO of DataBank, which has 14 data centres in eight U.S. markets.

Raul is a seasoned C-suite telecommunications and data centre executive, most recently serving as a senior member of the team at diversified tower, data centre, small cell and fibre investors and operators Digital Bridge. Prior to this, Raul was Chief Executive Officer of Net Access, a New York metro based co-location and managed services company acquired by Cologix. Prior to that, Mr. Martynek was the CEO of three other telecommunications infrastructure companies including Voxel dot Net, Smart Telecom Co., and Infohighway Communications, as well as the COO at Eureka Networks.

---

Raul Martynek and Alex Gellman, CEO of Vertical Bridge, will be presenting on “Micro data centres at the base of communication towers to enhance data delivery through edge computing” at the 5th TowerXchange Meetup Americas, taking place on 20-21 June at the Boca Raton Resort and Club. For more information and to register, visit: https://meetup.towerxchange.com/event/3aaed1e4-9f49-42de-a4e5-c91e6164d4ea/summary.