The Capital Call

The Opportunities and Challenges of the Sub-Sea Cable Market for Digital Infrastructure Investors

September 7, 2021

My guess is that when you ask most mainstream institutional investors what they think digital infrastructure looks like, in actual physical form, they’ll probably trip off a list of visualizations including mobile phone towers, data centers, and finally, everyone’s favorite, hard hat workmen laying urban cable – in recent years they might even mutter about satellite dishes and low orbit constellations. But I’d wager that very few would get excited about sub-sea cables which is peculiar because one of the most successful IPOs of recent years in this space is that of D9 Digital Infrastructure is in large part, currently, a sub-sea play via its takeover of Aqua Comms. This successful and profitable business owns a bunch of assets including trans-Atlantic cables including America Europe Connect-1 (AEC-1), America Europe Connect-2 (AEC-2), and CeltixConnect-1 (CC-1). It’s also just announced an investment in a new cable into the middle east out of Europe.

If I were charitable, I suppose one could forgive the markets ignorance of sub-sea by way of taxonomy. Perhaps investors pigeonhole sub-sea cable as part of a wider fiber play which consists of a number of segments including long-haul (terrestrial and subsea) linking countries and cities together, as well as metro rings (the backbone network within a city) plus customers favorite, FTTP, (running from the metro ring to houses and businesses). And while we’re on that taxonomy point. one might also assume that terrestrial fiber might be a more attractive option because…well…it’s on land! That makes the process of laying cables and getting permissions and worrying about technology upgrades that much easier. So, if as an investor you’ve bought into the idea of all those exponential growth curves in data usage, better to focus on terrestrial and not be too bothered with all that tricky stuff involving water.

I say tricky because sub-sea sounds more than a tad challenging, and I’m not just talking about the obvious challenges of laying a cable in the deeps as well as repowering said fiber every 20 to 30 kilometers. Terrestrial cables typically have complicated rights of way issues to navigate. And if all that isn’t bad enough, experts such as Thor Johnson at D9 warn that backlogs and waiting times for the construction of new sub-sea projects is getting longer all the time. Whereas waiting for the cable supply and then installation would typically take two to three years, Johnsen reckons its now closer to three to four years and growing all the time.

Lurking in the background though is a more practicable challenge: the corporate structure of the sub-sea industry. Traditionally, the funding model for sub-sea has involved big consortia of telco businesses and carriers. The end result is a legacy network where many of the old cables, especially in the all-important Atlantic circuit, are nearing retirement or in need of extensive modernization to allow greater capacity.

That said, despite all these obvious challenges, sub-sea as one segment of the fiber market seems to be growing fast, as are most other digital infrastructure platforms (to be fair). Many of the most important sub-sea cables (at least in terms of capacity) have been built within the last five years. And some routes already look busy.

Take for example the all-important trans-Atlantic routes. At the last count, there were 18 cables going across from the US or Canada to Northern Europe (and especially the UK). D9 owns two of the more well-known cables AEC 1 and AEC 2 but many of the other owners have familiar names such as Vodafone, Tata and Google. Which of course reminds us of an essential fact – that sub-sea markets tend be highly concentrated in terms of ownership, much more so than long haul terrestrial fiber.

The majority of digital data carried via sub-sea today tends to be from a handful of giant companies such as Google, Amazon, Microsoft and Facebook. Rather than relying on the traditional model (telecom consortia) for access to subsea cables, these organizations are deploying their own sub-sea cables. They are doing so to cut costs and expand control and, as a result, the scale and pace of this deployment is significant. According to one industry insider, at the moment more than two-third of digital data moving across the Atlantic is on private networks owned by Google, Microsoft, Amazon or Facebook.

This all leads to one important point: that the sub-sea business is increasingly a scale game played by large operators who can a) afford the capital overlays and then b) continuously upgrade to allow for ever more data traffic.

Both of these trends can be glimpsed through the example of Google. According to, Google currently has either a stake or privately owns 14 undersea cables,  mostly in consortiums run by the likes of Aqua. Many of the most important span the Atlantic but the data giant has also expanded into other routes such as the Middle East-Africa-Asia route which until recently did not have a Google (or Facebook/Microsoft) cable on it, so one was announced. Soon thereafter, Google announced a project to traverse the Mediterranean and then wrap around Africa.

The technological upgrade path is also clear. Google’s first undersea cable, the Unity system, as part of a consortium across the Pacific, was ready for service back in 2010 and had eight cable pairs with system capacity of just 8 Tbps. The most recent cable, Equiano, which is 15,000 kilometers in length and connects Portugal to South Africa, has 12 pairs but a gigantic capacity of 240 Tbps.

Given this race to scale (and the need to modernize) and link up vast global networks, it’s arguably something of a surprise that there are still niche players such as Aqua in the business, competing against rivals such as Telxius (part of Telefonica) and GTT.

D9, for instance, through its Aqua business, has investments in a number of transatlantic lines: AEC 1 is their first cable with six fibers with four sold to a variety of corporations while another pair is leased, leaving one available pair. D9 also runs AEC 2 across the Atlantic plus AEC 3 which is imminent, coming on shore in Cornwall, UK.

The revenue stack for these independent operators is varied. Some pairs are leased on long term 15-to-20-year deals with the OTT Big Data plays; most of the FAANGs (Facebook, Amazon, Apple, Netflix, Google) seem to be Aqua customers. But mixed in with these (probably lower margin) revenue streams, there’ll also be shorter duration deals in the one-to-five year range for data hungry customers, especially those in the second tier. For example data hungry players such as Akamai and Disney.

This increasingly complex ecology of infrastructure players – traditional carriers battling it out with Big Data vendors and independents playing all sides – becomes even more interesting in certain geographies. One insider tells me that the most interesting competition at the moment is either in shorter haul interconnector cables (for example, UK to the Nordics) or into specific regional markets, with India topping most lists.

And there’s one final observation on this competitive ecology that worth noting: “Amazon and AWS don’t seem to be as active as we’d expect in this market currently. They seem happy to buy capacity rather than lay their own cables,” one industry consultant tells me. Given Amazon’s vaulting global ambitions, I can’t imagine that state of affairs will continue for much longer.

That mention of Amazon though should act as a warning to bigger institutional investors interested in this space. Some investors I talked to admit they like the space but as a result of this competitive landscape tread cautiously around it. Says one: “We’d argue that a sub-sea fiber on a heavily competed route with few customers, poor contract terms and looming competition would naturally be less interesting than a long-haul fiber route with strong customers on good contracts, solid growth potential (new laterals and new customers) and limited opportunity for a competitor to find rights of way.”

Specialists like Aqua/D9 will of course disagree and point to the extraordinary growth in demand across the planet, with new routes opening all the time. They also remind us that even the old routes will need upgrading soon, but one is left wondering whether the sheer scale of capex needed for the next phase of sub-sea modernization might be a challenge too far even for deep-pocketed institutional investors.

Welcome to Space 3.0 Digital Infrastructure

August 2, 2021

As a life long sci-fi fan, I have to admit that the news that Branson and Bezos have in the last few weeks properly kickstarted the space tourism model left me a tad cold. I freely admit that I am terrified of heights and fast moving things so I’m probably not the target market but my skepticism was more nuanced than that. Sure, space tourism is exciting but I think it is a side show to a much bigger story, namely the brand new space-based digital infrastructure model that is taking shape above us.

To understand the real revolution underway lets first rewind. Space 1.0 was largely state run, funded by militaries and bureaucrats and comprised huge, bulky satellites. Space 2.0 took off in the last decade and spawned a fleet of listed satellite owners of communications platforms in search of specialist niche markets. As David Williams, co-founder of Avanti Communications Group plc, the United Kingdom’s first start-up satellite operator and now boss of satellite focused quantum cryptography form Arqit freely admits, this model had its flaws which became increasingly obvious to stock market investors by the middle part of the last decade. These Space 2.0 stalwarts were, he admits, “cost plus” platforms, forced to focus on broad products such as maritime comms to generate enough cashflow to pay the bills. We all know how unsuccessful that model was compared to functionally specific satellite-based platforms such as Sky in Europe and Dish in the US. These operators made their money out of using the satellites for a much bigger content first product. For Sky, a satellite firm, space-based costs were probably just a couple of per cent of total costs.

Space 3.0 is radically different again. The cost of sending a kilo via a launch outfit such as SpaceX is now one hundredth of the cost in the 1980s and there are now over 150 other rocket launch companies competing for business launching thousands of low earth orbit satellites for every purpose imaginable.

Media and investment commentators such as yours truly tend to focus on one itsy bitsy aspect – rural broadband in the developed world. Here in the UK, Space X has just launched its open beta for the Starlink product and from many subscribers I have talked to, the feedback is positive so far. Latency is massively reduced; speeds are as promised and at the moment there’s no big challenges around data usage (though I’ll bet that will change very very soon). Another operator OneWeb, rescued in part by Her Majesty’s UK government, also looks to be making solid progress and has just announced what I think could be a very fruitful partnership with incumbent telco BT to operate village level satellite hubs that can redistribute the signal to nearby rural houses.

But even this focus on rural broadband – obviously, a crowd pleaser for rural politicians seeking infrastructure spends – is I think missing the bigger picture. The real story is that the multitude of these low earth orbit constellations are pioneering a new product set – satellite data products. Rural broadband is one part of a broader mix of products that ranges from autonomous cars to land mapping.

One way of gauging this new digital infrastructure market is through the lens of recently stock market listed UK space fund Seraphim, which raised £180m to invest in this space using a venture capital approach. They also helpfully run a Space VC investment index. Cynics will snort that this an early stage side show with its obviously high valuations but I’d suggest that the headline numbers for Q1 2021 below tell you something is happening out there.

  • $8.7bn invested in last 12 months
  • $2.7bn invested in Q1 ($2.6bn in Q4 20)
  • 68 deals closed in Q1 (highest since Q1 18)
  • $850m biggest deal closed in Q1 (SpaceX)
  • $46m average deal size in Q1 21 (vs. $50m Q4 20)
  • 11 space-related SPAC mergers announced
  • $7.2bn funding committed to these SPAC mergers.

A better way of understanding the Space 3.0 model is to focus on investment stories, or more specifically some of the businesses in that Seraphim portfolio. Perhaps the most traditional example is a firm called AST Group which has just reversed into a SPAC with the ticker ASTS. The best way of describing this business model is that they provide cell towers in space.  

A more radical product comes in the shape of Iceye which provides 25cm resolution day or night of every metre of the earth every hour. That presumably requires a huge amount of data that needs to be processed back on earth in huge data centres? Or perhaps not.

Circling the earth are another constellation of satellites from a firm called D Orbit which is currently trialling something called Nebula, an on-demand, on-orbit cloud computing and data storage service. According to a recent press release this new product “features an intelligent automation SpaceCloud iX5-100 radiation tolerant computing module by specialist Unibap, will demonstrate a range of innovative applications for advanced geospatial Earth Observation (EO) and Space Surveillance and Tracking (SST) applications using sophisticated, Artificial Intelligence/Machine Learning (AI/ML) algorithms for extremely low-latency decision support”.

In layman’s terms, this a data centre in space. Sci-fi nerd heaven. All we need now are space elevators!

Stepping back from the company-by-company details, a new model emerges for space-based infrastructure: it’s just a tool by which new products can be slotted into a digital framework. So just as investors have pumped money into everything from cloud-based firms, or Google and its search technology through to enterprise as software business models, we have low earth, lower cost space based constellations as adjuncts to new products. With all the same bonkers valuations.

That said, despite those pie in the sky (!) valuations, I’d maintain there are some very earth bound investment implications for infrastructure investors.

The first, as evidenced by Seraphim’s successful fund raise and their wider index data, is that space is moving out of the shadows of early stage VC and becoming a mainstream sub asset class. Even fairly ‘traditional’ infrastructure investors such as Cordiant admit that they’ve been looking in detail at these Space 3.0 propositions. And that’s all because these are fast turning from ideas into market ready products.

I’d also counsel caution for the space based rural connectivity argument. In countries such as the UK our incumbent telcos have certainly been slow off the market, but I think fibre will end up becoming the default solution for between 90 and 95 percent of the homes in the developed world, eventually. It might take five years to get there but I find the up front expense of solutions such as Starlink and the high monthly running costs unattractive.  I also think they’ll run into capacity issues as that final one to five percent of difficult to reach homes swarm on to the space based networks.

In the developing world, by contrast, I think space base broadband solutions are viable and exciting but I’d wager that upstart space telcos will face intense competition from the Big Tech giants. I can absolutely imagine services from Facebook or Netflix delivered directly by satellite – cutting out the telco – for a monthly fee by Bigtech.

But all of these markets are a side show to what I think is the real big deal. Autonomous cars. I’d be a rich man for every telco and digital infrastructure payer who says the future of 5G is autonomous transport. I believe them until I think what happens when my self-driving car skips out of a 5G network zone. Do I trust the major mobile telcos to have 100 percent connected within a decade to 5G? Of course not and until that happens and I can be guaranteed seamless connectivity all the time, in all national geographies, I’ll avoid autonomous cars. The solution? Look to the heavens for seamless global connectivity with a small receiver built into the car. Is this imminent, not yet but IT IS the future.

Switching on Advisors to the Digital Infra Opportunity

July 6

In my humble opinion, one of the biggest challenges facing digital infrastructure businesses as well as funds is to find a way to get to talk to advised clients and their advisors. Institutions by and large understand the space as do many sophisticated private investors, especially the very active ones. But advisors tend to lump these digital assets into an alternative box alongside more traditional infrastructure assets and then treat with some suspicion, the suspicion of the ‘new’.

Yet there are signs that is changing and the example of a recently launched UK fund of funds might offer some useful clues as to what advisors might want from the digital infra sector. Gravis is a successful UK based infrastructure house that has also built up a fund of funds business that sells heavily to advisors. Its idea is simple. All these alternative investment ideas structured as funds (or opcos) are difficult to research for most advisors, so you need someone to build an overlay that finds the right funds (and single businesses) and then assembles them together in an income-oriented fund of funds. Over time its existing infrastructure and listed property equities fund have accumulated pretty big sums of money – as have rival offerings from peers – and now its launched what it thinks over time will be its biggest fund ever: a digital infrastructure equities fund of funds which invests in towers, data centres, fibreoptic networks, logistics warehouses. Or as the manager describes them physical structures, which are all tangible, and all have a bit of concrete poured in them, as well as contractual leases, producing high cashflow predictability.

In this universe, they’ve identified around 120 listed, global digital infrastructure companies although it’s interesting to note that within their definition the number of funds and businesses has doubled in terms of numbers over a decade and has increased more than six-fold in terms of aggregate market cap. More specifically Gravis analysis reveals that this universe has increased in market cap terms from £73.5bn in 2010 to £461.3bn in 2020. As for returns, between 2011 and 2020, returns have averaged c.18.9% per annum with volatility of 15.43%. From this growing universe, the Gravis team led by Matthew Norris has identified 29 businesses and funds for the portfolio which is now live and growing steadily in size.

Graphic : The growth of the digital infrastructure universe

Source: Gravis Capital Management, Ltd, 2021

As you’d expect the big sales pitch is one we’re all familiar with – as Norris observes “we’re living in the fourth industrial revolution” but the first key message from this fund is that it includes a very heavy helping of tech-enabled logistics park operators alongside the more expected tower cos and fibre operators.

According to Norris “If you go back in time, especially in the UK, data centres are actually born on logistic parks.  One of the biggest owners, possibly the biggest owner of data centres in Europe is probably property business Segro, and that data centre is on the Slough trading estate. The Slough trading estate gave birth to Segro 100 years ago and then as good fortune would have it, they found themselves close to London as well as a big fibreoptics cable from the Atlantic plus it helps that there’s access to power.” Segro has its own power station.

It is easy to see why logistics park operators will appeal to professional clients. You have classic property-based assets, with long leases and lashings of technology. As a result, the Gravis fund is exposed 45% to logistics businesses with the remainder in classic digital infra: 25% in data centres, 25% in tower cos and 5% in another bucket that includes fibreoptic networks and battery storage.

The next big lesson is that Gravis, like many advisers I’ve talked to, likes tower cos and are overweight exposure to this sector compared to their universe weights. Norris echoes those who’ve called these businesses “the best business ever”.  He likes the fact that the steel towers being put up last 25/50 years, and that according to one of the tower co operator’s maintenance capex on each is just 900 dollars per tower per year – “it’s a great asset, long-life asset”. And there’s also the increasingly obvious densification drive coming out of 5G especially in the US “where you have carrier neutral towers and the mobile network operators go to the tower codes and put their dishes on.  We are likely to see co-tenancy on towers, so competing operators on the same tower- that’s great news for the owner of the tower.

The next useful insight is that the digital infra space needs to be more vocal about its intrinsic valuation strengths as an equity asset class and not run scared of those who argue the sector is overpriced and a bond proxy. On the multiples question, Norris points to businesses that have contractual cashflows, are earning steady money, paying dividends, and increasing those dividends on average between 2 and 3%. Unlike say classic property REITs , which produce higher yields, the digital infrastructure space has very obvious drivers of huge growth “over many, many sequential years, so what you might have to sacrifice in yield you should make up for in terms of dividend growth. “

As for the cacophony of voices that say these assets are bond proxies and will falter as interest rates pick up, Gravis reminds us that these are not in fact anything remotely like fixed-income assets. “We’re talking about growth income here,” observes Norris “and there are two types of growth that you get from owning a data centre and a tower co. One is your contractual rental growth.  And then the second growth element is actually releasing up more space, growing the rental income, by hanging another dish on the same tower.  So, my response would be I hear what you say but this is not fixed income, this is growth income, plus market growth.”

I also sometimes detect another variation on the bond proxy argument which is that although there are obvious growth opportunities, most digital assets do not have explicit inflation protection, unlike say classic public-private partnership infrastructure assets. It’s a fair observation but I’ve always felt it is a misguided one. Sure, there may not always be explicit CPI agreements in place but as Norris reminds us “as long as the income can grow in line or faster than inflation then it’s going to be inflation proofed. With tower cos and data centres their top line should be rising faster than inflation just because of the growth characteristics and their cost base should be rising at or below inflation, you should see some level of operating leverage coming through.  So, I think those two sub-sections will perform very well.”

My own hunch is that these characteristics of the asset class are, in reality, fairly well understood but there’s another risk that is I think less so. Security.

I’ve visited more than my fair share of data centres and towers over the years and the first question I always ask is a security-related one. Cybersecurity tops many lists but in reality, that’s actually more of a risk for the tenant rather than the physical infrastructure owner. But physical security is very much a concern for everyone, especially the real asset owner. And here Norris at Gravis spies an opportunity, rather than a risk.

“I think that creates a barrier to entry, I think that’s why you and I wouldn’t be very successful at setting that kind of business up because you have to have a track record of building very secure, reliable centres. The ones that I’ve visited, especially the ones that have government servers in them, they literally do have those bollards around the centre to prevent terrorist attacks and when you enter the data centre you go in through one of those man trap scales that weigh you on the way in and weighs you on the way out.  So, absolutely, security is an issue but also it creates a competitive advantage.  If you are an established player in the market, you have the reputation. “

This brings us to the last key insight – ESG. I see it topping more and more financial professionals lists of concerns, and not always for the right reasons. ESG has become akin to a gateway drug that encourages all sorts of slightly inchoate concerns, mostly but not always based around energy efficiency.  According to Norris “we have done a lot of research on this and, it’s much more efficient to have servers in a data centre, in a bespoke environment. Here at Gravis, we still have a server in the corner of our kitchen.  That’s bonkers.  That’s not the right environment to have it, so we waste energy cooling that server in the corner of a kitchen.  If all businesses like ours put their servers into dedicated data centres, less power would be used…..  I think there’s more work to be done proving the point that the place for your server is a data centre because you’ll use less energy”.

This I think is a key message. Rather than focus all the attention on digital infrastructure operators own ESG policies – important though they are – also try and remind investors of their own obligations, their own inefficiencies that make the problem much worse!