OCOLO’s fearless leader, Tony Rossabi, reports back after TD Securities Equity Research’s Annual Communications Infrastructure Summit that, as usual, he was able to glean many illuminating insights over two days in Boulder, Colorado into what’s happening across the broader digital infrastructure sector, and, most important, what’s next for what we’re most focused on here at OCOLO: data centers.
Coming right after Michael Elias and his team’s companies under coverage reported Q224 earnings, the conference was perfectly timed to offer a read on the sector at mid-year, and lay out some scenarios for the rest of 2024 and beyond. Today, the OCOLO team will share some highlights from the keynote fireside chat sessions that Michael Elias and Gregory Williams, CFA summarized following the conference close. There’s some pretty heady stuff here, and it’s not just because of the altitude of 5,400+ feet!
Takeaways from the DigitalBridge Keynote fireside chat with Marc Ganzi, CEO
Management updated its views on the GenAI opportunity set for data centers from ~38GW at last year’s summit to 50-80GW now, citing power infrastructure – specifically power transmission and not power generation – as the true hurdle to meeting demand-driven data center capacity needs in the US.
Management is allocating ~80% of its greenfield capex across its portfolio today to data centers to meet exploding AI infrastructure demand, versus ~10-15% to fiber and ~5% to towers. Brownfield capex is more opportunistic and balanced at ~30-35% data centers, ~25% towers, ~5% for small cell/RAN infrastructure and ~25-30% for fiber, as it seeks another fiber deal in 2024 with a “couple” in the pipeline.
Management highlighted the ongoing disconnect between public and private market valuations, given how much private capital remains on the sidelines. Tower companies range from ~16-22X for public transactions vs. ~22-23X for privates. Management cited private data center trade multiples at “mid-20s to low-30s.”
Management described a “once in a lifetime opportunity to build [digital] infrastructure,” citing enormous and accelerating capex. Hyperscaler AI infrastructure buildout capex was ~$150B in 2023 and on pace to hit ~$200B in 2024 and $250B in 2025.
While acknowledging the current environment may be over-hyped, management pointed to an important distinction between this cycle and prior infrastructure buildouts: this time, data centers are largely pre-leased versus largely speculative building in the last round.
Takeaways from the CyrusOne Keynote fireside chat with CEO Eric Schwartz
Management described hyperscale demand as “strong,” noting that its largest customers expect to grow their footprint “substantially” with “largely the same” mix of lease vs. self-build strategy vs. five years ago.
Management pointed out that the hyperscalers seem to have figured out how to leverage third-party data center operators in managing incremental growth and reminded the audience that in prior generations of technology “we historically underestimated every one of them.”
Management sees enterprise demand as “robust,” given ubiquitous adoption of AI for its range of internal and external business impacts and benefits, joking that there is no CEO “taking a pass” on AI.
Management pointed to the supply side of the equation as the bottleneck for hyperscalers’ ability to meet demand, noting as well that planning cycles have lengthened as along with lead times and supplier delays. Not surprisingly, the biggest pain points are in large-footprint, contiguous capacity. The power crunch is definitely influencing site selection and management does not see self-generation through vertical integration as optimal.
CyrusOne strategy remains: 1) meeting customer expectations and demand both operationally and developmentally, 2) delivering and exceeding return expectations given the capital intensive nature of the business, and 3) creating a company that lasts by delivering high-quality customer service. Geographically, the company remains focused on its current markets of the US, Europe and Japan. Management expects the ability to deliver against the aggressive needs and demands of AI workloads (e.g., density, cooling) to be a key differentiator for the sector for the foreseeable future.
Takeaways from the Zayo Group Keynote fireside chat with COE Steve Smith
Following a 4.5 year journey of integrating 50 acquisitions and fortifying its leadership team, Zayo Group is back on track to being an organic growth story. While in the later innings, management still sees demand for cloud-related apps and cloud onboarding as robust. Demand – as well as prices – for both dark and lit services remains strong among hyperscalers, providing a strong organic growth foundation for new GenAI fiber business to build upon.
Management sees AI fiber as the most significant trend they’ve seen in many years, with the first phase connecting to Training Data Centers. Next, management expects Enterprises to get more active with the pivot toward network demand at Inference compute data centers, then move toward the Edge with latency-dependent applications. As with the cloud rollout years ago, management believes AI will change the way we work in an irreversible way.
On M&A, Zayo Group is taking a breather to focus on deleveraging, bolstering liquidity, exploring the ABS market and executing its organic growth trajectory, but expects to pick back up again in the next 18-24 months to fill any gaps it sees.
Takeaways from the Vertical Bridge Keynote fireside chat with Alex Gellman, Chairman
Management remains confident on long-term growth trends in US towers, despite what Michael Elias describes in his note as “more than a year of a 5G spend hangover slowdown.” While timing remains elusive, listening to recent relevant rhetoric from companies like American Tower, DISH Network, T-Mobile and Verizon, management believes it’s a matter of “when not if” the densification phase of the cycle begins.
Management believes another catalyst could be a potential spectrum crunch in a few years if mobile demand continues to relentlessly increase without meaningful sub-6 GHz spectrum coming to market. A crunch like this, coupled with future GenAI applications, could drive a long-awaited return to small-cell growth.
On M&A, management also referred to the pervasive public vs. private spread in multiples being a deterrent. Public provider multiples have been cut in half, while privates haven’t see the capitulation they have in the past. Management has no desire to increase leverage and will raise senior debt then flip it to the ABS market if they need capital.
Good stuff coming out of Boulder, and more proof of the snowballing growth of GenAI driving an avalanche of demand for data centers globally (see how I did that??)! Next, we’ll relay broader takeaways from the conference the Communications Infrastructue team at TD Securities Equity Research shared following the conference. Spoiler alert: expect more discussion of GenAI…
Many thanks as always to the Communications Infrastructure team at TD Securities Equity Research for sharing these conference takeaways: Michael Elias, Gregory Williams, CFA, Cooper Belanger and Anton Rinnert!
