BẢN CHÉP ÂM THANH 🔗
Stephanie: Hello everybody, I'm Stephanie Price, the telecom and data center analyst here at CIBC. Thank you for joining me. Today I'd like to spend some time discussing one of the most important themes emerging across technology and infrastructure markets: data centers. We recently launched coverage of the US data center sector, including initiations on Equinix and digital realty. While artificial intelligence may be the catalyst capturing the headlines today. The reality is that the demand for digital infrastructure has been building for years, and we believe we are still in the early innings of a multi-year cycle. In many ways, data centers have become critical to the digital economy. Every cloud application, every streaming service, every enterprise workload, every AI model, and every digital interaction ultimately relies on physical infrastructure. Data centers are where the computation happens. Historically, the sector's often been viewed through a real estate lens. Investors tended to focus on occupancy, leasing spreads, and development activity, much like they would with office buildings or industrial properties. Today, we think the sector should increasingly be viewed as critical infrastructure. The market is no longer simply leasing space. It is providing power, connectivity, computational capacity, and increasingly the foundation upon which AI services are delivered. That distinction is important because infrastructure businesses typically benefit from long-lived demand drivers, high barriers to entry, and significant competitive advantages for incumbents. And those dynamics are becoming increasingly visible in data centers. Let's start with demand. Whenever we discuss the industry, the first question is typically: how much of this is really about AI? The answer is quite a lot, but not everything. AI accelerating demand. But it's not creating demand from scratch. Cloud adoption continues to expand. Enterprises continue migrating workloads out of corporate server rooms and into third-party facilities. Digital transformation remains ongoing. Data volumes continue growing. The average enterprise today processes substantially more data than it did even five years ago. Cloud growth also continues, and AI sits on top of all of these trends and accelerates them. What makes the opportunity especially compelling is that we see multiple secular growth vectors operating simultaneously. It's relatively uncommon to find an industry benefiting from cloud growth, enterprise digital transformation, network demand, and AI-related compute requirements all at once. That combination is helping drive one of the strongest demand environments the sector has experienced. To understand the opportunity fully, I think it's important to distinguish between AI training and AI inferencing. AI training is the process of creating large language models in other AI systems. These workloads often occur in large, purpose-built campuses containing massive GPU clusters and
Load full transcript
are very power-intensive. Training has attracted significant investor attention because of the enormous scale involved. McKinsey expects AI training to grow at a 35% kegger through 2030. Our view is that inference ultimately becomes just an important and potentially more important than training over time. Inference occurs whenever an AI model is actually used, such as when a user asks a question of ChatGPT. So every ChatGPT query, every AI generated search result, every enterprise AI assistant, every customer service chat pot. All these are influence workloads. And unlike training, which happens periodically, inference happens continuously. AI inferencing could potentially double between 2027 and 2029 as AI adoption continues and becomes embedded within products and business processes. From a data center perspective, inference has implications that are particularly interesting. Many inference workloads benefit from being located closer to end users. Latency matters. Whether you're running a financial application, a healthcare application, or a customer service platform, response times are increasingly important. We believe this creates advantages for operators in dense metro market footprints and highly connected ecosystems. That observation becomes especially relevant when discussing Equinix. But before we get there, It's worth discussing the other sub side of the equation, supply. On the surface, high demand should simply attract new supply. It's normally how markets work. But data centers are different because of the critical inputs, power. Across major North American markets, access to electricity has become the defining industry constraint. Developers are increasingly discovering that identifying land is relatively straightforward. Securing the power. Is not. Utilities face growing interconnection cues, transmission infrastructure takes years to develop, permitting timelines continue to extend. Transformer shortages remain a challenge. In several major markets, operators are being quoted delivery dates years into the future. As a result, existing data center operators possess a significant competitive advantage. Companies with existing campuses, established utility relationships, and already secured power have become increasingly valuable. One of the simplest ways to think about the sector today is that powered land has become scarce, leading to a higher pricing environment to lease data center capacity. Leasing activity remains robust, development pipelines continue to fill, and renewal spreads are strengthening. Data center vacancy rates are falling to decade lows as rents rise. Now let's get into the specific data center operators. We'll begin with Equinix. Equinix is our top pick in this sector. The company occupies a unique position within the global digital infrastructure ecosystem. Most people know Equinix as a data center operator, and while that's technically correct, we would argue the company's most important asset isn't merely its real estate footprint. It's the ecosystem that resides inside those facilities. Equinix has spent decades building connectivity density. Networks connect to cloud providers, cloud providers connect to enterprises. Enterprises connect to partners, customers, and service providers. Over time, these connections create network effects. This dynamic helps explain what Econix generates meaningful interconnection revenue and experiences relatively strong customer retention. Once customers become deeply integrated into the platform, moving elsewhere becomes more complicated, leading to a stickier customer base. As a result, its business model exhibits many infrastructure characteristics while also benefiting from network effects more commonly associated with software businesses. Another reason we like Equinix is its positioning around AI inference. Many investors immediately associate AO with hyperscale campuses and massive GPU clusters. Those opportunities clearly matter. However, inference demand may increasingly favor facilities located closer to population center and enterprise users. Equinix's global footprint is particularly well suited to that environment. The company operates in a metropolitan markets around the world. and maintain dense interconnection ecosystems within these locations. As AI usage expands, we believe that metro market footprint becomes increasingly valuable. The second company we launched on is Digital Realty. While Econix is associated with interconnection and enterprise ecosystems, Digital Realty provides greater exposure to hyperscale and AI infrastructure spending. Digital Realty's platform spans enterprise co-location, wholesale deployments, hyperscale capacity, and connectivity services. This diversified approach allows the company to participate across multiple customer segments. Digital Realty also has a strong development pipeline. The company controls a substantial inventory of future development opportunities with roughly 9 gigawatts of future development capacity and 1.4 gigawatts under construction. The development pipeline is primarily driven by large scale products, with 90% of the 1.4 gigawatt expansion for facilities with capacity greater than 25 megawatts, and 63% are preleased. Importantly, we are also seeing evidence that customers are increasingly willing to commit to capacity earlier in the development lifecycle. This behavior reflects the broader supply constraints affecting the industry. When customers worry about future availability, they become more inclined to reserve capacity in advance. For operators, this creates greater revenue visibility and reduces development risk. Another reason we're constructive on digital realty relates to pricing. In recent years, the industry often faced questions around commoditization. Today's environment looks very different. When supply becomes constrained and power availability becomes scarce, existing capacity becomes increasingly valuable. We're already seeing
evidence of stronger pricing across portions of the sector. Renewal spreads have improved, and DLR increased its full year renewal spread guidance by 250 basis points to 9% to 11% with Q2. Development projects are continuing to attract interest, and operators remain disciplined regarding capital allocation. These factors support continued earnings growth and healthy returns on invested capital. Stepping back, one of the most interesting aspects of the investment case is that the current opportunity extends beyond near-term AI enthusiasm. Even if AI enthusiasm moderates, many of the underlying demand drivers remain intact. Cloud adoption does not disappear. Enterprise digital transformation does not stop. Data volumes continue expanding. Software applications continue to become more compute intensive. In many respects, AI is accelerating trends that were already underway. That gives us confidence that the opportunity extends beyond a single investment cycle. Interestingly, valuations in the sector do not reflect the accelerated growth opportunity, with DLR and Equinix trading roughly in line with their 10-year historical averages. Unlike the Industrials Index, which is trading two turns above its historical average given the physical infrastructure buildup, data centers have not seen a similar lift. Applying a similar re-rate to the data center names, guess that's to our price targets on DLR and Equinix. Of course, no investment thesis is without risk. One risk would be a meaningful slowdown in AI related demand. A second risk would be the dramatic improvement in compute efficiency that reduces infrastructure requirements. A third risk relates to power. Ironically, the same power constraints that support the industry's favorable supply demand balance can also limit future growth. Operators ultimately need additional power to develop new capacity. If power delivery timelines continue extending, development schedules could be pushed out farther. Finally, data center operators remain capital intensive businesses, and interest rates and financing costs continue to matter. Investors should remain mindful of those factors. So, where does this all leave us? Data centers have evolved from a niche real estate category into one of the most important infrastructure sectors globally. The industry is benefiting from multiple secular growth drivers at the same time. AI is accelerating demand, supply remains constrained, power availability has become a competitive advantage. And industry leaders are increasingly positioned to benefit from strong leasing activity and improving pricing dynamics. Within this landscape, Equinix remains our top pick in the space because of its differentiated interconnection platform, network effects, and attractive positioning around AI inferencing. At the same time, digital realty offers compelling exposure to the hyperscale expansion, AI infrastructure spending, and one of the industry's largest development pipelines. The broader message for investors is that data centers are no longer simply supporting the digital economy. They are becoming the digital economy's essential infrastructure. And from our perspective, that creates a compelling long term opportunity. Thanks for listening. I hope you found the discussion helpful, and I look forward to speaking to you again soon. If you are a CIBC client, you can view the related reports, Data Centers, the Next Infrastructure Supercycle, Equinix, Position for Growth, Initiating at Outperformer, and Digital Realty, Hyper Growth, Initiating Coverage at Outperformer, by visiting the CIBC Research website located at ResearchCentral.cibccm.com or by reaching out to your CIBC sales representative.