Key Takeaways
- The artificial intelligence boom is forecast to more than double data centre power consumption by 2030, creating a significant bottleneck and opportunity around energy availability.
- Specialised operators are emerging to meet this demand, with companies like GPUone (set to list as GPO) pursuing aggressive expansion plans by securing low-cost power in non-traditional locations.
- GPUone’s strategy targets a greater than fivefold increase in capacity to over 1 gigawatt by 2026, based on a model of acquiring power for approximately 3-4 cents per kilowatt-hour.
- Valuations for these high-growth infrastructure plays are demanding, pricing in flawless execution and reflecting a significant premium over established data centre REITs.
- The defining competitive advantage in the next phase of data centre growth may shift from access to capital towards access to secured, low-cost power contracts.
The relentless expansion of artificial intelligence is catalysing an unprecedented demand for data centre capacity, with some industry forecasts projecting that global power consumption from the sector will more than double by 2030. This structural shift is creating enormous tailwinds for the digital infrastructure ecosystem, moving the primary bottleneck from capital to energy and rewarding operators who can secure reliable, low-cost power. Against this backdrop, a new class of specialised, high-growth companies is emerging, exemplified by firms like GPUone, which aims to leverage power arbitrage to scale its capacity at a remarkable pace.
The New Digital Gold: Power Capacity
For years, the data centre narrative was dominated by location and connectivity, with primary hubs like Northern Virginia and Silicon Valley commanding premium valuations. Today, the script has flipped. The sheer energy required for training and operating AI models has strained the electrical grids in these established zones, making power availability the single most critical constraint on growth. A recent McKinsey & Company analysis suggests US data centre power consumption alone could surge from 17 gigawatts (GW) in 2022 to 35 GW by 2030, an expansion that existing grids are ill-equipped to handle.1
This has created a compelling opportunity for operators capable of thinking outside the traditional geographical boxes. By venturing into regions with abundant and cheaper power, often from renewable sources, these companies can offer capacity at competitive rates. This strategy not only circumvents the gridlock in Tier 1 markets but also aligns with the increasing demand from hyperscalers for sustainable energy solutions.
A Case Study in Aggressive Expansion
The model is perhaps best illustrated by the Canadian operator GPUone, which is set to become a public company through a merger agreement with Power & Digital Infrastructure Acquisition II Corp. (XPDI) and Nabors Energy Transition Corp. II (NBIS), eventually trading under the ticker GPO.2 The company’s strategy is not merely to build data centres, but to build them where power is cheap and plentiful, primarily in Quebec.
GPUone’s investor materials outline a plan that is nothing short of ambitious. Starting from an existing capacity of approximately 220 megawatts (MW), the company targets an expansion to over 1 GW by the end of 2026. This growth is underpinned by a business model focused on securing power at highly advantageous rates, reportedly in the range of 3 to 4 US cents per kilowatt-hour (kWh), and then leasing the resulting data centre capacity to clients with high-performance computing needs.
Projected Financial and Operational Trajectory
The company’s financial projections provide a clear picture of the expected growth curve, assuming successful execution of its expansion plans. While all forward-looking statements carry inherent risk, the figures illustrate the scale of the market opportunity management is targeting.
| Metric | Projected 2024 | Projected 2025 | Target Capacity (End of 2026) |
|---|---|---|---|
| Revenue (USD) | $109 Million | $244 Million | Not Disclosed |
| Adjusted EBITDA (USD) | $55 Million | $150 Million | Not Disclosed |
| Total Capacity (MW) | ~220 MW | ~550 MW | >1,000 MW |
Source: GPUone June 2024 Investor Presentation2
The pro forma enterprise value of the merged entity is cited as approximately $1.4 billion. Based on 2025 projections, this implies a forward EV/Revenue multiple of roughly 5.7x and a forward EV/EBITDA multiple of about 9.3x. These are not trivial figures; they reflect a market prepared to pay a premium for high growth and a strategic position in the AI infrastructure value chain, pricing in a significant amount of future success.
Investment Implications and Second-Order Risks
For investors, the rise of specialist operators like GPUone presents a different risk-reward profile compared to established data centre REITs such as Equinix or Digital Realty. While the latter offer stability and scale, their growth is more constrained by the challenges in legacy markets. The new breed of operators offers explosive growth potential but comes with considerable execution risk. Building out hundreds of megawatts of capacity on schedule and on budget is a monumental logistical and financial challenge.
The second-order effects are also worth considering. The primary risk is not a lack of demand, but a failure to secure the power contracts needed to fuel expansion. Any delay or renegotiation of these contracts could severely impact unit economics. Furthermore, as more operators adopt this “power-first” strategy, competition for favourable energy sources in untapped regions will inevitably increase, potentially eroding the arbitrage advantage over time.
A speculative but plausible hypothesis for the future of this sector involves vertical integration. The most defensible long-term position may belong not just to the companies that secure power contracts, but to those that own the power generation itself. We could see a future where data centre operators acquire or develop their own renewable energy assets—solar farms, wind turbines, or hydroelectric plants—to guarantee supply and lock in costs. Such a move would transform them from digital real estate companies into a new hybrid of technology and utility, fundamentally altering the competitive landscape.
References
1. McKinsey & Company. (2024, May 30). AI could drive $76 trillion investment in data centers, maybe, claims McKinsey. Data Center Dynamics. Retrieved from https://www.datacenterdynamics.com/en/news/ai-could-drive-67-trillion-investment-in-data-centers-maybe-claims-mckinsey/
2. PASCAL. (2024, June 20). GPUone, a Leading GPU-Focused Data Center Provider, to Go Public Through Business Combination with Power & Digital Infrastructure Acquisition II Corp. PR Newswire. Retrieved from https://www.prnewswire.com/news-releases/gpuone-a-leading-gpu-focused-data-center-provider-to-go-public-through-business-combination-with-power–digital-infrastructure-acquisition-ii-corp-302177651.html
3. @thexcapitalist. (2024, June 28). [Post regarding McKinsey data center demand forecast and NBIS guidance]. Retrieved from https://x.com/thexcapitalist/status/1806680459521360064
