Power Shortage Worsens Data Center Bottleneck

Oracle declared force majeure on September 24 for its 2.45GW mega AI data center 'Jupiter Project,' which was under construction in New Mexico.

This occurred due to an extension of the construction period caused by permitting issues related to the power supply contract, making Oracle, as the tenant, liable for financial costs and lease payments.

Consequently, the gas pipeline operation date was pushed back by nine months to February 2027. New Mexico cited air pollution from natural gas power generation and concerns about depleting local water resources as reasons for opposition.

As power shortage issues become prominent, proactive investments by power supply chain companies may expand, leading to a caution against over-interpreting an issue with a specific weak link as a crisis for the entire system.

Confidence in AI infrastructure investment for the second half of the year is still considered positive.

As data center construction demand spreads, Dy CEO explained that labor and material bottlenecks will worsen until 2028, and the Oracle case is merely a temporary issue at the permitting stage.

Data center construction demand and intensity are actually expanding and increasing, and there are no instances of perceived demand decline at Energy Transfer sites.

The CEO of 'Dy,' a data center construction company, mentioned that labor and material bottlenecks are expected to worsen until 2028, indicating strong market demand.

With natural gas and electricity demand hitting decades-highs, Kinder Morgan's utility pipe contracts continue irrespective of individual project delays.

The surge in natural gas and electricity demand presents the greatest expansion opportunity in decades, and there have been no cases of utility pipe contract suspensions despite individual issues.

This is a very positive sign, considering the conservative stance of European companies.

Demand for core power infrastructure equipment, such as large transformers, remains strong, and considering the typically conservative stance of European companies, this is interpreted as a very positive sign.

Such 'noise' can actually reinforce the need to simultaneously pursue more projects to expand power infrastructure.

They engage in investment activities not only by forming funds but also by directly owning infrastructure assets or acquiring stakes.

As part of this plan, an initial $10 billion fund was established, with $5 billion already raised. Brookfield has shown active movements, including acquiring energy storage company I-Power and renewable energy company Boralex, and increasing its partnership with Bloom Energy.

Additionally, through its subsidiary Westinghouse, it is promoting the development of Small Modular Reactors (SMRs), focusing on securing power supply capabilities, which are crucial for AI infrastructure.

Future data center power demand in the US over the next 10 years will reach 100GW, but the available supply is only 30GW. This is the background for the spread of the BTM model, where power is generated directly on-site, and AI infrastructure capital expenditure forecasts have been raised from the previous $7 trillion to $10 trillion. In particular, data center construction costs have surged more than fivefold to $50 million per MW.

Future data center demand in the US over the next 10 years will reach 100GW, but the currently available supply is only 30GW, indicating a severe power bottleneck is expected.

To address this situation, the spread of the BTM (Behind The Meter) model, which generates power directly on-site, is accelerating.

Blackstone is investing $25 billion in data center-specific gas power generation and grid construction, while KKR is raising a $19 billion fund to deploy $75 billion in digital infrastructure and power sectors.

Blackstone is also investing $25 billion in expanding data center-specific gas power generation and grid infrastructure, with several major private equity firms, including Energy Capital Partners and TPG, announcing multi-billion dollar investments to build out the digital infrastructure ecosystem.

Investment scope is expanding from relying solely on big tech's cash flow to financial securitization. With the active inflow of private equity, demand for data center asset securitization is becoming visible, which is expected to lead to a revaluation of the entire power supply chain value chain.

Investment methods have moved beyond relying on big tech's cash flow to allow for an expansion of investment scope through financial securitization.

Institutional investors only enter AI data center projects when sufficient calculations and estimates are available, even while accepting long-term uncertainties of 3-7 years or more.

This is based on the conviction that data center demand will not disappear within 2-3 years, and the current market situation where several companies are moving simultaneously serves as the basis for this conviction.

Peter Thiel focused on energy supply chain investments by acquiring stakes in Vistra, a power generation and retail supply company, and X-Energy, a next-generation SMR developer.

As data center locations spread beyond Virginia to cheaper areas around Michigan and Ohio, he targeted power infrastructure beneficiaries by investing 10% in each of the four utilities (AEP, FE, DT, CMS) in these regions.

This is a strategy of investing in companies related to power construction, the actual bottleneck of AI infrastructure, and investments in Bitcoin mining companies are analyzed as focusing on the potential for conversion to AI infrastructure rather than Bitcoin itself.

The power shortage for AI data centers is not just a negative factor but a critical bottleneck diversifying AI-related investment areas, including fuel cells, natural gas, power grids, and data center construction.

While the stock market still harbors significant concerns about an AI data center bubble or construction due to rising interest rates, companies on the ground, confident in AI's growth, are increasing their sectors and numbers.

This indicates that AI data center construction is a process of rebuilding social infrastructure beyond simply securing GPUs, and investors need to broaden their scope of investment consideration by focusing on the expanding range of companies driven by AI growth.