Data Centre Investment Hits $580 Billion and Overtakes Oil Supply Spending for the First Time

Data Centre Investment Hits $580 Billion and Overtakes Oil Supply Spending for the First Time The IEA's World Energy Outlook finds AI-driven digitalisation is reshaping commodity demand from power generation through critical minerals to gas-fired backup. Global data centre investment will reach 580 billion dollars in 2025, surpassing the 540 billion dollars being spent on oil supply, according to the IEA's World Energy Outlook. The crossover is a striking marker of how AI-driven digitalisation is reshaping energy capital allocation. For the first time, the infrastructure to process data commands more investment than the infrastructure to extract the fuel that powers most of the world.⁴ That matters because every dollar of data centre capital creates downstream demand across multiple commodity layers. Power generation equipment. Gas turbines. Battery storage. Grid infrastructure. Copper for wiring. Critical minerals for chips and batteries. The AI supply chain is not a single market — it is a cascade of commodity inputs, each bidding up prices in sequence. The generation layer is already responding. Babcock and Wilcox landed a 2.4 billion dollar design-build contract with Base Electron for 1.2 GW of natural gas-fired power, driving backlog up 470 percent to 2.8 billion dollars. Shares closed at $14.54, up 129.34 percent year to date. A second 1.2 GW option is under evaluation and the global pipeline exceeds 12 billion dollars. Management guided 2026 EBITDA to 70 to 85 million dollars, roughly 80 percent year-over-year growth, excluding any data centre upside.¹ But generation investment is running ahead of grid investment. Spending on electricity generation has surged nearly 70 percent since 2015, yet spending on power grids has increased at less than half that rate. The mismatch creates bottlenecks that limit how quickly new generation — whether gas, nuclear or renewable — can actually serve data centre load. Battery storage firms in the US are seeing surging interest from AI data centres, Reuters reported, but lengthy interconnection queues and supply chains heavily dependent on China are hampering the industry's ability to scale rapidly.^4,6 The critical minerals layer adds another constraint. The IEA warned of growing vulnerabilities in supply chains, with one country dominating refining for 19 of 20 key strategic minerals, averaging a 70 percent global market share. That concentration creates chokepoint risk for the entire AI hardware stack — from the semiconductors in GPU clusters to the batteries in grid storage to the permanent magnets in wind turbines that provide some of the power.⁴ The Economist offered a sceptical counterpoint to the supply chain hardening narrative. Attempts to make supply chains resilient are likely to fail, the magazine argued, invoking Napoleon's observation that the torment of precautions often exceeds the dangers to be avoided. Governments and companies want to protect themselves from disruptions but the cost of doing so — duplicate facilities, strategic stockpiles, reshored manufacturing — may exceed the cost of the disruptions themselves.⁵ That scepticism has historical support. Supply shocks are often a trap for commodity investors. When waterways are blocked, tankers reroute. When demand collapses, producers curtail and then ramp quickly when conditions improve. The Economist noted that new supply can often be brought online faster than the market expects during disruptions. The AI commodity cycle may prove different in scale but the pattern of overinvestment followed by surplus is well established.⁷ Natural gas sits at the centre of the AI commodity chain as the marginal fuel for power generation. NYMEX Henry Hub front-month natural gas regained momentum with prices rallying on hotter weather forecasts and stronger power-sector demand. Working gas in storage fell by 52 Bcf for the week, well below the five-year average withdrawal of 168 Bcf, with inventories 141 Bcf above year-ago levels. The gas market is not yet tight enough to confirm the AI demand thesis in physical terms.^3,2 A wave of LNG projects is reshaping gas markets, with 300 billion cubic metres of new annual capacity expected by the end of the decade. That supply growth will compete with the AI-driven domestic gas demand story, potentially relieving the tightness that current forward curves are pricing. The unresolved question is whether the 580 billion dollar data centre investment wave generates enough physical commodity demand to sustain the price increases it has already triggered in equity markets. Babcock's legacy balance sheet — negative 131.5 million dollars in stockholders' equity and a 6.50 percent note refinancing due this year — is the micro signal. If the company that just booked 2.8 billion in backlog cannot refinance its debt, the gap between AI demand narrative and energy sector fundamentals is wider than the stock prices suggest.¹