BlackRock published analysis rejecting AI infrastructure capex-bubble concerns and advocating 'Stars Surrounding the Moon' diversified investment strategy across the AI supply chain.
BlackRock Fundamental Equities' Asia-Pacific Chief Strategist Ben Bei stated that artificial intelligence has achieved a 56% global adoption rate within roughly three years of its first major product launch—far outpacing the early penetration of personal computers and the internet—with AI model intelligence improving at approximately tenfold each year. He emphasized that market concerns over excessive capital expenditure by tech giants and a potential bubble lack foundation, and advocated a "Stars Surrounding the Moon" macro strategy to position across three major investment dimensions.
**AI as Catalyst: A Supercycle Built on Profit**
Ben Bei noted that AI functions like a bright moon, driving prosperity across a constellation of surrounding industries. Data center investors should not limit their focus to individual chipmakers but instead position simultaneously across three dimensions to seize this long-term opportunity.
Addressing the most pressing concern over capital expenditure, Bei cited data showing that cumulative AI-related capital expenditure from 2026 to 2030 is estimated at $11 trillion, while cumulative EBITDA over the same period will reach approximately $10 trillion—demonstrating that this investment wave is backed by robust profit growth. Data center payback periods are approximately five to six years, after which returns are essentially pure profit.
Compared to traditional non-tech companies, the top ten tech stocks boast profit margins twice as high and earnings growth rates three times as fast, yet trade at relatively lower valuations—approximately 18x versus 24x price-to-earnings. The chip, server, and cloud model supply chain, protected by formidable cash flow moats, remains the premier choice for this supercycle. AI model capabilities continue to exhibit exponential growth, with applications expanding from chatbots to agentic AI, world models, physical AI, quantum computing, and scientific research. As physical applications such as autonomous vehicles, robotics, defense technology, healthcare, and chemistry expand, computing demand and power consumption will be further elevated.
**Power Infrastructure and Critical Minerals: Structural Supply Gaps**
"Without electricity, data centers are nothing more than elaborate decorations," Bei stressed. Global data center electricity demand is projected to reach 1,600 terawatt-hours by 2035, with power demand over the next decade growing to four times current levels. The proliferation of physical AI—such as 4.7 million industrial robots and drones—will continue to drive electricity consumption higher.
U.S. utility capital expenditure is projected to increase from $178.2 billion in 2024 to $248.4 billion in 2029, with transmission and distribution, grid upgrades, and energy storage equipment supply chains set to benefit significantly. Renewable energy and energy storage systems can be deployed more quickly and are positioned to meet incremental power demand first, while natural gas generation is constrained by turbine equipment supply and may not add new capacity until after 2030. New nuclear requires even longer construction timelines.
Mining remains a critical bottleneck. Capital expenditure by the five major hyperscalers climbed from $66 billion in 2018 to $379.5 billion in 2025, and is projected to reach $779 billion in 2026—yet mining capital expenditure has increased only modestly. A new mine takes an average of 17 years from exploration to production, roughly three times longer than 30 years ago, with capital expenditure remaining primarily focused on sustaining existing capacity.
Copper supply pressure is particularly acute: only 25% of copper mine development projects planned in 2015 have been completed; actual production from 2020 to 2025 fell short of earlier estimates by approximately 6 million tonnes. By the end of 2025, supply disruptions accounted for roughly 7% of global copper supply. China's electrolytic aluminum capacity is approaching the 45 million tonne ceiling, while demand spans real estate, passenger vehicles, power grids, solar, packaging, home appliances, and electronics. Under surging demand and rigid supply, copper, aluminum, and other critical metals will face structural supply-demand gaps—the mining mega-cycle has already taken shape.
**Gold: Portfolio Hedging Against Concentration Risk**
While embracing high growth, Bei cautioned that portfolios must establish defensive mechanisms. Global equity market concentration continues to rise, with the top ten constituents of the MSCI ACWI increasing from 23% at the end of 2022 to approximately 41% currently, significantly weakening the diversification benefits of traditional 60/40 stock-bond allocations. Over the past five years, diversification benefits provided by traditional 60/40 portfolios have been markedly lower than in the preceding two decades.
Gold mining equities exhibit a three-year correlation coefficient of only -0.11 with the S&P 500, effectively dampening volatility. An AI stock portfolio shows a correlation of 0.84 with the S&P 500 but -0.117 with the FTSE Gold Mines Index. Beyond central banks worldwide purchasing over 1,000 tonnes of gold annually for consecutive years, stablecoin issuer Tether acquired 84 tonnes in a single year, second only to the Polish central bank's 102 tonnes, demonstrating that large strategic buyers exist beyond the official sector. With gold miners' free cash flow improving significantly, net debt-to-EBITDA declining, and strict capital discipline maintained, high-quality gold mining equities represent the strongest defensive allocation for investors hedging geopolitical and macroeconomic risks.
Bei concluded that AI investment opportunities can be captured across three interconnected dimensions: directly positioning in companies with AI innovation capabilities; investing in power equipment, grids, energy storage systems, infrastructure, and mining supply chains that support AI scaling; and establishing positions in gold and high-quality gold mining equities to provide diversification benefits amid rising market concentration, geopolitical risks, and macroeconomic uncertainty.