“As Artificial Intelligence (AI) gets more efficient and accessible, we will see its use skyrocket, turning it into a commodity we just can’t get enough of” – Microsoft CEO, Satya Nadella.
According to Mckinsey’s AI in workplace report, 92% of Fortune 500 companies will adopt AI in their workstreams, from advanced computing to automating daily tasks. While adoption accelerates, the “big-tech” companies are strategically investing in data center buildouts. However, two critical bottlenecks are emerging: power to run these data centers and commodities (from steel to rare earth). What are the current challenges with scaling AI infrastructure and how can policies shape outcomes?
Let’s start by understanding the various activities that are performed, today and in future. First, traditional enterprise workloads — for example, file storage and sharing — will continue as more workloads are digitized at company level. These types of activities are expected to grow at a compound annual growth rate (CAGR) of 7%. Second, GenAI-related workload, where advanced models will be used to generate insights and predictions, are expected to grow at a CAGR of 65% until 2028. These workloads are being scaled up in many industries such as drug discovery, materials science, logistics, and other sectors could further drive growth of over 120%.
While there are many investments, flowing into data center expansions, as shown in this Red Chalk chart, capital investments into data center expansions, primarily driven by hyperscale’s (ex: quantum computing) is expected to grow by 33-40%, equating to $30 billion per year for the next decade.
However, this growth is increasingly facing two critical bottlenecks. The first is energy supply and grid integration. According to an IEA report, global electricity demand from data centers is expected to double by 2030, with the U.S. alone increasing its power consumption by 130%. Currently consuming about 4% of global electricity, data centers could account for up to 12% by 2030. That’s roughly equal to the EU’s current electricity—making them one of the fastest-growing energy consumers. This places significant strain on grid infrastructure, permitting, and energy transition goals.
The other challenge is the supply of materials and strategic risk.
As energy demand garners attention, material supply risks are becoming a hidden bottleneck. According to the USGS, the materials that make the core “computing” components in a data center (as shown in the table), are becoming increasingly constrained due to geopolitical tensions, underinvestment in refining, and competing demand from other sectors (e.g., EVs, batteries, renewables).
| Material | Current Applications | Growth Driver | Risk Level |
| Copper | Cabling, power distribution | High compute power | High |
| Aluminum | Racks, casings, heat sinks | Modular design & scaling | Medium |
| Rare Earths | Motors, fans | Energy-efficient cooling | High |
| Lithium | Backup energy (BESS) | Renewable microgrids | Medium |
Strategic Implications and the Path Forward
As material bottlenecks deepen, they pose a geo-economic risk to AI competitiveness. If access to critical minerals is becoming increasingly difficult, AI development in the United States could stall, impacting national security, innovation, and economic growth. To address these risks, a multi-step approach is required:
- Policy Reform: Fast-track permitting timelines (reduce 5-7 year lag to less than 3 years) A solar energy project takes between 6.5-10 years and a critical materials project could take upwards of 10 years
- Mineral Security Strategy: U.S. government intervention (e.g., DOE $400m investment in MP materials)
- Corporate Supply Chain Resilience: Tech companies securing upstream materials (e.g., Apple’s rare earth supply deal)
- Additional Energy Integration: Build-out of nuclear energy to support growing demand and renewable capacity to support infrastructure
- Innovation: Co-investment in refining, recycling, and grid modernization (ex: Amazon’s investment in Redwood Materials on battery recycling)
For the United States to remain competitive in the global AI race, it must align innovation with bold policy reforms. Meeting the base-case energy demand and securing critical minerals will be essential to scale AI infrastructure.
Readers’ Question:
What happens when data center construction slows – not due to lack of demand, but due to a lack of power to drive these data centers and critical minerals such as, copper, lithium, or rare earths, to build them?