Benchmark: AI Boom Fueling Battery Metals Demand, but EVs Remain King

The rapid expansion of artificial intelligence (AI) infrastructure is emerging as a powerful new driver of demand for battery raw materials. Hyperscale data centers require vast energy storage, copper and critical minerals to support escalating electricity needs.

According to Benchmark Mineral Intelligence, AI-related power consumption is reshaping the battery supply chain. Lithium, graphite, phosphates and copper are all expected to see surging demand as tech giants build out global data center capacity.

Benchmark’s Battery Energy Storage Lead Shan Tomouk pointed to the rapid growth in AI adoption during the April 30, “Behind the AI Boom: How Data Centre Demand is Reshaping BESS, Lithium, LFP and Copper Markets” webinar, underscoring that behind each AI query lies a growing energy burden that is scaling rapidly alongside adoption.

Tomouk noted that a typical AI query consumes roughly 10 times the energy of a standard internet search, while increasingly sophisticated models require enormous amounts of computing power during training phases. GPT-4, for example, consumed an estimated 50 gigawatt-hours of electricity during training, up sharply from earlier generations.

“A number that’s been thrown out for the average query on a ChatGPT question is around 0.3 watt hours of energy,” he said, noting that this is “about 10 times that of a Google (NASDAQ:GOOGL) search.”

As a result, global data center electricity consumption is projected to rise significantly through 2030, led primarily by the United States, where most major AI developers are based. Benchmark expects hyperscale facilities, once defined as sites above 50 to 100 megawatts, to increasingly exceed gigawatt-scale power requirements.

That surge in electricity demand is creating new opportunities for battery energy storage systems (BESS), particularly as utilities struggle to expand grid infrastructure quickly enough to meet demand.

“Natural gas is really winning” as the preferred near-term power source for many data center operators, Tomouk said, citing projects where companies are building dedicated on-site gas generation to avoid delays tied to grid connections.

At the same time, energy storage is increasingly being deployed to provide backup power, manage peak electricity demand and stabilize sudden load swings caused by AI training workloads.