AI Data Centers Are Hungry for Metal. One Fund Is Betting on Recyclers

By Daniel Okafor

AI's demand for raw materials like aluminum and copper is structural, not a passing phase, according to Tazia Smith, CEO of Closed Loop Capital Management. "That integration and demand for certain resources, as you alluded to, feels pretty structural going forward," Smith said during Business Insider's Tech Insider video event on Wednesday. "Not a cyclical blip, but a more permanent systemic transition."

Smith runs a firm that funds businesses reusing raw materials in their supply chains. Her thesis is simple. Hyperscalers are racing to build data centers across the country, and every new facility needs metal, chips and water systems. That buildout creates a parallel market: recycling and repair at scale.

What does the AI buildout demand?

Massive data centers are resource-intensive. Smith's point is that the AI sector's appetite for aluminum and copper will keep growing, because the infrastructure itself keeps multiplying. Microsoft and Meta consider these often gargantuan server farms instrumental to their ambitions to dominate the rapidly expanding AI industry, and the facilities handle millions of chatbot queries daily.

The vocabulary is shifting too. OpenAI now wants people to call data centers "compute campuses." Whatever the label, the physical inputs remain the same.

"The large players that are focused on AI are also focused on resilient supply chains and material security," Smith told Business Insider's Dan DeFrancesco.

Where does the recycling money sit?

Smith's portfolio points to concrete mechanisms. Components inside data centers need repair or replacement relatively frequently, and she works with companies that recycle these expensive parts rather than send them to landfill.

Two examples stand out:

• Water leak repair. One company Smith works with specializes in plugging water leaks inside data centers, minimizing environmental harm at the facility level.
• Chip repair. A company called Moog repairs valuable computing components, including semiconductor chips, so data center builders avoid buying replacements every 5 years.

The economics are direct. "That equals value," Smith said. "That equals cost savings and opportunity to have wider margins."

The mechanism matters more than the slogan. Repair extends component life. Longer component life cuts capital spending. Lower capital spending widens margins for operators spending billions on buildout. For recyclers and repair firms, each avoided replacement is revenue.

Why did the midterms matter?

Data centers took a beating in last week's midterm elections. Voters shot down ballot measures supporting their expansion, handing the industry a political setback at the very moment capital is pouring into compute.

A principal critique of data centers has been what critics describe as their deleterious environmental impact. That critique cuts both ways for investors like Smith. It raises the political risk for hyperscalers. It also raises the value of companies that can cut water waste, recycle components and tighten material loops inside facilities voters might otherwise block.

What comes next?

Smith's framing suggests the recycling opportunity scales with the buildout. If AI demand for copper and aluminum is, in her words, a "more permanent systemic transition," then repair and reuse become recurring revenue rather than a one-off trade. The open question is political: whether voter resistance slows construction enough to shrink the pipeline of components needing repair, or instead forces operators to adopt circular supply chains faster to win local approval.