WASHINGTON, July 7 — The copper thieves hitting America’s AI data centre construction sites are not just a security nuisance; they are now a line item on the balance sheets of the world’s largest technology companies, and the cost is being passed on to every business that uses cloud-based artificial intelligence. According to the International Data Corporation, U.S. investment in AI data centres reached $150 billion in 2024. Theft-related losses and the security spending required to combat them are now estimated at 2-3% of total project budgets.
That translates into an annual hit of $3-4.5 billion, a figure that has begun to reshape the economics of the entire AI infrastructure build-out. The detail that matters is the contingency.
Microsoft’s $80 billion investment in AI data centres for 2025 includes a $2 billion contingency specifically set aside for theft and security. That is not a rounding error; it is a direct acknowledgment that the industry now treats organised copper theft as a structural cost of doing business, rather than an occasional risk. Under the hood, the technical reality is that these delays compound.
Theft-driven disruptions have pushed some data centre completions back by 6-12 months. That lag has a direct market consequence: Gartner reports a 15% increase in the cost of cloud computing services for AI workloads in 2025.
The higher construction costs are passed on to AI service providers, and from them to the end consumer — every startup training a model, every enterprise running inference, every developer paying for API access. The problem has also hit the stock prices of the construction firms building these facilities. Turner Construction and DPR Construction have both reported lower margins attributable to theft.
Investors, as ever, are adjusting their models. Some venture capitalists are now demanding proof of security measures before they will fund AI infrastructure projects, adding another layer of due diligence to an already capital-intensive sector.
The mechanics of the theft
The target is copper — the wiring that powers and connects the enormous server racks. Data centre construction sites are sprawling, often in remote locations with cheap land and power, making them difficult to secure. Thieves strip wiring from partially built structures, target backup generators for their copper components, and in some cases have stolen equipment that is still on the delivery truck.
The Semiconductor Industry Association, in a 2025 white paper, warned that if thefts continue to escalate, the long-term impact includes a potential slowdown in AI development itself. That is the detail that should give the industry pause.
The United States is in a race to build the physical infrastructure for AI — the data centres that train and run the models. If the construction pipeline is chronically disrupted by theft, the bottleneck shifts from chip supply to site security. The $150 billion in 2024 investment was supposed to accelerate capability; a portion of it is instead being consumed by locks, cameras, guards, and replacement copper.
What to watch next
The question for the second half of 2026 is whether the industry can bring theft under control without further inflating costs. The contingency budgets are already baked into project finance, but if losses continue to run at the upper end of the 2-3% range, the 15% cloud cost increase reported by Gartner may prove to be a floor rather than a ceiling. Investors are watching the construction margins at Turner and DPR as a leading indicator.
If those margins do not stabilise, the cost of AI infrastructure — and therefore the cost of AI itself — will keep rising.
{“@context”: “https://schema.org”, “@type”: “NewsArticle”, “headline”: “Copper Theft Costs AI Data Centre Projects Billions In Annual Losses”, “description”: “WASHINGTON, July 7 — The copper thieves hitting America’s AI data centre construction sites are not just a security nuisance; they are now a line…”, “datePublished”: “2026-07-07T09:30:00+00:00”, “author”: {“@type”: “Person”, “name”: “Sebastian Smith”}, “publisher”: {“@type”: “Organization”, “name”: “InfoPulse Today”, “url”: “https://newsfeed.infopulsetoday.com/”}}
