Cargo thieves stole more than $150 million in artificial intelligence equipment between June 2025 and June 2026, increasingly using fraudulent carrier identities and falsified shipping documents to steal high-value shipments, according to a report from Overhaul.
The company’s AI Infrastructure Intel Report, released Oct. 21, analyzed 111 different AI equipment theft incidents across the United States from 2023 through June 2026. The number of incidents increased 34% since 2024, according to Overhaul.
The report comes weeks after the recent theft of $11.4 million in NVIDIA hardware from a shipment traveling from Oakland, California, to Fort Worth, Texas. Oakland police said the theft was reported Sept. 13 and remains under investigation, with no arrests announced.
Overhaul’s $150 million estimate covers a 12-month period from June 2025 through June 2026. By comparison, Scott Cornell, chair of TAPA Americas and an executive at SPG Cargo & Logistics, estimated that $110.6 million in data center freight had been stolen between June 3 and Aug. 14, 2026, when he shared the figure during an appearance on FreightWaves’ Fraud Watch podcast. The estimates cover different periods and categories of freight. Cornell’s estimate also predates the $11.4 million NVIDIA shipment theft, which was picked up on Aug. 29.
In the report, fraud or deception was involved in 92% of the reported thefts. Overhaul described incidents as deceptive pickups when criminals use tactics such as identity theft, fraudulent carrier identities and altered bills of lading to obtain shipments.
The report includes incidents identified within Overhaul’s own network, as well as thefts reported across the industry through open and closed intelligence sources. Overhaul sells supply chain visibility and risk management services.
The sheer value of AI equipment makes it an attractive target for cargo thieves. Overhaul estimates a single flagship AI server rack costs roughly $3 million, meaning a truckload of computing hardware can represent millions of dollars in potential losses.
“As thieves become increasingly deceptive, visibility into a shipment alone isn’t enough to stop it from being taken,” Barry Conlon, CEO and founder of Overhaul, said in the release. The report’s conclusion that visibility alone is not enough supports the case for Overhaul’s broader services.
The increase in AI equipment theft comes amid a broader rise in cargo crime across North America. Verisk CargoNet recorded 2,646 confirmed thefts in the United States and Canada in 2025, an 18% increase from the previous year. Estimated losses reached nearly $725 million, up 60%, while the average value per theft climbed to $273,990.
The financial impact of AI equipment theft can extend beyond the value of the stolen cargo. Overhaul estimates that for every dollar of stolen hardware covered by an insurance claim, roughly $3 in additional losses go unclaimed. The company said the 3:1 ratio is a conservative estimate based on a RAND research paper. Those indirect costs include deployment delays, lost contract eligibility, regulatory and compliance exposure, and additional security investments.
The incidents highlight a challenge for shippers and logistics providers moving high-value technology: Tracking a shipment’s location may not be enough when criminals can impersonate legitimate carriers and manipulate shipping documents. Verifying the identity of the person collecting a load and confirming that they are authorized to transport it are also critical to preventing theft.
Why this matters
The growing value of AI infrastructure is creating new opportunities for cargo thieves, who are increasingly using carrier impersonation and fraudulent paperwork to intercept shipments. For shippers and logistics providers, tracking technology alone may not be enough to prevent losses as verifying carrier identities and pickup authorization becomes increasingly important. Overhaul provides supply chain visibility, risk management and cargo security solutions designed to help companies monitor shipments and prevent theft.
