Lenovo Whitsett AI Server Line Doubles Plant To 890,000 Square Feet

Lenovo has doubled its Whitsett, North Carolina plant to 890,000 square feet as it expands U.S. production and liquid-cooling test capacity for AI servers, Tom's Hardware reported after touring the site.

The facility employs 400 line workers and has room for 1,000.

The expansion combines additional floor area, utility capacity, liquid-cooling equipment and test workflow for higher-density AI racks.

The figures come from a facility tour and Lenovo executive remarks.

Whitsett Line Adds 20 MW Of Power

The Tom's Hardware tour recorded a nine-month conversion from an empty warehouse to a working server production line.

Duke Energy supplied electricity capacity of up to 20 MW, and the site moved from temporary power to full power within 7 months.

The manufacturing site is near Lenovo's U.S. headquarters and R&D hub in Morrisville.

Lenovo said its local workforce plan draws on the Research Triangle and nearby universities, including NC State, Duke, UNC-Chapel Hill and Wake Forest.

Ashley Gorakhpurwalla, executive vice-president and president of Lenovo's Infrastructure Solutions Group, said customers are looking for U.S.-produced hardware and an infrastructure base they can trust.

Ben Massey, Lenovo's vice-president of supply chain, said the North Carolina location could shorten the path through development, mass production, customer delivery and cost control.

Rack Assembly Reaches 200 Racks A Week

Lenovo said the kitting area covers 26,000 square feet, with 10 lines and 50 team members processing more than 150,000 parts per day.

It put current assembly at 200 racks per week and capacity at 400.

Lenovo said assembly occupies 53,000 square feet across seven lines.

It put capacity at 1,400 server nodes per day across 28 stations, equivalent to one server every 60 seconds, and described a full configuration of 1,000 operators across three shifts after roughly 1,000 training hours.

The expansion sits inside the physical supply chain for AI compute rather than a software-only AI story.

Processors, memory and storage are brought together at the kitting stage, then server nodes move through assembly and testing before customer delivery.

For enterprises and hyperscalers seeking domestic hardware supply, the weekly rack rate, daily node capacity and power connection are the useful evidence beyond Lenovo's broader AI language.

Liquid-Cooling Pods Test GB300 And Vera Rubin Loads

Testing capacity shows how the Whitsett line is being prepared for dense AI systems.

Lenovo said it has 10 thermally controlled L10 testing pods, each with 16 test racks, and air handlers with 1 MW of cooling capacity.

Lenovo said the site has 30 Vertiv Liebert XDU direct liquid-cooling units and room for 50, with capacity of up to 300 kW each.

Lenovo also put Nvidia GB300 rack demand at 140 kW and Vera Rubin at 230 kW under peak load, and said it expects its first Vera Rubin system at Whitsett by the end of August.

Lenovo did not identify customers, purchase orders or shipment volumes for the expanded AI server line.