Boston Dynamics Brings Atlas Into Chip Fabs With GlobalFoundries Deal

GlobalFoundries becomes the first semiconductor fab to pilot the Atlas humanoid, while scaling up a Spot fleet that already hit a 99% mission success rate.

·
·
·
Boston Dynamics Brings Atlas Into Chip Fabs With GlobalFoundries Deal
Read4 min
TypeNews
  • Boston Dynamics is expanding its partnership with GlobalFoundries across four global fabs.
  • GF becomes the first semiconductor manufacturer in the Atlas humanoid early adopter program.
  • It is also Atlas's first announced customer outside Hyundai Motor Group.
  • Two Spots in Singapore have logged 900 km and 90,000+ inspections at 99% success.
  • Atlas onsite testing at GF facilities begins in early 2027.
  • Sites include Malta NY, Burlington VT, Dresden Germany, and Singapore.

Boston Dynamics is taking its electric Atlas humanoid beyond automotive manufacturing through an expanded partnership with GlobalFoundries (GF). The chipmaker will become the first semiconductor manufacturer in the Atlas early-adopter program, while expanding Spot deployments to four fabrication plants, or fabs, on three continents. Atlas trials are scheduled to begin in early 2027, giving the robot its first publicly named manufacturing user outside Hyundai Motor Group.

Atlas enters the chip fab

Atlas’s commercial rollout has centered on Hyundai, which owns Boston Dynamics and plans to use the humanoid across Hyundai Motor and Kia factories. Google DeepMind is also working with Boston Dynamics to integrate Gemini Robotics, a family of general-purpose models for robot perception and action, into the platform.

Semiconductor plants present a different operating environment. Fabs enforce strict particle controls, handle hazardous chemicals and demand precise, repeatable work. Their sub-fab areas, which support production floors, contain dense networks of pumps, valves, gas lines and other utility equipment. Any Atlas deployment may also require contamination controls, specialized safety validation and integration with existing maintenance systems.

GF and Boston Dynamics have not identified the first Atlas tasks or disclosed whether trials will occur in cleanrooms, sub-fabs or other parts of the facilities. Those decisions will shape the hardware adaptations, perception systems and safety controls required for deployment.

Spot brings field data

GF’s case for testing Atlas builds on four years of work with Spot, Boston Dynamics’ four-legged inspection robot. At the company’s Singapore site, two units patrol sub-fab areas and collect readings from equipment that employees would otherwise inspect manually.

Metric GF’s reported result
Distance traveled Nearly 900 kilometers since April 2025
Inspections completed More than 90,000
Mission success rate 99%
Inspection rollout 875 routine routes placed into service within two weeks
Planned expansion More than twice as many routes, including visual-inspection tasks

Repeated inspections can support predictive maintenance by flagging abnormal readings before equipment fails. That capability has direct economic value in semiconductor manufacturing, where a disrupted tool or utility system can halt tightly coordinated production steps and reduce output across a facility.

Atlas starts with task discovery

The early-2027 Atlas work will begin with workshops and trials focused on identifying suitable workflows, mapping the required systems integration and training the robot for selected tasks. The announcement specifies no production volume, deployment count, performance targets or timetable for moving from evaluation to routine operations.

Boston Dynamics outlined a similar schedule at CES 2026, where it unveiled the production version of electric Atlas and said that year’s initial run was fully committed. The company planned to begin work with additional customers in early 2027, and GF is the first publicly identified participant in that phase.

Factory humanoids multiply

Hyundai Motor Group plans to deploy more than 25,000 Atlas units across its global plants by 2030 and is targeting annual production capacity of 30,000 robots by 2028. Other manufacturers are testing competing systems, including Tesla’s Optimus, Figure’s humanoids, Agility Robotics’ Digit and Apptronik’s Apollo. BMW and Mercedes-Benz have also announced factory trials, while several Chinese automakers and robotics companies are developing their own platforms.

A semiconductor deployment would expose Atlas to requirements that differ from the material-handling and assembly tasks common in automotive and warehouse trials. Reliable performance around sensitive equipment, constrained spaces and strict operating procedures could give Boston Dynamics evidence for expanding into other process industries.

Five tests for the robotics stack

Because the partners have not disclosed specific Atlas workflows, the project’s engineering value will emerge through details about how the system performs and integrates. Five areas will indicate whether the pilot can support a production deployment:

  • Task fit: whether Atlas can complete useful work that benefits from a human-shaped body, reach and mobility.
  • Perception: whether its models can interpret gauges, controls, tools and equipment across varied lighting and layouts.
  • Reliability: how often the robot completes tasks without human intervention, recovery or remote assistance.
  • Integration: how fleet software, maintenance systems, access controls and operational logs exchange data.
  • Safety and contamination: how the deployment manages people, hazardous equipment, emergency stops and particle-sensitive areas.

Data from chip fabs could also broaden the range of environments and tasks represented in models used for manufacturing robots. GF’s chief manufacturing officer has described Spot and Atlas as part of the company’s AI-enabled manufacturing strategy, while Boston Dynamics expects deployment data to inform future models and task training.

A production case will depend on measurable results such as task completion rates, intervention frequency, cycle time, safety incidents, maintenance burden and total operating cost. The 2027 trials should show whether Atlas can extend Spot’s inspection record into useful humanoid work inside a semiconductor fab.

Trending
  • No trending articles

Comments

avatar

Next Reads