Foundry Raises $100M Series A for Autonomous Factories

Building specialized hardware in the United States has spent three decades trapped in a painful manufacturing cycle. When an aerospace venture designs a reconnaissance drone or a defense supplier needs specialized wiring harnesses, constructing a dedicated factory production line takes eighteen months and tens of millions of dollars in mechanical tooling. If market demand shifts or the engineering team alters a metal bracket design, the physical production line becomes obsolete, requiring complete teardowns and expensive machinery retooling. American hardware startups were forced to outsource physical assembly to overseas plants, creating supply chain vulnerabilities and intellectual property risks. Bay Area robotics startup Foundry Robotics moved to solve that industrial bottleneck on Wednesday, October 28, 2026. The company announced a $100M Series A financing round led by defense-focused investment firm Marlinspike Partners, joined by prominent venture syndicates including Khosla Ventures, Red Glass Ventures, and SE Ventures. The financing sits in the top three percent of all American deeptech Series A rounds on record, providing capital to build out a 130,000-square-foot autonomous production center in Emeryville, California. The capital raise arrives as venture funds back physical automation platforms, continuing hardware trends we tracked when Sequoia backed Mecka AI for robot training data.

The strategic differentiation that convinced investors to write a nine-figure Series A check sits in Foundry business model: manufacturing-as-a-service. Instead of licensing software code or selling robotic arms to existing manufacturers, Foundry owns and operates the physical factory floor directly. Founded by Chief Executive Officer Adarsh Kulkarni, the enterprise designed flexible robotic cells powered by its proprietary FoundryOS software platform. The system connects multi-axis industrial robots, computer vision cameras, and cloud computation into a unified assembly network. When a customer submits computer-aided design blueprints for an uncrewed aerial vehicle or satellite electronics subassembly, Foundry robotic cells reconfigure their physical toolheads and assemble the product within weeks, eliminating the multi-month tooling delays that plague traditional contract manufacturers. Duane Fotheringham, president of HII Unmanned Systems, stated that physical production capacity has become the primary national constraint on deploying autonomous systems. Kulkarni argued that reviving domestic manufacturing requires rebuilding factories from the ground up rather than retrofitting outdated machinery. Autonomous factory software integrations match principles we examined when CloudNC automated precision factory operations.

The Tooling Trap of Traditional Assembly Lines

To grasp why Foundry raised $100M, one must inspect how physical products are assembled in traditional factories. Conventional manufacturing relies on fixed automation: stamping presses, injection molds, and pick-and-place arms bolted to concrete floors, engineered to perform one repetitive movement millions of times.

Fixed automation works well for mass-market consumer gadgets like plastic toothbrushes or television remotes, where production runs exceed five million identical units. Modern aerospace, defense, and industrial hardware operate on different commercial rules: high complexity, rapid design changes, and low production volumes. An uncrewed defense drone or orbital satellite subassembly might require only five hundred units per year. Building dedicated metal stamping dies and specialized assembly fixtures for small production runs drives per-unit expenses into the tens of thousands of dollars. When military engineers update a sensor package, the assembly line stops while technicians spend weeks machining new fixtures. Foundry replaces specialized mechanical fixtures with adaptive robotic arms guided by real-time neural vision. The robots adjust their grip angles, torque limits, and movement trajectories based on live optical camera feedback, assembling complex electro-mechanical devices without custom physical tooling. Robotic coordination across distributed facilities was detailed in our review of Destro AI coordinating automation systems.

FoundryOS: The Brain Behind Adaptive Robotic Cells

The technical core of the Foundry platform is FoundryOS, an integrated operating system that bridges digital engineering blueprints with physical robotic actuators. In traditional factory automation, programming a single robotic arm to assemble a wire harness requires weeks of manual programming in specialized languages.

FoundryOS automates this programming pipeline through synthetic simulation and spatial reasoning models. The workflow executes through three distinct operational phases:

• Digital Blueprint Ingestion: The software parses customer 3D CAD models, mapping assembly hierarchies, fastener torque specifications, and tolerance boundaries automatically.
• Simulation Path Planning: Inside virtual simulation environments, physics models test thousands of robotic movement paths, calculating optimal toolhead selections and avoiding mechanical collisions before physical motors turn.
• Closed-Loop Execution: On the physical factory floor, multi-spectrum camera sensors track robotic movements, measuring parts down to five micrometers and feeding telemetry back into the model to refine subsequent assembly cycles.

Because the software learns from every production run, knowledge transfers across product categories. A robotic cell that spent Monday assembling a drone power distribution box can spend Wednesday assembling an aerospace satellite transceiver with zero physical downtime. The software simply loads a new motion policy, swaps magnetic end-effector tools, and begins production immediately. Precision engineering automation was explored in our analysis of Flow Engineering securing growth capital for industrial workflows.

Comparing Contract Manufacturing Business Models