AUSTIN, Texas, March 2026 — Elon Musk has confirmed that Tesla’s dedicated facility for manufacturing its next-generation AI inference chips will begin production in 2026, marking a major step toward reducing the company’s dependence on external suppliers such as Nvidia.
In a post on X and during a recent earnings call, Musk stated that the new chip factory—located near Tesla’s Gigafactory Texas in Austin—will focus on high-volume production of custom silicon optimized for Tesla’s Full Self-Driving (FSD) neural networks, Dojo supercomputer training, and next-generation vehicle inference workloads.
The facility is expected to produce chips using advanced nodes (likely 4nm or 3nm class) sourced from TSMC or Samsung, with Tesla handling final packaging, testing, and integration.
Musk described the project as critical for cost reduction and performance gains: “We need to control our destiny on AI compute. This fab will let us iterate faster and bring down the cost per FLOP dramatically.”
Tesla has been developing its Dojo D1 chip since 2021 and introduced the Dojo ExaPOD system in late 2025. The new facility will scale production of follow-on designs, including inference-focused chips for in-vehicle use. Musk indicated initial output will support Tesla’s robotaxi fleet and Optimus humanoid robot programs.
The announcement comes amid ongoing chip shortages and high costs for Nvidia’s H100 and Blackwell GPUs, which Tesla has used extensively for training. By bringing more of the supply chain in-house, Tesla aims to achieve greater vertical integration and shield itself from external supply constraints.
No specific production volume, exact timeline within 2026, or cost estimates were provided. Musk noted that the facility will also support limited external sales or partnerships in the future, though the primary focus remains internal use.
Tesla shares rose modestly following the news, reflecting investor optimism about long-term cost savings and AI autonomy leadership.
