Tesla Cortex 2.0 Powers Up in April: What It Means for Optimus

What Is Cortex 2.0 — And Why Does It Exist?

Tesla's original Cortex supercomputer cluster was built to train FSD. Cortex 2.0 is the next-generation evolution, purpose-built at Gigafactory Texas with a dramatically larger power footprint. At 500MW total capacity, it dwarfs most AI training facilities in the world.

According to Elon Musk's own statements, the GigaTexas compute cluster is built around approximately 100,000 Nvidia H100/H200 GPUs, paired with massive video storage infrastructure for training both FSD and Optimus. That GPU count alone puts it in the same league as the largest AI clusters operated by any company on the planet.

The primary focus for Cortex 2.0, per multiple sources, is Optimus — Tesla's humanoid robot currently in limited production. Training a robot to perform real-world physical tasks requires an enormous amount of compute: video data, simulation runs, reinforcement learning loops. The April activation of the first 250MW phase means that training pipeline is about to get a massive injection of horsepower.

📊 Key Figures

🔭 The BASENOR Take

The April activation of Cortex 2.0's first phase is a significant infrastructure milestone — not just for Tesla, but for the broader AI race. Here's the strategic read:

For Optimus: Tesla has been clear that compute is the bottleneck for robot training. Cortex 2.0 at full capacity is designed to run the kind of large-scale simulation and video-based learning that turns a robot prototype into a deployable product. The faster this cluster comes online, the faster Tesla can iterate on Optimus capabilities — and the faster it can scale production beyond the current limited runs. Every month of additional compute translates directly into robot dexterity, task completion rates, and real-world reliability.

For FSD: While Optimus is the primary stated focus, Cortex 2.0 will also feed the FSD training pipeline. More compute means Tesla can process more edge-case video data, run more neural net training iterations, and push toward higher autonomy levels faster. Owners waiting on FSD improvements have a direct stake in this infrastructure coming online on schedule.

The power numbers matter: 500MW is not a rounding error. For reference, a typical large data center runs 20-100MW. Tesla is building something in a completely different class — one that signals the company views AI compute as a core competitive moat, not just a supporting function. The decision to co-locate this at GigaTexas rather than a standalone facility also speaks to Tesla's vertical integration philosophy: manufacturing, energy, and AI training all under one roof.

Watch the mid-2026 date: The second 250MW phase coming online by mid-2026 will be the real inflection point. That's when the full training throughput becomes available — and when the downstream effects on Optimus capabilities and FSD performance should start becoming visible in product updates. For our FSD coverage, this is a key upstream variable to track.

📰 Deep Dive

What makes Cortex 2.0 structurally different from Tesla's original Cortex cluster isn't just scale — it's the deliberate pivot toward robotics training. The original Cortex was conceived primarily as a video-ingestion and neural net training machine for FSD, processing dashcam footage from Tesla's fleet at scale. Cortex 2.0 expands that mandate significantly, adding the simulation-heavy workloads that humanoid robot training demands. These are fundamentally different compute profiles, and building a single facility capable of handling both is an engineering statement in itself.

The infrastructure buildout has been visible from the air for months. Drone operators and Tesla watchers have documented Megapack installations on the north side of GigaTexas — the battery storage necessary to provide the power stability a 500MW compute cluster requires. Chiller systems and high-density electrical infrastructure have also been photographed in various stages of completion. The fact that a 200MW permit was approved in December 2025 and the first phase is now reported for April activation suggests the construction timeline has stayed largely on track.

One detail worth watching: the gap between the approved permit capacity (200MW) and the reported total build-out (500MW) suggests additional permitting and infrastructure work is still in progress for the second phase. The mid-2026 target for full 500MW operation is credible given the construction pace, but it carries the usual caveats around permitting, grid interconnection, and hardware delivery timelines. Tesla has not made an official public announcement confirming these specific dates.

For Tesla owners, the practical takeaway is this: the AI infrastructure that will define how good Optimus and FSD become over the next two to three years is being switched on in the next few weeks. Cortex 2.0 isn't a product you'll interact with directly — but its output will show up in every software update, every Optimus demo, and every FSD capability milestone that follows.