Tesla Optimus Gen 3: The Fremont Line Conversion, Cortex 2.0, and the Reality Check Musk Just Admitted To

The setup: Tesla just ended a 14-year production run of the Model S and an 11-year run of the Model X to free up the Fremont production line for Optimus. The dismantling started in early May 2026. Optimus Gen 3 production is scheduled to begin late July or August 2026. Musk has called Gen 3 “by far the most advanced robot in the world.” He also admitted on the Q1 2026 earnings call that zero Optimus units did useful factory work in 2025, walking back his earlier prediction of 10,000 units that year. Both statements are true. This post walks through what’s actually shipping, what’s still aspirational, and where Optimus sits relative to Figure today.

🔗 See the robot: tesla.com/optimus — Tesla’s official Optimus page with the latest demo videos and program updates.

What Musk actually said on the Q1 2026 earnings call

The most honest Optimus update of the program’s existence happened on Tesla’s Q1 2026 earnings call in late April. Key admissions:

• Zero Optimus units did “useful work” in Tesla factories in 2025. Musk’s exact framing was that Optimus “is not in usage in our factories in a material way.” This is a walkback from his January 2025 prediction of 10,000 Optimus units that year — a target Tesla missed by approximately 10,000 units.
• Gen 3 production starts late July or August 2026, not Q1 2026. Musk had originally signaled a Q1 2026 Gen 3 reveal. That got pushed to “probably middle of this year.” Then on March 31, Musk posted on X that Optimus 3 is “mobile but requires some finishing touches before it is ready to be shown to the world.”
• “It is impossible to predict” 2026 production volume. Musk explicitly declined to give a 2026 production target, only saying initial output will be “quite slow” and that initial skills will be “simple skills in the factory.”
• Optimus has 10,000 unique parts. The Fremont line conversion involves dismantling the existing automotive production equipment “from the ground up” and installing an entirely new production line for those parts.

For a CEO known for aggressive timelines, this earnings call was unusually measured. Telling investors “it is impossible to predict” production rates and walking them through the practical realities of converting a car line to a robot line is the kind of reality injection that’s been missing from previous Optimus discussions.

The Fremont line conversion

The mechanical part of this story is concrete and verifiable. Tesla announced in late January 2026 that it would discontinue both the Model S and Model X and convert the Fremont production lines to Optimus manufacturing.

• Model S and Model X production ended in early May 2026. Combined, the line produced more than 610,000 vehicles over its 14-year run. Sales had dwindled to roughly 30,000 vehicles per year — a fraction of the line’s 100,000-unit capacity.
• Line dismantling started early May 2026. Tesla is tearing down the existing line “from the ground up,” starting with smaller parts production equipment and working forward to final assembly.
• Optimus production starts late July or August 2026. Per Musk, this is a four-month conversion. For context, the Cybertruck took over a year to ramp from first production to meaningful volume, on existing automotive expertise. Optimus is a fundamentally new product class.
• Giga Texas is the second Optimus factory. Tesla broke ground on the second factory on the Giga Texas north campus expansion. Production there is expected to begin around summer 2027. The design capacity is 10 million units per year — a number that’s aspirational but indicates intent.

The Fremont conversion plus Giga Texas roadmap is what Tesla is betting on. Vertical integration — Tesla owns the actuator design, the battery cells, the chip (AI5), the training infrastructure (Cortex 2.0), and the assembly — is the structural advantage Tesla has over any humanoid competitor. The bet is that this advantage compounds at scale even if the AI software is initially behind Figure’s Helix.

”Gen 3” is really the hands (for now)

A key clarification that surfaced after Tesla’s January 2026 announcements: “Gen 3” specifically refers to the upgraded 22-degree-of-freedom hands with 50 actuators (25 per forearm), not a complete new robot body. The body remains the Gen 2 design (1.73 m tall, ~57 kg) for current factory deployments. The fully new V3 body is still being finalized.

The Gen 3 hands themselves are a substantial upgrade:

• 22 DoF per hand (up from 11 in Gen 2)
• 50 total actuators (25 per side) — a 4.5x increase over Gen 2
• All 25 actuators housed in the forearm via tendon-pull biomimetic design, enabling a slimmer hand profile
• Production-ready as of February 17, 2026, per a Musk post
• 0.08mm hand positioning accuracy, per Tesla’s reported specs at AWE 2026

Tesla Board Chair Robyn Denholm confirmed the Gen 3 hands “can wash clothes, clean dishes, and shake hands with highly sensitive touch.” That last capability — sensitive touch — is the one that matters for home tasks, and is the gap Tesla has been trying to close against Figure’s 3-gram tactile sensitivity.

For factory deployments through 2026, the realistic configuration is Gen 2 body + Gen 3 hands. Full V3 production starts late summer 2026 per current guidance.

Cortex 2.0 — the training supercomputer

The infrastructure piece that gives the production timeline credibility is Cortex 2.0, Tesla’s next-generation AI training cluster at Giga Texas. The build-out:

• First 250 MW phase online in April 2026
• Full 500 MW capacity targeted by mid-2026
• 230,000 H100-equivalent GPUs across the combined campus
• Purpose-built for robotics training alongside FSD

Cortex 2.0 is timed almost exactly to support the summer 2026 Optimus production launch. Tesla’s bet is that the same vision-based training pipeline that produced Full Self-Driving can train Optimus, with the AI5 chip (delivering roughly 5x the memory bandwidth of AI4) running inference onboard the robot.

The AI5 chip and Cortex 2.0 also enable Grok integration — Tesla and xAI have been wiring Grok into Tesla vehicles in Europe (since update 2026.2.6 in February 2026), and the same Grok pipeline is expected to provide natural-language task instructions to Optimus.

AWE 2026 Shanghai and the $20K target

Tesla showcased Gen 3 at the Appliance and Electronics World Expo (AWE) 2026 in Shanghai on March 11, 2026 — Tesla’s first major product showcase in China after a multi-year absence. On-site staff stated mass production could begin by end of 2026 (a slight optimistic shading versus Musk’s “late July/August” framing).

The pricing target Tesla has been consistent about: under $20,000 at scale of 1 million units per year. This is the number that, if achieved, changes the entire humanoid market. No other Western humanoid program has a credible path to that price. Figure 03’s $20K consumer target relies on Figure achieving similar scale economics; Atlas at ~$150K and Digit at ~$250K are in a different price tier entirely.

Whether Tesla can actually hit the $20K number depends on three things going right: (1) the Fremont conversion delivering by late summer 2026 as planned, (2) Cortex 2.0 generating the training data needed for general-purpose autonomy, (3) the supply chain — particularly the 10,000 unique Optimus parts — maturing fast enough to support 1M-unit annual production. Missing any one historically stalls a hardware ramp. Tesla has missed prior Optimus production targets twice.

Where Optimus sits versus Figure right now

The most useful data point for ranking Optimus against the competition isn’t a spec sheet. It’s a comparative tracker published by New Market Pitch in April 2026 that scored both programs across 15 questions in three categories. Figure scored 78.9 out of 100. Tesla scored 45.1. The 33.8-point gap is concentrated in one category: real-world traction.

• Real-world traction: Figure 79, Tesla 32 — a 47-point gap driven by BMW’s own published deployment metrics.
• Robot performance specs: the gap narrows to about 26 points.
• Manufacturing scale today: Figure has BotQ at 1 robot/hour now; Tesla has the line conversion in progress.

The tracker estimates Figure is 12 to 18 months ahead of Tesla on a deployment timeline basis. That’s not a permanent gap — Tesla’s manufacturing advantage compounds over multi-year horizons — but it’s the gap as of mid-2026.

For our broader take on where Optimus sits among the major humanoid programs, see the Top 5 Humanoid Robots of 2026 roundup.

Bottom line

Tesla Optimus Gen 3 is a real program with a real production timeline now, but it’s a program that’s behind where Tesla originally said it would be. The Q1 2026 earnings call was the moment Musk publicly recalibrated. Production starts summer 2026 at Fremont, ramps slowly through 2027, and aims for high-volume production at Giga Texas starting summer 2027 at the earliest.

The Gen 3 hands, Cortex 2.0, AI5 chip, and the Fremont line conversion are all genuine progress — Tesla is shipping infrastructure, not just slide decks. But Figure is still ahead on deployment, AI, and per-unit manufacturing throughput. Tesla’s long-term advantage is its supply chain and vertical integration; whether that advantage actually wins depends entirely on execution over the next 18 months.

If you only follow one number to know whether the program is on track: watch the Fremont conversion completing on schedule (late July/August 2026) and the first Optimus units rolling off the new line. That’s the milestone that turns Optimus from a Musk prediction into a Tesla product.