The golden age of autonomous ridesharing is arriving, and Tesla is making sure we can all picture a future that looks like the future. A recent teaser posted to X shows a Cybercab parked outside a home, and with a clear message that your everyday life may soon look like this when the driverless vehicles shows up at your door.

Tesla has begun the rollout of its Robotaxi service across US cities, and the production of its dedicated, fully-autonomous Cybercab vehicle. The first Cybercab rolled off the Giga Texas assembly line on February 17, 2026, with volume production now targeted for this month. Additionally, the Robotaxi service built around it is already running, without human drivers, in US cities.

Tesla Cybercab production ignites with 60 units spotted at Giga Texas

The Cybercab is built without a steering wheel, pedals, or side mirrors, designed from the ground up for unsupervised autonomous operation. Musk described the manufacturing approach as closer to consumer electronics than traditional car production, targeting a cycle time of one unit every ten seconds at full scale.

Drone footage from April 13, 2026 captured over 50 Cybercab units on the Giga Texas campus, with several clustered near the crash testing facility. Musk has noted that Tesla plans to sell the Cybercab to consumers for under $30,000, and owners will be able to add their vehicles to the Tesla robotaxi network when not in personal use, potentially generating income to offset the vehicle’s purchase cost. That model changes the math on vehicle ownership in a meaningful way, making a car something closer to a depreciating asset that can also earn by paying itself off and generate a profit.

During Tesla’s Q4 earnings call, the company confirmed plans to expand the Robotaxi program to seven new cities in the first half of 2026, including Dallas, Houston, Phoenix, Miami, Orlando, Tampa, and Las Vegas. The service already runs without safety drivers in Austin, and public road testing of the Cybercab has expanded to five states, including California, Texas, New York, Illinois, and Massachusetts.

Golden era pic.twitter.com/AS6pX2dK8N

— Tesla Robotaxi (@robotaxi) April 16, 2026

Tesla is planning to soon reveal its latest and greatest version of the Optimus humanoid robot, and a series of new patents for the hands and arms, with the former being, admittedly, one of the most challenging parts of developing the project.

Two new patents, which were coincidentally filed on the same day as the “We, Robot” event back in October 2024, protect Tesla’s mechanically actuated, tendon-driven architecture.

The designs relocate heavy actuators to the forearm, route cables through a sophisticated wrist design, and employ innovative joint assemblies to achieve human-like dexterity while enabling lightweight construction and high-volume manufacturing.

Core Tendon-Driven Hand Architecture

The primary patent, which is titled “Mechanically Actuated Robotic Hand,” details a cable/tendon-driven system.

Actuators are positioned in the forearm rather than the hand. Each finger features four degrees of freedom (DoF), while the wrist adds two more.

Tesla’s Optimus V3 robot hand looks to have been revealed in a new international patent published today.

The patent describes a tendon/cable-driven hand:
• Actuators in the forearm
• Each finger has 4 degrees of freedom
• The wrist has 2 degrees of freedom
• Tendon-driven… pic.twitter.com/eE8xLEYSrx

— Sawyer Merritt (@SawyerMerritt) April 16, 2026

Three thin, flexible control cables (tendons) per finger extend from the forearm actuators, pass through the wrist, and connect to the finger segments. Integrated channels within the finger phalanges guide these cables selectively—routing behind some joints and forward of others—to enable independent bending without unintended motion.

Patent diagrams illustrate thick cable bundles emerging from the wrist into the palm and fingers, with labeled pivots and routing guides. This setup closely mirrors human forearm-muscle and tendon anatomy, where most hand control originates proximally.

Advanced Wrist Routing Innovation

One of the standout features is the wrist’s cable transition mechanism. Cables shift from a lateral stack on the forearm side to a vertical stack on the hand side through a specialized transition zone.

Boom! @Tesla_Optimus 의 3세대 구조로 추정되는, 로봇 팔 및 관절에 대한 특허가 공개되었습니다.

아티클 작업에 들어가겠습니다.

Advertisement

1년 넘게 기다려 온, 정말 귀한 특허인데, 조회수 100만대로 터져줬으면 좋겠네요. 😉@herbertong @SawyerMerritt@GoingBallistic5 @TheHumanoidHub pic.twitter.com/CCEiIlMFSX

— SETI Park (@seti_park) April 16, 2026

This geometry significantly reduces cable stretch, torque, friction, and crosstalk during combined yaw and pitch wrist movements — common failure points in simpler tendon systems that cause imprecise or jerky motion.

By minimizing these issues, the design supports smoother, more reliable multi-axis wrist operation, essential for complex real-world tasks.

Companion Patents on Appendage and Joint Design

Two supporting patents provide additional depth. “Robotic Appendage” covers the overall forearm-to-palm-to-finger assembly, with a palm body movably coupled to the forearm and finger phalanges linked by tensile cables returning to forearm actuators. Tensioning these cables repositions the phalanges precisely.

“Joint Assembly for Robotic Appendage” describes curved contact surfaces on mating structures paired with a composite flexible member. This allows smooth pivoting while maintaining consistent tension, enhancing durability, and simplifying assembly for mass production.

Executive Insights on Hand Development Challenges

Tesla executives have consistently described the hand as the most difficult component of Optimus.

Elon Musk has called it “the majority of the engineering difficulty of the entire robot,” emphasizing that human hands possess roughly 27–28 DoF with an intricate tendon network powered largely by forearm muscles. He has likened the challenge to something “harder than Cybertruck or Model X… somewhere between Model X and Starship.”

Elon Musk shares ridiculous fact about Optimus’ hand demos

In mid-2025, Musk acknowledged that Tesla was “struggling” to finalize the hand and forearm design. By early 2026, he stated that the company had overcome the “hardest” problems, including human-level manual dexterity, real-world AI integration, and volume production scalability.

He estimated the electromechanical hand represents about 60 percent of the overall Optimus challenge, compounded by the lack of an existing supply chain for such precision components.

These patents directly tackle the acknowledged pain points: relocating actuators reduces hand mass and inertia for better speed and efficiency; advanced wrist routing and joint geometry address friction and crosstalk; and simplified, stackable parts visible in the diagrams indicate readiness for high-volume manufacturing.

Implications for Optimus Production and Leadership

Collectively, the patents portray the Optimus v3 hand not as a mere prototype, but as a production-oriented system engineered from first principles.

The 22-DoF architecture, forearm-driven tendons, and crosstalk-minimizing wrist deliver a clear competitive edge in dexterity. They align with Musk’s view that high-volume manufacturing is one of the three critical elements missing from most other humanoid projects.

For Optimus to become the most capable humanoid robot, its hand needed to replicate the useful and applicable design of the human counterpart.

These filings demonstrate that Tesla has transformed years of engineering challenges into patented, elegant solutions — positioning the company strongly in the race toward general-purpose robotics.

Tesla has finalized its chip design for AI5, as Elon Musk confirmed today that the new chip has reached the tape-out stage, the final step before mass production.

But in a brief reply on X, Musk clarified Tesla’s AI hardware roadmap, essentially confirming that the new chip will not be utilized for being “enough to achieve much better than human safety for FSD.”

He said that AI4 is enough to do that.

Instead, the AI5 chip will be focused on Tesla’s big-time projects for the future: Optimus and supercomputer clusters.

Musk thanked TSMC and Samsung for production support, noting that AI5 could become “one of the most produced AI chips ever.” Yet, the key pivot came in his direct answer: vehicles no longer need the bleeding-edge silicon.

And thank you to @TaiwanSemi_TSC and @Samsung for your support in bringing this chip to production! It will be one of most produced AI chips ever.

— Elon Musk (@elonmusk) April 15, 2026

Existing AI4 hardware, which is already deployed in hundreds of thousands of HW4-equipped Teslas, delivers safety metrics superior to human drivers for Full Self-Driving. AI5 will instead accelerate Optimus robot development and massive Dojo-style training clusters.

The Tesla CEO’s words mark a strategic shift. Tesla has long emphasized software-hardware co-design, squeezing maximum performance from every transistor. Musk previously described AI5 as optimized for edge inference in both Robotaxi and Optimus.

Now, with AI4 proving sufficient, the company avoids costly retrofits across its fleet while redirecting next-generation compute toward higher-value applications: dexterous robots and exponential training scale.

But is it reasonable to assume AI4 enables unsupervised self-driving? Yes, but with important caveats.

On the hardware side, the claim is credible. Tesla’s FSD stack runs end-to-end neural networks trained on billions of miles of real-world data. Internal safety data reportedly shows AI4-equipped vehicles already outperforming average human drivers by a significant margin in controlled metrics (collision avoidance, reaction time, edge-case handling).

Dual-redundant AI4 chips provide ample headroom for the driving task, leaving bandwidth for future model improvements without new silicon. Musk’s assertion aligns with Tesla’s pattern of over-provisioning compute early, then optimizing ruthlessly, exactly as HW3 once sufficed before HW4 scaled further.

Optimus and our supercomputer clusters.

AI4 is enough to achieve much better than human safety for FSD.

— Elon Musk (@elonmusk) April 15, 2026

Unsupervised autonomy, meaning Level 4 or higher, is not solely a compute problem. Regulatory approval remains the primary gate.

Even if AI4 achieves “much better than human” safety statistically, agencies like the NHTSA demand exhaustive validation, liability frameworks, and public trust.

Tesla’s supervised FSD has shown rapid gains in recent versions, yet real-world edge cases, like construction zones, emergency vehicles, and adverse weather, still require driver intervention in many jurisdictions. Competitors like Waymo operate limited unsupervised fleets, but only in geofenced areas with extensive mapping. Tesla’s vision-only, fleet-scale approach is more ambitious—and harder to certify globally.

In short, Musk’s post is both pragmatic and bullish. AI4 is likely capable of unsupervised FSD from a technical standpoint. Whether regulators and consumers agree, and how quickly, will determine if Tesla’s bet pays off.

The company’s capital-efficient path keeps existing cars relevant while pouring future compute into robots. If the safety data holds, unsupervised autonomy could arrive sooner than many expect.