Today, Elon discussed the 26,000 units of Optimus robots in 2025 and 500,000 units in 2027, which Tesla's management had previously revealed multiple times during investor communications in North America: (Thanks to US stock source information)
<ol>
<li>
By the end of 2025, it is expected to produce &#34;several K,&#34; which means a few thousand units (IR mentioned 1,000-2,000 units), for testing to identify 10-15 manufacturing application scenarios (for Tesla's internal use).
</li>
<li>
In 2026, collaborate with external manufacturing partners to achieve mass production of &#34;tens of thousands,&#34; which means several tens of thousands of units.
</li>
<li>
In 2027, open to the public and implement scalable training methods to enter more industries, especially home scenarios.
</li>
<li>
In 2026, the goal is to achieve a BOM of $70,000 to $80,000, aiming to reduce it to the cost level of automobiles. (If 500,000 units are indeed achieved in 2027, then it would be below $35,000.)
</li>
</ol>
Therefore, the demand and mass production rhythm are very clear: a few thousand units are for Tesla's internal factory use, several tens of thousands are likely to enter external factory customers, and hundreds of thousands must enter the home to C scenario.
Additionally, Tesla's management mentioned some long-term issues:
<ol>
<li>
Achieving technological maturity is not easy; the management's original words were “The difficulty of Optimus is about 10 times that of achieving full autonomous driving.” A car has only 4 core controls (left, right, throttle, brake), with a linear adjustment range of 0-100%. However, robot motion design has dozens of degrees of freedom and different force sensitivities (for example, the subtle difference in grip strength between holding an egg and a stone). Training a robot is much more challenging than training FSD; cars can rely on millions of real driving video datasets, while training robots is insufficient with just video observation; real physical interaction data is needed.
</li>
<li>
Why adopt a humanoid shape? It can be compatible with social environments designed for human interaction (not surprising). For tasks like cleaning, cooking, and climbing stairs, a humanoid shape can interact more effectively with household items and building layouts. Non-humanoid robots, such as factory robotic arms, can handle limited tasks but lack the versatility for broader applications.
</li>
<li>
Currently, there is only a small team of a few hundred engineers. Developing robots for specific tasks may achieve commercialization faster, but the application range is limited. Initially, Optimus will start with simple, repetitive tasks in Tesla factories (Pick and place). In the short term, the work speed still lags behind humans and cannot achieve cost reduction for humans. However, Elon’s vision for robots is very long-term, over 50 years. The current goal is not EPS profit contribution but exploring future possibilities.
</li>
</ol>
Therefore, Musk's announcement of the large-scale production was not sudden; the internal goals have been set this way. The timeline and demand scenario matching sound quite reasonable. Clearly, there are many issues to resolve, the chicken-and-egg problem of cost and quantity, whether to first have data training capabilities or to have the ability to enter scenarios to obtain data.
Can the Austin, Texas factory and supply chain prepare tens of thousands of units of capacity 26 years in advance, etc. **. FSD has also faced bumps and doubts along the way before seeing the light, and I believe that embodied intelligence will develop at a non-linear speed under the acceleration of AI.
In addition, Huang's speech at CES revealed another ambition; NV's goals are certainly not just to sell shovels. &#34;Three computers,&#34; &#34;mini supercomputers,&#34; COSMOS—NV seems to want to build a huge AI operating system based on computing power. For super applications of AI like robotics and autonomous driving, NVIDIA will definitely have a hand in it. The competition among giants is likely to accelerate the industry.
The market has risks, and investment should be cautious. This article does not constitute personal investment advice and does not take into account the specific investment goals, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Invest at your own risk

TSLA

Elon Musk revealed while discussing the Optimus robot that he expects to produce thousands for testing by 2026, with mass production in 2027, targeting 500,000 units. Tesla plans to collaborate with external manufacturers to enter the home market. Management pointed out that the technical difficulty of Optimus is extremely high, and the humanoid design helps it to be compatible with human environments. The current team size is relatively small, and developing robots for specific tasks may achieve commercialization more quickly

- Tesla plans to produce several thousand Optimus robots by the end of 2025 for internal testing, with a goal of tens of thousands in collaboration with external partners by 2026.  
- The company aims to achieve a bill of materials cost of $70,000 to $80,000, targeting a reduction to below $35,000 if production reaches 500,000 units in 2027.  
- Challenges include the complexity of robot design compared to vehicles, the need for humanoid form for interaction in domestic environments, and the current limited team size focused on simple tasks.

Let's talk about Elon Musk's robot satellite