Recently, Elon Musk announced on social media that a technical review of the Tesla AI5 chip design will take place this Saturday, and a series of meetings regarding artificial intelligence, the Optimus humanoid robot, and automotive production are scheduled for next week. This Saturday’s technical review will unveil Tesla’s ambitions for the next-generation AI chip and its synergy with three core businesses, quickly attracting the attention of the capital markets, leading to a 2.3% increase in Tesla’s pre-market stock price.
From a technical core perspective, Tesla’s chip technology continues to iterate.The AI3, which went into mass production in 2019, had a computing power of 144 TOPS, supporting the first-generation FSD system; the AI4, which went into mass production in 2023, reached 500 TOPS, supporting Transformer models. The AI5 chip, expected to be equipped with HW5.0, may exceed 2000 TOPS, specifically designed for end-to-end AI architecture.
This review has several key highlights: In terms of energy efficiency, it adopts TSMC’s 3nm process, reducing power consumption by 40% compared to AI4;video processing capability has significantly improved, supporting the processing of 480 frames of video data per second, while the current AI4 only supports 160 frames; it also features multimodal fusion capability, with a neural network architecture capable of synchronously processing visual, radar, and acoustic signals. Chip team leader Pete Bannon revealed that the AI5 will achieve “in-vehicle training” functionality, allowing vehicles to learn from driving environments in real-time and share data.
Musk’s scheduling reflects three strategic intentions.First, to achieve a closed-loop of self-developed hardware, with full-stack self-research from chips (AI5), servers (Dojo) to devices (robots/cars);Second, to leverage data network effects, with 5 million Tesla vehicles generating 400 PB of training data daily;Third, to adopt a cost-dilution strategy, using the AI5 chip for cars, robots, and energy storage devices, significantly reducing unit costs.
According to Tesla’s Q2 financial report, investment in AI training computing power increased by 280% year-on-year, with investments in the Optimus robot production line reaching $2.3 billion, aiming to reduce robot manufacturing costs to below $20,000 by 2026.
Special Statement: The following content does not constitute any investment advice, guidance, or commitment, and is for discussion purposes only.
Core Target Analysis
1. Top Group (601689)
Advantages:Self-developed planetary roller screw has a lifespan of 2000 hours, with costs 40% lower than overseas, serving as the exclusive supplier of joints for Tesla’s Optimus, providing rotary/linear joint actuator assemblies, with a production capacity of 1 million sets/year in Ningbo.
2. Sanhua Intelligent Control (002050)
Advantages:Provides serpentine cold plates for Tesla’s 4680 batteries, with a global market share of 32% in electromechanical actuators, being a core supplier for Tesla’s thermal management, responsible for the robot’s thermal management system and rotary actuators, with a revenue target of 1 billion yuan for the robot business by 2025.
3. Harmonic Drive (688017)
Advantages:Products compatible with forearm/wrist joints, with over 60% market share in domestic harmonic reducers, serving as a first-tier supplier for Tesla’s Optimus, with harmonic reducers accounting for 20% of the robot’s total cost.
4. Mingzhi Electric (603728)
Advantages:Among the top three in global market share for hollow cup motors, with Tesla’s dexterous hand using 17 units, is a potential core supplier of power sources for Tesla’s dexterous hand, providing hollow cup motors (for dexterous hand drive).
5. Lianchuang Electronics (002036)
Advantages:8-megapixel cameras compatible with FSD pure vision solutions, supporting high dynamic range imaging for adverse weather, being a core supplier of lenses for Tesla,holding over 50% share of the front/rear view lenses for Model 3/Y, providing onboard lens modules.
6. Desay SV (002920)
Advantages:L3-level autonomous driving solutions have passed Tesla’s FSD certification tests, leveraging the advantages of NVIDIA Orin chip foundry, becoming a collaborative developer of domain controllers, providing intelligent driving domain controllers.
★ Creating is not easy, and I hope everyone can support more. Your like, share, and casual sharing are the greatest motivation for the Stock Model Square public account to move forward~
If you need to reprint, please contact the backend and indicate the source while retaining the original author’s signature.
The content of this article is sourced from public information and is for reference only. No guarantees are made regarding its accuracy, completeness, or timeliness. This article cannot be used as an investment basis; investing involves risks, and caution is advised when entering the market.
Stock Universe Database
The Stock Universe Database is a technology financial enterprise focused on the research and development of financial database technology, data rebirth, data organization, data modeling, and data training, developing a new generation of financial database base model through cutting-edge data technology, building a more intelligent, rational, and scientific investment data stack support. It independently develops a financial holographic chain analysis database and model algorithm AI technology, while innovatively developing the database “model” system, empowering a new generation of investors, aggregating the power of the era’s data flood, and creating application paradigms.
Database Vision: Let data run for humanity
Core Technical Advantages of the Database:
1. Database base stacking-driven black land quantitative effect
2. Financial holographic chain analysis database-driven AI services
3. Model algorithm AI technology-driven zero-sum game
4. Data model-driven training academy (Zhibiao Academy)
5. Combination model-driven investment engine value (Universe Model Square)
Scan the QR code
FollowStock Model Square WeChat public account
Seeking likes
Seeking shares
Seeking likes
