Technology tycoons talk about chip opportunities, and the AI trend is ushering in a big turn!

1. Musk × Wei Zhejia: Betting on the future of Dojo, shifting focus to robots

• Meeting core:Tesla’s self-developed Dojo training chip guarantees stable production capacity
• Chip specifications: 5nm process, 5×5 on-chip system array, InFO-SoW advanced packaging, running as a single chip overall
• Real use: mainly used forVery large-scale neural network training, not directly for use with Optimus robots
• Strategic significance: The competitiveness of next-generation products such as autonomous driving and robots depends entirely on Dojo’s computing power.
• Boss’s stance: Musk speaks outThe future focus is on multifunctional robots, not cars; Wei Zhejia promises that funds will be in place and production capacity will be guaranteed.

2. A major shift in the trend of AI: GPU vs ASIC changes from complementarity to competition

• Past: General-purpose GPUs (represented by NVIDIA) dominate AI training and inference
• Now: The industry beginsGet rid of over-reliance on a single GPU, turnCustom ASIC route
• The signal is obvious:
  • Tesla self-developed Dojo
  • Broadcom and other ASIC manufacturers' stock prices strengthen, Nvidia faces pressure
  • Major manufacturers have customized exclusive chips in pursuit of lower costs, higher energy efficiency, and safer supply chains.

3. TSMC: No matter which side wins, it will be the biggest beneficiary

1. Absolute monopoly on process and packaging

   3nm continues to increase its volume, with advanced processes accounting for nearly 70%. Apple, Nvidia, AMD, Broadcom, and Tesla are all highly dependent on it.
2. unique neutral positioning

   It only does foundry and does not produce end products, so customers do not need to worry about competition. The ASIC camp especially trusts TSMC.
3. Foundry 2.0 builds barriers

   Extended from simple manufacturing toAdvanced packaging (CoWoS/SoIC), providing a one-stop solution for process + packaging,

   Top-end chips such as Tesla Dojo must also use TSMC technology in full.
4. Benefit both ways
   • In order to maintain its position, Nvidia increases its film investment
   • The rise of new ASIC forces also relies on TSMC’s advanced technology
     Both parties have jointly pushed up TSMC's orders, and there are sufficient funds for 2nm production expansion and advanced packaging construction.

4. Summary of industry trends

1. AI computing power structure reconstruction

   General-purpose GPUs still dominate, butThe share of dedicated ASICs is increasing rapidly, becoming the second growth curve.
2. Self-research driven by supply chain security

   Giants have stopped making cores one after another. The essence is to reduce costs, optimize structures, and ensure independent control of the supply chain.
3. TSMC’s position further strengthened

   Whether GPU or ASIC, they will eventually be produced at TSMC. The more the industry shrinks, the more stable the foundry will be.
4. The next wave of hardware hot spots: robots

   Technology giants such as Musk have made it clear that robots will replace cars and become the core engine of chip demand.