Here’s what we have for you today:
• In advance
• Cost of building
• Releasing substitutions
Cost of building upon AI advancements
Hold on: AI startup(s) in China have stockpiled Nvidia chips prior to the US government’s-imposed export restrictions towards them.
For example: CEO Kai-Fu Lee of 01.AI mentioned his company bought enough of the chips to last for the next 18 months or so.
Expensive build: According to Kai-Fu Lee building AI models are very expensive, because one would have to spend millions on chips to make the technology better.
Japan advancing for next generation chips
The amount: Japan’s Economy Ministry is preparing ¥2 trillion or $13 billion in subsidies to push investment in its chip industry in an effort to regain its status as a major semiconductor leader.
Catching up: The money will be used to speed up the country’s ability to design and manufacture next-generation chips and train AI models.
High priority: Japan’s METI stated “The global economic security situation over semiconductors is becoming more severe.”
Another set of chips from Nvidia for sale
Different designs: Nvidia reportedly releases three dedicated chips for the Chinese market (HGX H20, L20 PCIe and L2 PCIe) in November 2023.
Current state: Nvidia has more than 90% share of China’s $7 billion AI chip market.
The curb: US trade restrictions to China are likely to create more opportunities for domestic firms to sell domestically.