Elon Musk officially releases the Terafab project: aiming to produce 1 Terawatt of computing power per year, primarily deployed in space.

SpaceX and Tesla, Inc. (TSLA.US) jointly announced the Terafab project at 8 p.m. on Saturday night in US Eastern Time, and live-streamed it on the X platform. The goal of the Terafab project is to manufacture 100 billion to 200 billion advanced 2-nanometer chips annually. SpaceX founder and CEO of Tesla, Inc., Musk, retweeted the live stream content and stated that the target annual production capacity of Terafab is to produce 1 terawatt (trillion watts) of computing power. Since the electricity generation capacity in the US is only 0.5 terawatts per year, most of the future computing power will have to be shifted to space. In an earlier tweet, Musk also pointed out that the Terafab AI chip manufacturing will cover logic, memory, and packaging fields, with 80% of the produced chips being used in space and the remaining 20% for ground use. Terafab will be located in Austin, Texas, USA, and will be jointly operated by SpaceX and Tesla, Inc. Grand Plans Musk has previously stated that despite the continuous increase in semiconductor industry production, the development speed is still too slow to meet his expected chip supply. This is one of the key reasons he is pushing for Terafab. Musk emphasized at the Terafab launch event that the project is expected to produce two main categories of chips. One category will be optimized for edge computing and inference, mainly for use in Tesla, Inc. electric vehicles, autonomous taxis, and Siasun Robot & Automation. The other category is high-performance chips designed for space applications, available for use by SpaceX and xAI. Among them, AI5 is designed as a dedicated chip for Tesla, Inc.'s fully autonomous driving system and Siasun Robot & Automation, with a performance power increase of 40 to 50 times compared to the existing AI4. The next-generation AI6 is expected to complete design by the end of this year, mainly for use in the second generation of Siasun Robot & Automation and large-scale inference clusters. The D3 chip is designed for the space environment and will provide computing power for projects like Starlink and Starship. At the launch event, Musk also showcased the design of future mini artificial intelligence data center satellites, which are part of SpaceX's large satellite system aimed at conducting complex space computing. In January of this year, SpaceX applied to the US Federal Communications Commission to launch one million data center satellites. However, Musk himself has no experience in semiconductor production, and given the huge investment scale and complex and precise production equipment in the industry, Terafab may not be able to achieve mass production in the short term. In the live-streamed launch event, Musk did not provide a specific timetable for the construction or mass production of the Terafab project. Raising funds for the construction and launch of artificial intelligence data centers in space is one of SpaceX's main drives to conduct an IPO later this year. According to reports, SpaceX is expected to conduct an IPO this summer, with a valuation possibly exceeding $1.75 trillion. This article is from "Caishen.com", authored by Ma Lan, GMTEight, edited by Chen Siyu.