Intel announced last week that it will join Musk's giant chip project Terafab to participate in chip design, manufacturing and packaging processes, and help Terafab achieve the goal of annual production capacity of one terawatt of computing power. Although this cooperation is regarded as a win-win situation, it will not only allow Terafab to obtain professional technology and experience in chip manufacturing, but also allow Intel's chip foundry department to obtain new customer resources. However, compared with TSMC and Samsung Electronics, which have more mature technologies, Intel's final result of winning Musk's approval still caused some thinking in the industry.

A recently published technical article by Intel may explain the key to this collaboration. On the day the partnership was announced, Han Wui Then, senior chief engineer at Intel's Foundry Technology Research Institute, posted on a community forum that Intel had made breakthrough progress in gallium nitride chips.


new breakthrough

According to the article, gallium nitride chips are a compound semiconductor that is more stable than silicon in high-pressure environments. Intel has found a way to grow gallium nitride chips directly on standard 300mm wafers using standard semiconductor production equipment, allowing for low-cost production.

The researchers also used a new thinning process called stealth dicing before grinding (SDBG), which allowed Intel to create gallium nitride chips with a silicon substrate thickness of just 19 microns. For reference, 1 micron is equal to one millionth of a meter, and 19 microns is only one-fifth the diameter of a human hair.

In addition, Intel has also successfully integrated gallium nitride power electronics and silicon logic circuits on the same chip, which means that in the traditional manufacturing process, the problem of having to separate the power transistors from the logic circuits into two chips due to their excessive size and generating a lot of heat and electrical noise has been solved, thus further shrinking the chip space and reducing current consumption.

According to Intel, this integration performed well in subsequent tests, working properly and maintaining its stability under high-stress conditions. These technological improvements mean for Terafab that it can produce thinner and lighter chips, thereby reducing the weight of the rocket during launch, thereby reducing launch costs.

In addition to the performance optimization of the chip itself, gallium nitride chips have another advantage. They are more resistant to radiation than silicon chips, which means they are more suitable for space operations. One of Terafab's future application scenarios is space data centers.

However, it is unclear whether Intel will directly license Terafab to use gallium nitride technology, or whether it will jointly invest in the Terafab project with SpaceX and Tesla to develop this technology. And given the huge investment in Terafab, the future profitability prospects of Intel and Terafab will still need some time to be tested, and people may not understand the economic impact of this project until a few years later.