NVIDIA has delayed the launch of “China-focused” AI GPUs such as the H20, amid concerns from server manufacturers regarding the integration of the chip.
NVIDIA H20 & Other China-Focused AI GPUs Face Delay As Manufacturers Reportedly Having Trouble With Chip Integration
For a bit of a background context, with the recently imposed US ban on the export of NVIDIA’s AI GPUs, Team Green decided to yet again find a “workaround” and to do so, the company revealed to launch at least three new AI GPUs for the Chinese market which include the H20 SXM, PCIe L20, and the PCIe L2.
These GPUs reportedly are much more cut-down configurations than what the Chinese customers were getting before in the form of the H800 and A800, but NVIDIA doesn’t plan on leaving the AI markets in China alone, hence the company took the step of surrendering to the US policies once again.
However, it looks like NVIDIA’s “re-entry” into the Chinese markets would be delayed since industry sources, quoted by Reuters, cite that the company can’t seem to position itself into releasing the highest-end of all, the H20 AI GPU. While Team Green was rumored to launch the H20 somewhere in the middle of November, it seems that they have encountered a roadblock, and the release has been pushed into the first quarter of 2024, supposedly near February or March. In terms of the other variants, sources say that the L20 is scheduled to release on time but there is no word on the L2.
The reason behind the delay hasn’t been explained in depth, but it has something to do with the compatibility of the H20 with servers in the Chinese markets. The delay does make sense, considering that the H20 not only is the highest-performing of the new Chinese variants, but it does require a little more attention in terms of manufacturing, compared to the others. NVIDIA is apparently hurrying up to prevent any sort of space for competitors but the delay of H20 could prove to be a dealbreaker for the company.
After the newer export restrictions imposed on NVIDIA, competitors like Huawei saw an opportunity to capitalize on, which is why big tech firms like Tencent and Baidu are actually looking to employ Huawei’s Ascend AI chip for low-power AI tasks, which is one step towards the adoption of “in-house” solutions. This is something NVIDIA doesn’t want, since after complete dominance in the Chinese markets, getting pushed out would prove detrimental to the company, especially to its financial performance.
News Source: Reuters