But isn't there currently a bottleneck in Chinese GPUs?

There seems to be a massive gap between
ASML technology and
current Chinese technology.
Even if Changxin catches up with HBM,
it won't matter if the GPU performance is insufficient.