DeepSeek R2 launch delayed due to performance concerns
Chinese AI startup DeepSeek has put the brakes on its R2 model launch after CEO Liang Wenfeng said the performance just wasn't there yet.
The R2 was meant to boost coding and multilingual skills over the older R1, but it's still being tweaked, with no launch date in sight.
US export rules blocking new shipments
Another big reason for the delay? There aren't enough NVIDIA H20 GPUs—the hardware needed to run these models.
Even though Chinese tech companies have invested heavily in these chips, US export rules have blocked new shipments.
That means cloud providers can't roll out R2 widely yet, and everything depends on when more chips become available.
Customers are sticking with the R1 model
For now, DeepSeek's customers are sticking with the R1 model on existing NVIDIA H20 chips.
The company wants to level up with R2, but both tech hiccups and hardware shortages are slowing them down—especially as US restrictions keep tightening.