[kyboren]

For their handset SoCs, sure. But the real prize appears to be the base station market. As I understand it, Huawei's base stations are critically dependent upon Xilinx FPGAs. With those no longer available to them, Huawei is in a difficult position.

They can make their own FPGAs, true. These FPGAs will not be commercially competitive with Xilinx and Intel, so development costs won't be defrayed by commercial sales. IP laws may effectively prevent their sale in Western markets, anyway. And base stations made with these FPGAs won't be as commercially competitive, either (they'll be larger, more power hungry, produce more heat, etc.).

I don't know enough about 5G or base stations to say with confidence what other alternatives they might have. But I'm sure that they will be sub-optimal, at the very least imposing billions of USD in costs to Huawei/PLA/CPC.

[sudosysgen]

I wouldn't be so sure. 14nm is already competitive with Xilinx FPGAs in terms of efficiency. All Huawei has to do is build ASICs instead, and they should keep or improve their performance. Xilinx is on a mix of 20nm and 16nm, by the way.

The billions of dollars that the Chinese Government will invest in order to compensate for the sanctions might have the unintended side effects of actually making performance of Huawei gear more efficient. The jump from FPGA to ASIC is pretty big as far as efficiency, and being able to have it financed for free to defend against US sanctions is pretty sweet.

They also happen to have stockpiled enough FPGAs to buy time for the transition.

[kyboren]

They are not currently using ASICs, so ASICs are presumably not a good fit for the application. Maybe the NRE costs are too excessive given the very small volume, flexibility/updateability is essential, TTM is too long, something else, or some combination of the above. Any way you slice it, being cut off from Xilinx is going to cost Huawei.

Building a competitive ASIC will also be significantly more challenging now that Huawei has been cut off from all the major EDA companies.

Finally, I assume that Xilinx is well into development on future product lines based on newer processes. It would be extremely difficult for Huawei to bootstrap their way to commercially competitive FPGAs.