Custom chips run faster, draw less power, and need less cooling. It’s standard practice to use reprogrammable FPGA’s during test and low volume production. Nearly all designs move to custom when they reach enough volume, 10’s or 100’s of thousands.
Huawei has now shipped 400,000 5G base stations. That it is switching to dedicated chips normally wouldn’t be worth a story. However, experienced analyst Joe Madden told a major newspaper the Europeans wouldn’t be happy with the new units. I see no evidence Europeans carriers are pulling back from Huawei. Two confirmed to me they will happily buy the Huawei radios without the US parts.
He noted that Huawei will “run out of Xilinx FPGAs in November and would have to change over to their own parts.” For practical purposes, US companies Xilinx and Intel/Altera are the only suppliers of high-end FPGA’s. Would that cost sales?
Last week, I had a chance to ask <a very senior technical person at one of Europe’s largest carriers.> I also asked the CTO of a carrier with over 10M customers. I’ve respected both for over a decade. Without getting clearance from their pr departments, I can’t use their names.
Both told me they would not be deterred from buying the new radios without the US parts. Both are currently buying radios from Huawei and plan to continue doing so.
ZTE already is using a custom chip and I’m told Nokia, Ericsson, and Samsung are well along in producing their own custom chip. While the ability to reprogram FPGA’s in the field is valuable, the power and speed disadvantages are even more important. They are no longer a natural choice for 5G radios, now shipping in the hundreds of thousands.
China is adding 20-40,000 5K radios per month, Europe probably fewer than 2,500. Europe is going so slow on 5G any loss of sales there is dwarfed by the demand in China.
Important conflict of interest notice: Huawei pays my actual expenses to some of their events. I don’t have a large company behind me so have to accept expenses.