Industry Voices—Madden: Huawei's inventory bubble soon may burst

Joe Madden

Somehow, executives at Huawei knew in advance that the U.S. government would be coming after them. During late 2018, Huawei started ramping up their purchasing of key semiconductors for their 5G base station platform. By May 2019, the company had shipped at least 30,000 base stations and had inventory equivalent to 100,000 additional base stations using 64T64R MIMO. 

Yes, that’s a pile of 20 million amplifiers and millions of FPGAs. It’s a commitment of at least $400 million for "extra" inventory, in addition to their immediate production requirements. (If you are interested in the detailed semiconductor forecast, click here.)

This is a remarkable contrast with the ZTE situation last year. When ZTE was cut off from American semiconductors, they were shut down almost immediately.

Today, Huawei is shipping 5G base stations using their pile of inventory, and we estimate that they have about four months of runway left before the plane must fly on its own. They’re scrambling with “Platform C” implementation, which is a variation on their 5G base station design without any American components.

Nobody outside of Huawei’s lab has seen a prototype of “Platform C,” but we have some estimates of performance based on component level benchmarks:

  • The integration of ASIC and ADC/DAC is a key part of making 64T64R work without excessive heat. In the short term, a return to a HiSilicon ASIC without ADC/DAC integration would result in a very hot radio head.  
  • If HiSilicon integrates the ADC/DAC technology that they have access to, we believe that they’ll have difficulty with EVM and spurious emissions requirements.  
  • The ARM core is an important part of the HiSilicon product line, and takes care of some key functions in the radio head design at low power. Can Huawei either use the ARM core without a license, or somehow duplicate the ARM core’s function in their own design? We don’t see a straightforward answer.
  • Other components such as power amplifiers, filters, and fiberoptic transceivers should be available from Japan and Europe, so the list of truly critical devices has become shorter.

At a political level, it’s important for the U.S. government to understand that Huawei is not feeling the pressure in the base station market today. They can keep running for a few more months… and then the noose will tighten.   

Even when Huawei runs out of inventory of Xilinx RFSoCs, we believe that they will be able to produce base stations for the domestic Chinese market. The “Platform C” base station is likely to be klunky, with fans on the radio head and poor EVM/linearity performance. But, they are selling most of their 5G base stations to friendly customers in China! So we expect Huawei to get government waivers for compliance issues, and customer waivers for ugly fans and other workarounds.

The problem for Huawei will be in their overseas 5G sales. Will Huawei be able to sell 5G base stations with crappy performance in Germany or Denmark? No way. That’s where this game will get really interesting.

Joe Madden is principal analyst at Mobile Experts, a network of market and technology experts that analyze wireless markets. The team provides detailed research on small cell, base station, carrier Wi-Fi, and IoT markets. Madden currently focuses on trends in 5G, IoT, and enterprise markets for wireless infrastructure. Over 26 years in mobile communications, he accurately predicted the rise of digital predistortion, remote radio heads, small cells, and a mobile IT market. He validates his ideas with mobile and cable operators, as well as semiconductor suppliers, to find the match between business models and technology. Madden holds a physics degree from UCLA. Despite learning about economics at Stanford, he still obeys the laws of physics.

"Industry Voices" are opinion columns written by outside contributors—often industry experts or analysts—who are invited to the conversation by FierceWireless staff. They do not represent the opinions of FierceWireless.