Apple & Semis – Integration game

Qualcomm helps Broadcom for once.

• Although Apple is intending to design Broadcom out of its iPhones in favour of an in-house solution, I suspect that its inability to design a good 5G modem that works at millimetre wave frequencies will keep Broadcom in Apple devices for longer than many expect.
• I think that there is plenty of evidence that Apple is the finest designer of applications processors available today.
• The combination of excellent performance and battery life with the advantages of hardware and software integration has been leveraged to great effect and the rest of the industry is racing to catch up.
• However, when it comes to communications and modems, it is a completely different game and here, Apple is still reliant on the expertise of others when it comes to radio design.
• The two leaders here are Qualcomm for cellular modems and Broadcom for Wifi, Bluetooth etc both of whom still derive a meaningful portion of their revenues from Apple.
• It is no secret that Apple buys its 5G modems from Qualcomm through gritted teeth as it was its need for 5G that forced it to settle with Qualcomm in what was an ignominious defeat in its long-running legal battle (see here).
• It has been working hard ever since then to design its own 5G modem that can compete with Qualcomm (especially at 28Ghz and above) so that it can stop buying them from outside.
• This was expected to start happening in 2023 but I have long been of the opinion that 28Ghz and above is much harder to do than many think meaning that it will be a while longer before Qualcomm starts losing share at Apple.
• 28Ghz is difficult enough but on top of that 5G uses any number of radio bands meaning that making a single phone that can work anywhere is very difficult given the number and combination of radio bands that need to be supported.
• In practice, making cellular modems is very difficult and in every generation, since WCDMA (3G) Qualcomm has always managed to get working technology to market long ahead of anyone else.
• In 5G this meant that if Apple wanted a 5G modem that supported 28Ghz as many of its US customers demanded, it had to get it from Qualcomm.
• Bluetooth and WiFi are also pretty difficult but not in the same league as cellular which is why it should be relatively straightforward for Apple to design Broadcom out.
• However, I think that Apple is intending to replace both Qualcomm and Broadcom with a single SoC for connectivity which would save space, power and offer greater opportunities to integrate its software with its silicon.
• I don’t see much advantage to Apple in having its own WiFi and Bluetooth chips and keeping them separate from the cellular modem.
• This means that there is little point in removing Broadcom before, it is ready to remove Qualcomm and I think that this is going to take it some time yet.
• Many radio engineers will describe analog radio as a black art where it is more about one’s experience as opposed to knowledge of the laws of physics.
• This is why Qualcomm guided for more revenues than expected from Apple this fiscal year which I expect to be repeated next year and maybe even the year after.
• This is good news for Broadcom because this may delay the loss of its biggest customer and give it time to find other places to sell its wares.
• The outlook for semiconductors remains pretty difficult this year especially as Samsung has just kicked off Q4 2022 reporting with a 70% decline in profits and so I continue to look to the value end of the sector.
• Here I find TSMC, Qualcomm, MediaTek and Samsung to be the most attractive although I remain somewhat nervous about owning factories at the moment.