I have been working on a design for an internet-of-things type device: it doesn’t do very much, but it needs Wi-Fi.
Right now I’m prototyping with Microchip parts (PIC32 MCU, Wi-Fi transceiver), and while it’s a great combination, the cost is pretty high-- about $6.50 for the MCU and $12.50 for the Wi-Fi module (5000 unit volume).
I’m wondering if this is about as low as I’ll realistically get at these volumes, or if I could do better if I went with a different processor/Wi-Fi module combo, and if so, which one? I have had several Chinese vendors offer Wi-Fi chips with Broadcom or Marvell Wi-Fi parts for about $3, but they offer no meaningful documentation. Neither does Marvell or Broadcom, for that matter, since they don’t consider me a customer, and won’t let me sign a NDA unless I am one.
Also, I’m not wild about sourcing parts from random Shenzhen vendors since I have no way of knowing if the parts they get are genuine or not.
If your device “it doesn’t do very much” you may be able to use a PIC16 processor, which is much cheaper than the PIC32. I don’t think you are going to get a fully funcitonal Wi-Fi radio for less than $12.50. The Broadcom & Marvell chips require a lot of other parts, and design expertise, to make a Wi-Fi radio. Not to mention the certification costs.
You may want to look at the GainSpan GS1011Soc http://www.gainspan.com/gs1011soc1. Unlike the Microchip part that requires an external uP, this module has a on-board CPU that can run your application code. It has SPI, UART, I2C, PWM and ADC I/O. Single quantity price is $20 - the 5K pricing should be cheaper than the Microchip solution.
From experience I can tell you Broadcom is not interested in a 5K volume customer, it’s not worth their while to support you. Even at 100K they probably won’t talk to you, let alone believe you.