Mergers are mostly guided by revenue opportunities rather than common sense in terms of technical products. Been there, seen it happen in other markets.
I think it's very unfortunate too Microchip has not tried harder with their PIC24. It's a very nice answer to the shortcomings of their 8-bit chips. However the 16-bit MCU market hasn't really been growing for ages now. If you look at percentages of revenue, the 16-bit market is actually shrinking. Does it make sense to invest more into 16-bit? Nope. Their investors probably want MCP see investing in "upcoming markets".
Microchip has got their 8-bit covered from a business point of view. They announced a little while back very proudly they have the biggest revenue of the 8-bit market. How can that happen if you read the amount of bitching on their 8-bit architecture? It probably all has to do with "bolt on" peripherals, pricing and availability. Have you tried to buy some AVR chips in any quantity? They are very expensive .. any PIC and most MSP430 or modern ARM will beat that.
So it makes sense that Microchip acquired Atmel because of their ARM offerings. Atmel ARM listings far exceeds what type of AVR chip you can get. Atmel has Cortex A5 or ARM926 chips that run some flavor of Linux, back down to cortex m0+ for IoT and whatnot.
The trend is already that 32-bit is the new 8-bit, so suppose this solidifies completely in the upcoming (let's say) 10 years it makes sense they really needed to strengthen their 32-bit portfolio. Because although I can personally appreciate PIC32MX having worked with them a few years back, out-sight is slightly different:
- Microchip misses all the love ARM gets. I appreciate the PIC32 and MIPS4K in it self though but honestly it's not enough to be convincing.
ARM wins IMHO on all aspects of development and firmware parts now, offering more alternative toolchains (from free mbed cloud to IAR/KEIL with trace and RTOS inspectors), RTOS's, complexity & options of it's peripherals, etc.
- Full debugger + dev boards + toys for ARM can be had for 20$. If you just want to evaluate a MCU peripheral it's no big deal to order such board and use it for only 1 hour to figure out if a peripheral will work out for you.
From my point of view, there is no answer from Microchip. Their PIC32MZ EF starter kit is already 100$, that still only gets you a PICKIT3. Although it works, it is outperformed by all onboard debuggers on ARM boards I've seen.
- Energy friendly with PIC32 is impossible.
PIC32MX250: 500uA/MHz, 44uA sleep current.
PIC32MZ EF: 460uA/MHZ, 1.5mA sleep current.
SiLabs Pearl Gecko (48MHz Cortex m4): 60uA/MHz, 2uA sleep current
Most other ARM chips are probably in 100-200uA/MHz range with <10uA sleep currents.
- No PIC32 that can run Linux.
- No PIC32 that can handle SDRAM
- LCD TFT hardware controller with 2D DMA anyone?
- No PIC32 with house-keeping DAC. Only 12-bit house-keeping ADC on their latest PIC32MZ
- PIC32MZ has been a bit of a fail in terms of silicon bugs. Since then ARM Cortex m7 has come out with in general more peripherals, yet higher clock speeds and an even faster instruction pipeline (up to 2IPC)
PIC32 still applies for many other applications (not everyone will jump straight into a Linux project with LCD TFT and SDRAM, or ultra-low power IoT wireless nodes), but I can see how Atmel chips would fill huge gaps in Microchip's 32-bit offering.