About a billion Cortex-M based MCUs from Atmel, ST, TI, NXP, etc. They typically have: GPIO with PWM, UART, SPI, I2C, I2S, Ethernet, USB (host, device, OTG), CAN, RS-423, SDIO...

What peripherals do you think might be missing?

Look at it from a different perspective. I agree that an argument centered on peripheral availability is no longer as applicable as it was just a couple of years ago.

- No chip that needs an OS can truly replace an AVR.

- No chip that takes more than two seconds to boot up can truly replace an AVR.

- No chip with a 2,000-page data sheet can truly replace an AVR.

- No chip that requires binary blobs to do anything interesting can truly replace an AVR.

- No chip that's available only in a BGA or other leadless package can truly replace an AVR.

- No chip that requires a 1-gigabyte toolchain installation (or, Woz forbid, a "cloud"-based toolchain) can truly replace an AVR.

In general, no chip that wasn't explicitly designed to be easy to use by nonexperts can truly replace an AVR. Even lack of 5-volt compatibility has proven a handicap for some suitable candidates like the Cortex-M3 part used in the Arduino Due.

Nobody in the hobbyist market cares if the chip costs $0.50 or $5.00. Their time is worth more than that. An 8-bit AVR can run 32-bit C code all day long without requiring the user to become a Linux sysadmin or wade through a bewildering selection of thousands of RTOSes, none of which are perfect for any one job. That is a huge deal.

You asked what ARM could replace an AVR. I said there were lots of ARM Cortex-M based microcontrollers that could. Your response makes no sense.

> No chip that needs an OS can truly replace an AVR

They don't need an OS.

> No chip that takes more than two seconds to boot up can truly replace an AVR

No boot up time. They start running your code right away.

> No chip with a 2,000-page data sheet can truly replace an AVR.

Maybe more like a few hundred. Just like an AVR datasheet.

> No chip that requires binary blobs to do anything interesting can truly replace an AVR.

They don't. They don't even have GPUs.

> No chip that's available only in a BGA or other leadless package can truly replace an AVR.

There available in hand-solderable TQFP packages. There are even a couple of DIP ones (8-pin and 28-pin).

> No chip that requires a 1-gigabyte toolchain installation

I'm pretty sure GCC isn't anywhere near a gigabyte, but it's also the same toolchain used with AVRs.

> (or, Woz forbid, a "cloud"-based toolchain) can truly replace an AVR.

While they exist, they're not in the slightest bit necessary.

> no chip that wasn't explicitly designed to be easy to use by nonexperts in a hardware project can truly replace an AVR.

AVRs were also not designed for non-experts. Try reading a datasheet sometime and try to tell me that's written for non-experts.

> Even lack of 5-volt compatibility has proven a handicap

Lack of 5-volt "compatibility" is an interesting spin on what is essentially a legacy. One could just as easily argue "lack of 3.3V compatibility".

Sounds like you need to educate yourself a bit. There's more to microcontrollers than Arduino.

> Why do you suppose that is?

Marketing and the Arduino IDE.

(I edited my comment to point out that the AVR chips predate the Arduino platform by several years, and that there were technically superior parts on the market even then.)