WiFi is a difficult protocol, so it makes sense to use a radio coprocessor. The ESP8266 can also have trouble running non-trivial applications while reliably handling WiFi, and in 2020 a Cortex-M0 chip is probably at least as cost effective as an 8-bit micro.

They don't need those chips to make your coffee, they need them to connect to the internet and to avoid hiring expensive experts. 5 years from now this might use a single ESP32-S2, but it doesn't seem outrageous, especially when the machine also has a display.

Plus, as a bonus, you can use it to mine Monero! Small correction to the article, STM32F0s have a PLL so they aren't limited to the crystal's 8MHz frequency. They're rated for 48MHz, but you can also overclock them.