The magic number to remember is 2.5.
You will find that most scopes have a max sample rate of around 2.5 times the maximum bandwidth.
Take the Rigol DS1054Z - 4 channels, 100 MHz, 1 GHz sample rate. Multiply 4 ch * 100 MHz/ch * 2.5 samples/Hz and you get the magic number 1 GHz. This is related to the 2x sample rate requirement given by Shannon and Nyquist - sampling theory.
The question is how many channels are active. The DS1054Z is smart enough to use a faster sampling rate if fewer channels are enabled. As are most scopes...
There are two major players in the entry level scope arena: The Rigol DS1054Z at well under $400 which, when unlocked (details elsewhere) to 100 MHz, is a very capable 4 channel scope. Since you are planning to mess around with microcontrollers, I'm going to recommend 4 channels and the ability to decode SPI.
The other is the newish Siglent SDS 1104X-E - a more recently released 4 channel 100 MHz scope. It's around $500 and, according to reviews, the user interface is much more responsive. Look for threads in the Test Equipment forum here on EEVblog.
The Rigol has the advantage of being around for 3 or 4 years and has gone through several firmware revisions. The Siglent is newer and I'm not tracking firmware issues.
For my money, I would probably buy the Siglent SDS1204X-E to get 4 channels and 200 MHz. But it's around $750 and I'm not quite willing to spend that much right now. I do have the Rigol.