The first address of each subnet may be the broadcast address.
"WAIT", I hear you scream. "I thought the LAST address of each subnet was the broadcast address!"
Yeah, that is what they teach you, isn't it? "They" just expect people to be trusting and follow whatever rules people teach.
Well, the secret is: it depends on what equipment you're using. Some older equipment actually used the first address as the broadcast address. (This may be from back in the days when other technologies, like perhaps Token Ring instead of Ethernet, were competing with the standards that are more commonly used today.) People started treating the first address as a "Network ID", perhaps in part because they were trying to avoid using that address. People spread the advice to treat that as "unusable". The common reason is just that it was a "network ID". And the results seemed positive: as long as people follow that advice, people managed to avoid conflicting with that old equipment. And eventually, some networking hardware/software may have started to make the assumption that people follow that recommended advice. So rules were enforced to follow that recommended advice, which means that now even more equipment (including some newer stuff) may not be able to use such addresses. And so now, the result is that even with modern standards, instead of having only 75% of IPv4 addresses in a /30 being recognized as being usable, we have 50% available. Instead of 25% waste, we have 50% waste.
I cover this, perhaps in more detail at: ][CyberPillar][ glossary: Network IDNote 1 and/or the glossary entry for Broadcast address. (For instance, even more RFCs are referenced there.)
Hopefully that helped to explain why.
Note 1 Full disclosure, I started the ][CyberPillar][ site