The bigger you are, the more it costs. If you're a popular service, you're likely running many, many servers doing many, many things across a broad geographic area. You've carefully implemented your infrastructure to balance cost, stability, reliability, and performance. Adding just one new feature can completely upend this calculus. If you're running multiple server farms in multiple data centers, this gets expensive quickly.
You can't afford "aw shucks oopsie woos". Whoops! Your new feature caused some unexpected behavior for 15% of all users, resulting in 18 hours of downtime! If you're a small web operation, you're sending out a lighthearted email apologizing for the inconvenience and promising to do better. Maybe you're even offering a week's worth of free service. If you're a major player, you're in the world news. Your enterprise customers are screaming at you--or worse, they're not screaming at you and are looking for your replacement. You're working on figuring out just how much this will impact the bottom line, because if you're going to need to cut back somewhere, you want to know that as early as possible. Mess up hard enough, and you're looking at a subpoena from your governmental bodies of choice.
You can't afford to annoy your users. Ooooh, we've all had that time when we rolled out an awesome new feature and the user response ranged from "meh" to "change it back right now you gibbering twits." That's never fun, is it? Gotta roll back to yesterday's configuration, apologize, and try to figure out how to move forward. If you're a major player, "rolling back" may be nigh impossible, and if you've already reconfigured your infrastructure to accommodate your new feature, that's money already spent (and worse, your new configuration may even be sub-optimal in the absence of said new feature.) You're basically looking at the same outcome as the previous point, perhaps minus the subpoenas and plus a bit more global mockery on social media.
Messing up will cost you users, and those users are unlikely to return. If you're small, this can be weathered, and is almost expected. There are way more fish in the sea, and you if can iron things out, you've still got plenty of room to grow. If you're big, everyone already knows about you and what you do. You've got a lot smaller pool of "new" people to bring on compared to the people you've already reached. Big companies that mess up need to work to retain unhappy customers, because there aren't that many fish in the sea who haven't already heard of them.