The technology that YouTube provides was hard to build when YouTube started a decade and a half ago, but surely today it’s not a huge challenge.

PDP has 20 billion total views. He doesn’t need traffic from the algorithm suggesting his videos, everyone else is trying to game the algorithm to get redirected by PDP!

The problem is that building a platform to enable those 20-billion views carries enormous fixed costs that only make sense when they are amortized across a truly massive amount of users, both in terms of uploaders and users. Video delivery at scale is one of the most difficult engineering problems out there. The only companies that have mastered it (YouTube, Vimeo, PornHub, Netflix, Amazon) are all billon dollar enterprises.

Sure, PewDiePie could pay to build out his own video service. But would it be as good as YouTube? It's very doubtful that it would have the level of polish that YouTube offers. YouTube is far more than just tossing up a bunch of .mp4 files on a web server.

Finally, I think you're underestimating the power of YouTube's algorithms. When Logan Paul (another YouTube celebrity) got delisted from YouTube, he suffered a massive revenue hit, even though his videos were still on the platform (but not showing up in search results). So I do think that PewDiePie is beholden to the algorithm. I would be willing to bet that if PewDiePie got delisted from YouTube, he would rapidly be forgotten, and would be replaced by the next YouTube celebrity willing to walk the fine line between "outrageous enough to be entertaining" and "so outrageous as to cause offense".

Edit: Scott Alexander has addressed the part of your question regarding hosting other comedians on his excellent post, Freedom on the Centralized Web. He correctly points out that the initial group of switchers are all going to be people who YouTube has deemed undesirable. However, YouTube deeming people undesirable is an effect. The cause is that these people have offended some powerful group (copyright holders, activists, etc). If all of these people abandon YouTube and start their own platform, the same forces that kicked them off YouTube will ensure that their new platform is starved of funding and respectability. For a good example of this, look at what happened to Gab. I don't support Gab, but the saga of Gab shows how difficult it really is to set up an entirely independent platform, which supports content that society doesn't approve of.

Just to understand the scope of the task:

In the last month PDP averaged over 14 million views per day, with peaks above 20 million. At 1MB per minute for the lowest video quality (240p) that gives you a floor of 14 terabytes per day per minute of video watched. Multiply that by 10-12 for 1080p video. Assuming users watch only 10 minutes of PDP per day, that gives you somewhere between 140 and 1400 TB per day of video delivered reliablely, planet wide, everyday. As a floor that requires delivery of 1.66 GBytes/sec, 24 hours a day.

My guess would be that this has a lot to do with IQ differences of both the audience and the creators. First, Sam Harris listeners may have a much easier time learning new apps and websites than PewDiePie viewers. Second, PewDiePie is primarily "famous for being famous" and there are many people producing videos of more-or-less equivalent quality to PewDiePie's, whereas there aren't that many nearly identical podcast hosts ready to step in and replace Sam Harris as soon as his podcast becomes slightly more inconvenient to access.

Yes, there are costs to building your own platform, especially for video, but my guess is traffic is the main limitation. YouTube, Facebook, etc are trying to find things to put in front of people to entertain them, and if you can do well at this you can have an enormous audience. Streaming video from your own website to fans who care enough to seek it out gives you freedom but adds too much friction to seeing your stuff.

Putting technical limitations aside (which are a huge deal, at the very least for video), the problem is that the audiences were built using the platform, and don't carry over easily.

The creators were able to build their audiences because, notably

• The platforms have idle eyeballs actively looking for good content *on the platform*. No one google for content these days, only for answers.
• The recommendation algorithms sometimes work, or at least you can make them work for you. Even if you have to figure out the peculiarities in the algorithm, this is vastly simpler than cracking global marketing. And again, active digital marketing for content typical passes through social media anyway! This is were the people are, and it's where they look, and it's where they will stumble on you if they're not looking.
• The alternative is being so damn appealing that you'll spread by word of mouth. And even then, you'd do better on a platform, it's just an incredible force multiplier.

The audiences don't carry over because, simply put, they are living on the platform. It's centralized. They consume many things there, so they will check it. Most people don't know RSS and it's being phased out of many browsers. You'll lose most of your subscribers.

And you are wrong, the algorithms do account for many of the views of the top creators, on top of their subscribers.

Could they survive without the platform? Of course! Would they do better? No chance.

Finally, anger at the platform is generally at being less good than it used to be. But think about, for instance, demonetization on YouTube. Well, you can still sign your own deals and include your own ads in the videos. If you leave the platform, you have to do this. But if you stay there, it's still an option.

Technical difficulties of development and maintenance of own platforms have been mentioned in other comments.

However, many own platforms lack revenue opportunities provided by centralized platforms. YouTube specifically has a huge benefit of built-in monetization. Most content creators on YouTube start earning money much earlier because YouTube manages ads for them. General trend I see is creators start getting sponsored videos sometime between 500,000 and 1MM subscribers. Depending on channel that can take about a year getting videos out at a regular pace. I hazard a guess that many would've given up much earlier if they had to think about monetization on their own instead of relying on the platform for that.