I don’t think this is entirely or even mostly Epic’s fault.
Unreal Engine 5 is a very forward-looking engine in terms of its features, which does translate to relatively high hardware requirements if you use every one of them in an improper manner, however: If you adjust the settings for Nanite and Lumen and use the engine for what it’s good at, you can quite dramatically speed it up at a low cost to visual fidelity (example: Fatekeeper, which is gorgeous, runs well and is being developed by a tiny team; exactly the promise of this engine) - or you can more selectively employ only some of them for a more performant game (example: ARC Raiders, which employs UE5 without Nanite nor Lumen - still looks good and runs on a toaster).
The main problem however is that Epic inadvertently chose the worst timing possible for releasing these expensive new features (regarding processing power and memory cost) to their engine. All versions of UE - or, to be more precise, games released with them - have had relatively high hardware requirements when they were new. The kind of hardware that was necessary to run the first Unreal back in 1998 the way it was meant was eye-wateringly expensive, adjusted for inflation. However, what used to happen was that prices for suitable hardware quickly dropped at the same time as performance increased. This was annoying in some ways - a high-end 1998 gaming PC was pretty much obsolete for the latest titles by the year 2000 already - but it also meant that as the years went on, more and more people could afford to enjoy playing good-looking games based on this increasingly capable and more accessible engine.
The quadruple-whammy of the crypto boom, AI mania, rising global inequality and hardware capabilities plateauing or even degrading (the 2022 RTX 4090 is more powerful and cheaper than the rumored upcoming 24 GB 5080, which is set to replace the 5090 as Nvidia’s gaming flagship) means that releasing an engine with seductive labor-saving and artist-friendly features like geometry that doesn’t have to be simplified (Nanite) and lighting that just works™ (Lumen) results in below average and average developers more or less happily rushing out games that would have previously been impossible to make without incredibly talented teams - but the downside is that due to a discrepancy between increasing engine complexity (easy to learn, hard to master) and the cost of hardware, those games do not run as well as they should.
Empowering developers with a capable, yet accessible engine is always a double-sided coin: On one hand, it can allow people with tons of talent and limited budgets to punch way above their weight class (like the aforementioned developers of Fatekeeper), but on the other hand, it also results in low quality games from less intelligent (or just greedy) people that tarnish the name of the engine behind them. None of this is new: When Unity gained in popularity, we got both great Indies and a still ongoing avalanche of abysmal asset flips (and I predict the same for Godot). Unreal Engine 3 was the first version of UE that was widely accessible to smaller and mid-sized studios, but many of especially early games for it looked and ran extremely poorly, as even experienced developers were like kids in a candy store trying out every flavor (new rendering technique) all at once, even if the result was often a shiny, blurry, bloomy mess that stuttered more than it ran on contemporary hardware.
I don’t think this is entirely or even mostly Epic’s fault.
Unreal Engine 5 is a very forward-looking engine in terms of its features, which does translate to relatively high hardware requirements if you use every one of them in an improper manner, however: If you adjust the settings for Nanite and Lumen and use the engine for what it’s good at, you can quite dramatically speed it up at a low cost to visual fidelity (example: Fatekeeper, which is gorgeous, runs well and is being developed by a tiny team; exactly the promise of this engine) - or you can more selectively employ only some of them for a more performant game (example: ARC Raiders, which employs UE5 without Nanite nor Lumen - still looks good and runs on a toaster).
The main problem however is that Epic inadvertently chose the worst timing possible for releasing these expensive new features (regarding processing power and memory cost) to their engine. All versions of UE - or, to be more precise, games released with them - have had relatively high hardware requirements when they were new. The kind of hardware that was necessary to run the first Unreal back in 1998 the way it was meant was eye-wateringly expensive, adjusted for inflation. However, what used to happen was that prices for suitable hardware quickly dropped at the same time as performance increased. This was annoying in some ways - a high-end 1998 gaming PC was pretty much obsolete for the latest titles by the year 2000 already - but it also meant that as the years went on, more and more people could afford to enjoy playing good-looking games based on this increasingly capable and more accessible engine.
The quadruple-whammy of the crypto boom, AI mania, rising global inequality and hardware capabilities plateauing or even degrading (the 2022 RTX 4090 is more powerful and cheaper than the rumored upcoming 24 GB 5080, which is set to replace the 5090 as Nvidia’s gaming flagship) means that releasing an engine with seductive labor-saving and artist-friendly features like geometry that doesn’t have to be simplified (Nanite) and lighting that just works™ (Lumen) results in below average and average developers more or less happily rushing out games that would have previously been impossible to make without incredibly talented teams - but the downside is that due to a discrepancy between increasing engine complexity (easy to learn, hard to master) and the cost of hardware, those games do not run as well as they should.
Empowering developers with a capable, yet accessible engine is always a double-sided coin: On one hand, it can allow people with tons of talent and limited budgets to punch way above their weight class (like the aforementioned developers of Fatekeeper), but on the other hand, it also results in low quality games from less intelligent (or just greedy) people that tarnish the name of the engine behind them. None of this is new: When Unity gained in popularity, we got both great Indies and a still ongoing avalanche of abysmal asset flips (and I predict the same for Godot). Unreal Engine 3 was the first version of UE that was widely accessible to smaller and mid-sized studios, but many of especially early games for it looked and ran extremely poorly, as even experienced developers were like kids in a candy store trying out every flavor (new rendering technique) all at once, even if the result was often a shiny, blurry, bloomy mess that stuttered more than it ran on contemporary hardware.