Nvidia is allegedly investing a lot of money into the RTX 4000 series' manufacturing to avoid the problems it faced with RTX 3000 GPUs.
As reported by Hardware Times, Nvidia is paying chip foundry TSMC about $10 billion to ensure it has reserved enough capacity to meet demand. Whether that eye-watering sum is enough to ensure a steady supply is another matter though.
The Nvidia RTX 4000 cards are apparently using new Nvidia Lovelace architecture built on a smaller 5nm process, down from the 8nm process used to make the current Ampere-powered cards. Decreasing the process size allows increased and more efficient performance into the same sized chip, so we could see a particularly big performance jump between the current and upcoming RTX cards.
AMD is itself rumored to be adopting the 5nm tech for its RDNA 3 architecture cards, and many other companies are no doubt looking to use it too. That would explain why it has cost Nvidia so much to reportedly book TSMC's services.
Nvidia is apparently planning to launch its new generation of graphics cards before the end of the year, although the bigger question is perhaps if you'll be able to actually get them. It's been well observed how difficult it's been to buy a GPU over recent years, between the global chip shortage, scalping and cryptocurrency mining. While we have been diligently tracking RTX 3080 restock activity and those for other graphics cards from Nvidia and AMD, it's not been easy for users to find one in stock and buy it.
There is reason to hope that with improvements in supply, we won't see the same level of scarcity, and you'll be able to upgrade your PC more easily and without paying way too much for new GPUs. These cards will probably be quite expensive though, given the promised performance increase and still-high demand.