Should You Buy an RTX 5090 Right Now? An Honest Look at the Options

Short answer: probably not, unless the card earns money for you.

That’s the honest starting point for anyone thinking about whether to buy an RTX 5090 right now. The card carries a $1,999 MSRP from NVIDIA. On 11 September 2026 it was actually selling for $6,788.88, about 240 percent above that figure, according to the price tracker at videocardprices.com. You’re not paying for a graphics card at that point. You’re paying for a graphics card plus roughly two and a half more of them in premium.

There are still people for whom that maths works. Just far fewer than the number of people currently considering it.

Who should actually buy an RTX 5090 today

If the card generates revenue, the conversation changes completely. Someone running paid rendering work, video production on deadline, or compute workloads that bill by the hour can look at $6,788.88 as a capital cost with a payback period. Work out how many months it takes to earn the premium back. If that number is small, the price stops being outrageous and starts being a line item.

The same applies to anyone whose current hardware is the actual bottleneck in something they get paid for. Waiting has a cost too, and if the waiting cost exceeds the premium, buying is rational even at a silly price.

Everybody else is in a different conversation. Wanting maximum frame rates is a perfectly fine reason to want a card. It just isn’t a reason that survives contact with a 240 percent markup.

Headroom is the expensive thing you are really buying

Most enthusiasts aren’t buying performance they need. They’re buying performance they might need later, and that’s a much weaker case at this price.

Think about what the top card in a generation actually gives you over the tier below it. Higher settings you may not notice in motion. Headroom for games that haven’t shipped yet. Comfort that you won’t feel behind in two years. All real, all nice, none of them worth several thousand dollars in cash today.

And here’s the awkward bit about future-proofing. By the time those demanding games arrive, there will be newer cards, possibly at saner prices. Paying a shortage premium to be ready early often means paying the most for hardware that will look ordinary by the time you use it properly.

The price history should make you hesitate

Tracking since 10 June 2026 shows a low of $4,199.99 on 30 July 2026 and a high of $6,810.90 on 20 August 2026. That’s a swing of more than $2,600 in roughly three weeks.

Sit with that. The card’s price moved by more than its entire official MSRP inside a month. And the current $6,788.88 sits right up near the top of that range, not the bottom.

Anyone who bought on 30 July got the card for about $2,600 less than someone buying in late August. Same card. Same month, nearly. That’s not a market where timing is a minor detail, it’s a market where timing is most of the outcome.

What the used market gives you, and what it costs you

Used cards sell around 26 percent below new prices. That’s a genuine saving in absolute terms, and at these levels it’s a large number of dollars. It’s also the only meaningful discount available.

But look at what you give up. No manufacturer warranty in most cases, or a fraction of one. No recourse if the card fails in month three. And crucially, no way to know what the card has been doing. The same AI demand that inflated prices means plenty of cards have run flat out, continuously, for long stretches. That’s a different life from occasional evenings of gaming.

Fans wear. Thermal paste dries out. Solder joints go through thousands of heat cycles. None of this is guaranteed to bite you, and plenty of used cards run fine for years. But you’re making a several-thousand-dollar bet on someone else’s honesty about how the card was used, usually with no way to verify it.

If you do buy used, buy through a platform with real buyer protection, insist on seeing the card running, and treat any seller who won’t answer questions about its history as a seller with something to hide.

Waiting versus buying, without the wishful thinking

The case for waiting is straightforward. Prices at the top of a tracked range tend not to stay at the top forever. The pressure driving this comes from AI industry demand for the same silicon, and demand like that can shift when business conditions change.

The case against waiting is equally straightforward. Nobody knows when it eases. It could take a very long time. Meanwhile you’re using the machine you have, and if that machine is genuinely holding you back, months of waiting is a real cost.

So the sensible middle ground is to decide in advance what would make you buy an RTX 5090 and what wouldn’t. Pick a price you’d be comfortable paying and set an alert. If it hits, buy without agonising. If it doesn’t, you’ve lost nothing except the cards you never bought at a price you’d have regretted. Making that decision when you’re calm beats making it while staring at a listing.

Never buy at the top of a shortage

This is the general principle, and it holds well beyond graphics cards. When a product’s price is set by scarcity rather than by cost, the peak is always the worst moment to commit.

The reason is simple. At the peak, everyone who’s willing to pay a lot already has. The buyers still in the market are the ones who held out, which means the seller’s advantage is fading, not growing. Prices set by panic tend to come down messily and without warning, and the people holding hardware they bought at the top are the ones who feel it.

If you buy an RTX 5090 at $6,788.88 and it settles back towards the July low, you’ve lost more than $2,500 in value on a single component. That’s not a small mistake. That’s an entire decent computer.

The alternatives nobody wants to hear

A previous-generation flagship still plays everything. A current mid-range card will handle most workloads most people actually run, and at a fraction of the outlay. Neither of those is exciting, and both are considerably more sensible than what the top of the stack costs today.

There’s also the option of skipping the decision to buy an RTX 5090 altogether and upgrading something else. More memory, faster storage, a better monitor. These often do more for how a machine feels day to day than a faster GPU does, and they don’t require competing with data centres for supply.

I’d also gently suggest that the pressure to own the fastest card is partly manufactured. It’s a shortage, and shortages make people act quickly. Slowing down is usually the winning move.

Common questions

Is it ever smart to buy an RTX 5090 at these prices?

Yes, if it pays for itself. Professional work with a clear hourly value attached can justify almost any hardware price as long as the payback period is short. If the card is for entertainment, the same numbers don’t work.

How much cheaper is the used market?

Around 26 percent below new prices. At current levels that’s a significant sum, but it still leaves you well above the $1,999 MSRP, and you take on the risk of a card that may have been run hard.

Will waiting actually save money?

Possibly, though nobody can promise it. The tracked low was $4,199.99 on 30 July 2026 and the high was $6,810.90 on 20 August 2026, so the price clearly moves. Whether it moves back down depends on demand that has nothing to do with gamers.

What should I check before buying a used card?

Ask what the card was used for and for how long, see it running under load if you possibly can, buy through a platform with buyer protection, and confirm whether any warranty transfers. Vague answers are a reason to walk away.

The decision to buy an RTX 5090 comes down to whether the card makes you money or just makes you happy. One of those justifies a 240 percent premium, and the other really doesn’t.