• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

GPU prices — is the worst behind us ?

I was looking on Newegg last night and you could buy my ASRock Intel Arc B 570 10GB card with Battlefield 6 for free at $209 or the B 580 for $249, seems from the reviews on Newegg that Intel has made some happy customers, they seemed to have ramped up supply and pulled prices back down to MSRP.
 

"GPU Price Hikes Coming in 2026, Warns PowerColor

by AleksandarK Yesterday, 10:18 Discuss (29 Comments)
A series of shocking news about price hikes continues to unfold, as we now receive official "warnings" from GPU AIB partners about the situation. On the official PowerColor subreddit, a representative noted that Holiday 2025 discounts are about to commence, but also claimed that prices will rise in 2026, urging gamers to buy before the increase. While this could be a marketing tactic, the recurring theme of warnings is becoming more prominent in the supply chain. Previously, we reported that DRAM prices have risen significantly—171.8% year-over-year—driving costs to new levels for this essential component found in every GPU, smartphone, PC, and console.

South Korean memory giants, such as Samsung and SK Hynix, are currently unable to fulfill all orders, completing only 70% of them. Smaller OEMs and channel distributors have been informed to expect only 35-40% fulfillment through the first quarter of 2026. Even some larger AIBs, like PowerColor, may face shortages in DRAM supply, potentially leaving their GPUs without installed DRAM modules. This situation not only delays their planned product rollouts but also jeopardizes their expected revenue streams if it persists. They are faced with the choice of either gambling on the spot market for a significant markup or leaving their production lines idle. Additionally, the shortage could ultimately increase the price of GPUs that are shipped, as they now contain a rarer commodity—DRAM."
 
All of this feels like a concerted effort to push people towards game streaming. You'll own nothing and like it. Looking forward to 600 dollar 6060s.
It's more just the AI monster eating everything in its path. A bunch of companies are building datacenters as fast as they can to have the biggest, bestest LLM they can. It's not lie it is easy to scale up production of things like DRAM once you hit the limits of the factory, so prices are going up and the companies can pay because they have massive amounts of investor funds to waste.

In the long run who knows what happens? Maybe LLMs will advance to the point where they just replace almost all workers (this is what the companies are betting on and about the only way this investment ever makes sense) and then, well, GPU prices probably won't matter because you won't have a job to afford one anyhow. Or maybe this will all come crashing down when the economy enters a recession and investors get spooked and pull their money. At that point the AI companies have to figure out how to be profitable, which so far they haven't, and as part of that they'll have to start buying drastically less hardware which should reduce the prices.
 
The idea that this trillions worldwide compute built up and worldwide ram supply is a concerted effort to push people toward game streaming (as if they do not love people paying a fortune, all of it upfront and the electricity to run it if they want), feel like looking hard to search it.

so prices are going up and the companies can pay because they have massive amounts of investor funds to waste.
Lot of the buyer are companies that can pay because of the massive amount of cash flow as well, they are not diluting themselve for this, they just slowing buyback a little bit

At that point the AI companies have to figure out how to be profitable, which so far they haven't,
Some have been doing it for a long time by now, spending are growing but their revenues exploding

https://ca.finance.yahoo.com/news/googles-corporate-parent-posts-first-204206717.html
Alphabet performed like a powerhouse during the July-September period, delivering a profit of nearly $35 billion, or $2.87 per share, a 33% increase from the same time last year. Revenue rose 16% from last year to $102.3 billion. Both figures easily exceeded the analysts' projections that steer the stock market.

https://ca.finance.yahoo.com/news/meta-q3-earnings-stunted-nearly-202442769.html
Meta, the parent of Facebook, Instagram and more, saw its top-line revenue jump 26% to a whopping $51.2 billion in the third quarter — a new quarterly record. But a provision of President Trump’s One Big Beautiful Bill Act led to a huge tax charge, causing the internet giant to fall well short of Wall Street expectations.


A lot of that feeling that AI is not generating profit (and ahead of schedule even optimistic people a couple of years ago regarding the latest harder to run transformer models) to anyone but hardware sellers, is because the first biggest profitable way is one that was already working very well and been there for a long time, Machine learned suggestion and ads, AI improving content recommendation and ad targeting, the google, amazon, Meta is not saying it that much outloud that for them AI is working better and better and printing more and more money, not something to be particularly proud of. If AI does more than that for them (for amazon and google already do and will more and more) the build up for it would have been paid by better ads business and mind capture during it.
 
I mean, I'm not saying the bigger firms like Google or MS aren't profitable, they are, I'm saying AI isn't profitable. For all available data it is a money furnace right now. OpenAI went through like $11 billion last quarter. At some point, that party comes to an end and they have to either increase prices, decrease outlays, or both.
 
I mean, I'm not saying the bigger firms like Google or MS aren't profitable, they are, I'm saying AI isn't profitable. For all available data it is a money furnace right now. OpenAI went through like $11 billion last quarter. At some point, that party comes to an end and they have to either increase prices, decrease outlays, or both.
It’s the typical startup model these days, hyper scale or whatever then try to figure out how to be profitable.
 
It’s the typical startup model these days, hyper scale or whatever then try to figure out how to be profitable.
And I'm not saying they won't (though I do have some skepticism) just that they will have to cut spending on that path. Unless people are willing to pay a LOOOOOOT more for LLMs than the current prices, they are going to have to cut expenditures. When that happens, stress on the computer supply chain will relax.
 
I mean, I'm not saying the bigger firms like Google or MS aren't profitable, they are, I'm saying AI isn't profitable. For all available data it is a money furnace right now. OpenAI went through like $11 billion last quarter. At some point, that party comes to an end and they have to either increase prices, decrease outlays, or both.
google being profitable but not AI can become a bit messy, pretty much everything that bring revenue that they do is powered by it and there recent explosion in revenues (at least they claim) is due to AI getting better at capturing attention and matching sellers with buyer with it.

And I'm not saying they won't (though I do have some skepticism) just that they will have to cut spending on that path. Unless people are willing to pay a LOOOOOOT more for LLMs than the current prices, they are going to have to cut expenditures
GPT 5 was a lot about cutting cost, we do not know exactly marginal cost per token of course, but the consensus is that they are being sold at a way higher price then marginal compute for inference cost already, the money burning is from the giant training phase/R&D now.

Google cloud being profitable became a thing after they started selling tokens in the spring of 2023, it is growing fast and that despite a lot of free user base cost. (https://www.constellationr.com/blog...ue surged 34,operating income of $3.6 billion.)

I feel some of it are 12months+ old prediction that are repeated and did not update itself with Meta/Google surprising results since, even waymo the profitability is getting in sight
 
It's more just the AI monster eating everything in its path. A bunch of companies are building datacenters as fast as they can to have the biggest, bestest LLM they can. It's not lie it is easy to scale up production of things like DRAM once you hit the limits of the factory, so prices are going up and the companies can pay because they have massive amounts of investor funds to waste.

In the long run who knows what happens? Maybe LLMs will advance to the point where they just replace almost all workers (this is what the companies are betting on and about the only way this investment ever makes sense) and then, well, GPU prices probably won't matter because you won't have a job to afford one anyhow. Or maybe this will all come crashing down when the economy enters a recession and investors get spooked and pull their money. At that point the AI companies have to figure out how to be profitable, which so far they haven't, and as part of that they'll have to start buying drastically less hardware which should reduce the prices.
Hey man if i can't be a game dev or software dev any more i can easily work in sales or go back to serving tables in fine dining. I love making games but it's insanely stressful. I use Gemini 3 and GPT 5.1 every day. So far it's a productivity boost, but if it takes over it wont hurt me too too much. The other jobs sound far less stressful at this point. The odd thing is the entire economy will collapse and die if they're successful. The whole thing feels short sighted, but so is our economy.
 
It's more just the AI monster eating everything in its path. A bunch of companies are building datacenters as fast as they can to have the biggest, bestest LLM they can. It's not lie it is easy to scale up production of things like DRAM once you hit the limits of the factory, so prices are going up and the companies can pay because they have massive amounts of investor funds to waste.

In the long run who knows what happens? Maybe LLMs will advance to the point where they just replace almost all workers (this is what the companies are betting on and about the only way this investment ever makes sense) and then, well, GPU prices probably won't matter because you won't have a job to afford one anyhow. Or maybe this will all come crashing down when the economy enters a recession and investors get spooked and pull their money. At that point the AI companies have to figure out how to be profitable, which so far they haven't, and as part of that they'll have to start buying drastically less hardware which should reduce the prices.
Chip makers are the winners in all of this. When there is a gold rush you don't pan for gold to not miss out - you sell shovels and pans.
 
When I started this thread 4 years back, I anticipated something like the 9060xt 16gb to be affordable, right around this time (before a potential refresh with Super cards)

It is. Has finally dropped below MSRP. for now.

But immediate future looks bleak.

"AMD Prepares 10% GPU Price Hike Amid Memory Shortage

by AleksandarK Today, 04:11 Discuss (32 Comments)
AMD has reportedly notified its supply chain that it will increase pricing across its entire GPU product line by about 10%. The main driver behind this is the massive memory shortage driving DRAM prices to skyrocket and eating up all the margins for GPU makers. Apparently, AMD's AIB partners have been notified that AMD plans to implement a 10% price increase across the board, which AIBs will likely match in the coming weeks to preserve their margins. GPU makers like AMD (and NVIDIA/Intel) bundle their GPU dies with GDDR memory and send that to AIBs as a kit. In situations like this, when AMD raises GPU prices, it is due to the cost of the DRAM bundled with the GPU, as the costs of AMD's raw silicon remain stable for now.

Additionally, this is AMD's second recent GPU price increase. The company did some price restructuring in October, but it seems that the memory supply chain is now starting to show signs of weakness and that a full 10% increase is needed. We have reported that AMD's long-standing AIB partner, PowerColor, is now urging gamers to purchase GPUs during this holiday season, and that 2026 is when we will see this price hike. It could be that AIBs have locked their pricing mechanisms for now with the old AMD pricing, and that 2026 will be implementing the new pricing strategy. This is also happening just as AMD's Radeon RX 9070 and Radeon RX 9070 XT graphics cards have finally reached their MSRPs of $549 and $599, respectively."
 
I'm seriously considering a 9070 or xt though I should probably just get a 9060xt 16gb.
9060xt will last a long time

9070 / 9070xt will get replaced in 2 years time by the 10070

The 10060xt is expected to have lpddr5x/lpddr6 memory. So there could be regression in some scenarios vs the 9060xt, I think
 
Interesting, I hadn't heard that. I should go with the 9060xt as I doubt I really need the 9070 given im running at 1080p, I just dislike buying a 128bit card and it is supposedly only 20% faster than a 6700xt. In theory I should alsosell my 6700xt. Price/perf the 60 vs 70 seems on par with each other.
 
9070 / 9070xt will get replaced in 2 years time by the 10070
Not saying that not the timeline, but if is mid-end of 2027 for the 10070 to came out, AMD will have went a full 5 years without clearly and cleanly beat the AD103, a 379mm-TSMC 5-256 bits of 2022.... that many called a 4070 in disguise.

I would imagine them trying to launch something Q4 2026, treating the RDNA 4 launch as a fast one and keeping the expected cadence of RDNA 2 in 2020, RDNA 3 in 2022, RDNA should have been Q4 2024 and now 2026. Same for Nvidia, skip the super refresh and go on the 6000 series in 2026.

I just dislike buying a 128bit card and it is supposedly only 20% faster than a 6700xt.
25% according to TPU bench if it make things feel a little bit better, they have the 6700xt at 80% of an 9600xt16GB (https://www.techpowerup.com/gpu-specs/radeon-rx-6700-xt.c3695)
 
Last edited:
I'm very glad I purchased a 5070 from a forum member when I did. GPU prices are going to get crazy over the next few months.
 
So price speculation here... do you think used cards are going to see higher prices similar to what we saw in the past? Where 1 or 2 gen old cards were selling for original msrp?
 
Just got my XFX Mercury AMD Radeon RX 9070XT OC Gaming Ultra Bada$$ DeluXxX Edition in the mail. The over sized cooler doesn't disappoint.
 
Im glad I got my XFX 7900 XTX, should be good for a while now. Doubt I will run into any problem with games for the next several years. Plus by then Ill be due for a whole system upgrade anyway.
 
Igor says, if you are planning on buying the 9070xt, then buy it immediately

Will Radeon RX 9070 XT Soon Cost 200 Euros More? Insights into Material and Manufacturing Costs, Cost Calculations, and Manufacturer Margins​

22. May 2026 05:30
Igor Wallossek

The RX 9070 XT is currently less a normal product in price comparisons than an indicator of a delayed cost wave. The visible price apparently does not yet reflect the current FOB pressure everywhere. That is precisely why the card still appears comparatively attractive in May 2026, even though it is objectively not cheap. The real message is therefore not that graphics cards have become more expensive. The real message is that part of the price increase may not yet have fully reached the shelf.

https://www.igorslab.de/en/mid-range-gpus-2026-material-manufacturing-costs-margins-trends/
 
Im glad I got my XFX 7900 XTX, should be good for a while now. Doubt I will run into any problem with games for the next several years. Plus by then Ill be due for a whole system upgrade anyway.
In the next several years, there's probably gonna be one GPU from nVidia, one from Intel, one from AMD and you get to decide which subscription you want depending on how much power you want to unlock
 
Igor says, if you are planning on buying the 9070xt, then buy it immediately

Will Radeon RX 9070 XT Soon Cost 200 Euros More? Insights into Material and Manufacturing Costs, Cost Calculations, and Manufacturer Margins​

22. May 2026 05:30
Igor Wallossek

The RX 9070 XT is currently less a normal product in price comparisons than an indicator of a delayed cost wave. The visible price apparently does not yet reflect the current FOB pressure everywhere. That is precisely why the card still appears comparatively attractive in May 2026, even though it is objectively not cheap. The real message is therefore not that graphics cards have become more expensive. The real message is that part of the price increase may not yet have fully reached the shelf.

https://www.igorslab.de/en/mid-range-gpus-2026-material-manufacturing-costs-margins-trends/
This is exactly why I snagged my 9070xt back in early February; just wanted to get in while I could still stomach the cost. Knew the price increase was coming, but thought it'd take effect a bit sooner then what's reported above.
 
So far, there's no indication of stopping the eventual building of the AI datacenters, even though funding has slowed up. Even if AI isn't profitable, they're just going to rent out the hardware as cloud/compute. The best case scenario is somehow they just resell the hardware to consumers, but that'll be at least 2-3 years down the road. By then, newly released graphics cards costing $5,000 will be the norm, so you'll be paying $2,500 for a used four-year-old RTX 5070 Ti.
 
isn't H2 2026 cap-ex a new record high and by a lot ? Not sure anything has slowed yet.
Isn't nVidia including gaming sales as part of this "revenue" stream? That tells me they're doing everythign they can to artificially inflate the numbers
 
Isn't nVidia including gaming sales as part of this "revenue" stream? That tells me they're doing everythign they can to artificially inflate the numbers
Data center is one group, everything else falls into a group bucket.
 
So far, there's no indication of stopping the eventual building of the AI datacenters, even though funding has slowed up. Even if AI isn't profitable, they're just going to rent out the hardware as cloud/compute. The best case scenario is somehow they just resell the hardware to consumers, but that'll be at least 2-3 years down the road. By then, newly released graphics cards costing $5,000 will be the norm, so you'll be paying $2,500 for a used four-year-old RTX 5070 Ti.
my guess is the party ends in 2028

GPU prices return back to normal in H2 2028

An AI consultant tells Axios that one of their clients recently spent half a billion dollars in a single month after failing to put usage limits on Claude licenses for employees.


https://x.com/edzitron/status/2060006709557690841?s=20
 
Isn't nVidia including gaming sales as part of this "revenue" stream? That tells me they're doing everythign they can to artificially inflate the numbers
I think you are the first one to suspect Nvidia of using secret gaming revenues to boost their datacenter number.

But no data center is its own revenues segment and they separate it in 2 bucket, hyperscaler and the neo clouds, industrial-enterprise (i.e. datacenter hardware but not bought by the giant microsoft-meta-google-amazon....), hyperscaler Q1 2027 (Fiscal) was more than double Q1 2026.

datacenter 62,314 was 21.6% bigger than the previous quarter (21.6% a year growth would be excellent....)

Pace seem to look like this :
image-4.png


Did not saw/heard any sign of spending today being lower than ever before.

my guess is the party ends in 2028
Amount of compute per robots and how many of thoses there will be make this an very hard guess game, 2 intelligent robots per humans seem reasonable, the compute to train that fleet and to operate it being about 2x5080 per robots has well, thats 30 billions high end GPU worth of compute in the next 15 years just there.
 
Is the worse behind us? Absolutely not; that is yet to come I'm afraid. I wouldn't be surprise if this is part of the plan to kill off the component market. Cloud gaming, where you will own nothing.. or pre-built OEM PC's ... but no DIY... that is the future landscape
 
Back
Top