Nvidia has told major customers that prices on servers packed with its artificial intelligence chips are going up more than 15% in many cases, and the pain has already spilled into the consumer graphics card market. Newegg data tracked through late August 2026 shows RTX 50-series cards climbing as much as 39% since June, with the RTX 5070 alone jumping 36% in two months. AMD is following with roughly 10% increases of its own on select Radeon GPUs. For anyone shopping for a graphics card, or running an AI workload that depends on Nvidia silicon, the math on a 2026 upgrade just changed.
The trigger is not a chip shortage in the traditional sense. It’s a memory shortage. HBM4e and GDDR7 supply is being consumed by the data center build-out at a pace memory makers can’t match, and that scarcity is now working its way down through every product tier Nvidia and AMD sell, from $300 entry cards to $16,000 workstation GPUs to AI server racks worth tens of millions of dollars.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
What Happened: Nvidia’s August 2026 Price Notice
Bloomberg first reported, and Fortune and CNBC subsequently confirmed, that Nvidia notified some of its biggest customers in the week of August 17-22, 2026 that server systems built around its AI chips would cost more than 15% more in many configurations. The increase applies to systems shipping starting in early 2027, which means orders being placed right now, for delivery months out, are already being repriced upward before a single unit ships.
The increase touches Nvidia’s current and next-generation data center lineups alike. Coverage from The GPU newsletter and other trade outlets indicates the price bump applies “across the board,” including established server GPUs like the H200 and B200, not just the newer Vera Rubin and Grace Blackwell platforms still ramping toward volume shipments. Nvidia has not published a per-SKU breakdown, and the company has not issued a public statement detailing exact percentages by chip generation, which is typical for enterprise pricing communicated directly to hyperscaler and OEM accounts rather than through a press release.
What makes this notice different from routine enterprise price adjustments is the timing. It landed in the same month that retail GPU pricing trackers started showing double-digit jumps on consumer cards, suggesting the memory cost pressure driving the enterprise hike is not contained to the data center. It’s a single supply problem showing up in two markets at once.
RTX 50-Series Retail Prices Jump Up to 39%
Tom’s Hardware tracked median RTX 50-series listings on Newegg between June and August 2026 and found the increases are neither small nor uniform. The RTX 5060 Ti 16GB posted the steepest jump in the lineup at 39%, moving from a $569.99 median to $804.99. The RTX 5070 wasn’t far behind at a 36% increase, moving from $659.99 to $899.99. The RTX 5060 rose 27%, from $369.99 to $469.99.
Higher up the stack, the increases are smaller in percentage terms but still real in dollar terms. The RTX 5090 climbed 9%, from a $4,299.99 median to $4,699.99, and the RTX 5080 rose about 3%, from $1,461.99 to $1,499.99. The RTX 5070 Ti was the one card in the lineup that held flat at $1,099.99 across the two-month window, which Tom’s Hardware notes points to selective repricing rather than a blanket surcharge applied evenly across every SKU. Even the budget-tier RTX 5050 wasn’t spared, ticking up 5% from $299.99 to $314.99.
The pattern that emerges is that the mid-range, the tier most PC gamers actually buy, absorbed the biggest percentage hikes. That’s a notable shift from the 2020-2021 GPU shortage, when scalping and crypto-mining demand hit high-end cards hardest. This time, a 16GB card aimed at 1440p gamers got hit nearly as hard as a flagship.
| GPU Model | Median Price, June 2026 | Median Price, August 2026 | Change |
|---|---|---|---|
| RTX 5050 | $299.99 | $314.99 | +5% |
| RTX 5060 | $369.99 | $469.99 | +27% |
| RTX 5060 Ti 16GB | $569.99 | $804.99 | +39% |
| RTX 5070 | $659.99 | $899.99 | +36% |
| RTX 5070 Ti | $1,099.99 | $1,099.99 | 0% |
| RTX 5080 | $1,461.99 | $1,499.99 | +3% |
| RTX 5090 | $4,299.99 | $4,699.99 | +9% |
Source: Newegg median listing prices, tracked and reported by Tom’s Hardware, June-August 2026.
Why Memory Costs Are Driving the Nvidia GPU Price Hike
Every account of the August 2026 price hikes, from Bloomberg’s original report to Fortune’s and CNBC’s follow-ups, points to the same root cause: memory chip costs, not GPU die production. Nvidia’s own reporting to customers said the increases are “driven primarily by soaring memory chip costs,” with the size of the hike varying by chip generation and memory configuration rather than by GPU compute performance alone.
That distinction matters because it explains why a two-year-old architecture and a brand-new one are both getting more expensive at the same time. HBM4e, the high-bandwidth memory feeding Nvidia’s data center accelerators, and GDDR7, the memory in consumer RTX 50-series cards, are both manufactured by a small group of suppliers (SK Hynix, Samsung, and Micron) whose capacity is increasingly reserved years in advance by AI infrastructure buyers. When memory demand outstrips supply, it doesn’t matter whether the GPU attached to that memory launched in 2024 or 2026: the bill of materials goes up either way.
Analyst estimates on just how expensive next-generation AI racks are becoming vary widely but agree on direction. Wccftech has cited analyst models from BofA Global Research and Morgan Stanley putting a fully configured Rubin Ultra rack near $21 million, with HBM4e memory alone accounting for roughly $1.5 million of that total. A separate Bernstein estimate put a Vera Rubin rack closer to $9.1 million, with the gap between estimates itself a signal of how unsettled current AI hardware pricing really is heading into 2027 shipments.
AMD Follows With Its Own Radeon Price Increase
Nvidia isn’t repricing in isolation. AMD is preparing roughly 10% price increases on select Radeon GPUs and GDDR6 memory kits, effective in August 2026. The increase was originally slated for July before AMD pushed it back a month, and the company has attributed the move to the same AI-driven VRAM cost surge Nvidia cited, rather than to any change in its own manufacturing costs or margins.
That AMD is moving in lockstep with Nvidia on timing, even though the two companies compete directly for the same GPU buyers, is itself informative. It suggests the memory cost pressure sits upstream of both companies, at the supplier level, rather than reflecting a pricing decision either GPU maker had much room to avoid. When two rival chipmakers raise prices in the same month for the same stated reason, that’s a supply chain signal, not a coincidence of corporate strategy.
For AMD, the increase adds a wrinkle to a value pitch that has leaned heavily on undercutting Nvidia at every price tier. Cards like the Radeon RX 9070 XT built their retail momentum on being priced meaningfully below Nvidia’s closest competing SKU. A 10% cost bump narrows that gap unless AMD chooses to absorb some of the increase itself to protect its price-per-frame positioning against Nvidia’s RTX 50-series.
Data Center Chips Get More Expensive Too
The consumer GPU story is the one most readers will notice at checkout, but the bigger dollar figures are happening in the data center. Nvidia’s professional and workstation GPUs are already showing the increase in list pricing. The RTX PRO 6000 Blackwell Workstation Edition is now listed at $16,000 on Nvidia’s US marketplace, and as of mid-August 2026 it was showing as out of stock even at that price. That $16,000 listing is roughly 20.8% higher than the card’s previous $13,250 price point, and somewhere between 86% and 110% above the $7,600-$8,565 range reported around its original launch or preorder window.
Server GPUs further down the product line, including the H200 and B200 that already ship in volume to cloud providers, are reportedly seeing increases of similar magnitude to the headline 15% figure, according to coverage from The GPU newsletter’s daily briefing. That means the price increase isn’t confined to unreleased or newly launched hardware still commanding early-adopter premiums. It’s hitting chips that have been shipping and generating revenue for Nvidia for well over a year.
| Product | Prior Price | August 2026 Price | Change |
|---|---|---|---|
| RTX PRO 6000 Blackwell Workstation | $13,250 | $16,000 | +20.8% |
| RTX PRO 6000 Blackwell (vs. launch/preorder) | $7,600-$8,565 | $16,000 | +86% to +110% |
| AI server systems (Vera Rubin, Grace Blackwell) | Baseline | +15%+ | 15%+ across configurations |
| H200 / B200 server GPUs | Baseline | Reported increase “across the board” | 15%+ (unconfirmed exact %) |
| Select AMD Radeon GPUs + GDDR6 kits | Baseline | +10% (Aug 2026) | +10% |
Sources: Tom’s Hardware, The Verge, KuCoin News.
How This Compares to Past GPU Price Shocks
PC hardware buyers have lived through GPU price spikes before, but the mechanics behind this one are different from the two most recent shocks. The 2020-2021 shortage was demand-side: crypto miners and pandemic-era PC upgraders competed for a fixed supply of finished cards, and scalpers and bots amplified the squeeze at retail. Wafer and chip production wasn’t the constraint; getting a card into a shopping cart before a bot did was.
The 2025 memory crunch that preceded this year’s increases was closer to a preview of today’s problem: DRAM and NAND prices climbed as AI training clusters absorbed a growing share of global memory output, squeezing consumer PC and phone makers who compete for the same fabs. What’s happening in August 2026 is that squeeze reaching full maturity. Memory suppliers are now allocating HBM and high-end GDDR capacity years in advance to AI infrastructure customers willing to pay premium, guaranteed-volume pricing, leaving GPU makers with less room to absorb cost increases on the consumer side without either raising prices or shrinking margins.
The practical difference for buyers: 2020-2021 pricing was chaotic and often temporary, snapping back once mining demand collapsed and supply caught up. The current increase is structural, tied to a years-long memory capacity buildout that isn’t going to resolve with a single good manufacturing quarter.
Market Impact: Gamers, AI Startups, and Cloud Providers
The impact splits into three distinct groups, each absorbing the increase differently. PC gamers shopping for a new RTX 50-series card in the $400-$900 range are the ones seeing the sharpest percentage jumps, and for a segment of buyers already price-sensitive after 2025’s console and component price increases, a 27-39% jump on mid-range cards is enough to push an upgrade decision into 2027.
AI startups and smaller cloud providers renting or buying GPU capacity face a different problem: budgets set earlier in 2026 for infrastructure scheduled to arrive in early 2027 are now short of what they’ll actually need to pay. Since the increase applies to systems shipping starting next year, any company that locked in a roadmap based on current list pricing is now negotiating a gap, either absorbing the cost themselves, passing it to their own customers, or delaying planned capacity expansion.
Hyperscalers, meanwhile, have more leverage to negotiate around list pricing and typically buy on multi-year contracts that partially insulate them from spot increases, but even at that scale, a 15%+ increase on server systems compounds quickly across orders sized in the tens of thousands of GPUs. Analysts covering the space have started framing the increase as a cost that will ultimately be passed through to AI API pricing and cloud GPU rental rates, rather than absorbed indefinitely by any single layer of the stack.
Competitive Landscape: Nvidia, AMD, and Intel’s Pricing Strategies
Nvidia still holds the dominant position in both the consumer discrete GPU market and the AI accelerator market, which gives it more room to raise prices without immediately ceding share. Buyers who want the best available performance per watt in either category have few alternatives that match Nvidia’s software ecosystem, spanning CUDA, DLSS, and its AI framework integrations. That’s a large part of why Nvidia can raise prices across an entire generation rather than risk losing sales only on the SKUs where competition is fiercest.
AMD’s matching 10% increase suggests it isn’t trying to use this moment to undercut Nvidia and grab share through aggressive pricing, likely because AMD faces the identical memory cost pressure and doesn’t have the margin cushion to eat the increase while Nvidia raises prices around it. Intel, meanwhile, is taking a different tack by pushing into adjacent hardware categories rather than competing head-on in high-end discrete GPUs. Intel’s Arc G3 Extreme chip, which combines a 14-core CPU with an Arc B390 GPU featuring 12 Xe3 cores, went through handheld gaming testing in mid-August 2026 and now powers devices like the MSI Claw 8 EX AI+, a handheld launching around €1,649 with up to 32GB of LPDDR5x memory. That’s a bet on winning a hardware category, gaming handhelds, where Nvidia doesn’t currently compete directly, rather than trying to underprice Nvidia and AMD on desktop GPUs during a memory-constrained market.
What This Means for PC Builders Right Now
For anyone actively planning a build, the practical guidance splits by tier. At the high end, the RTX 5090 and RTX 5080 have seen the smallest percentage increases of the lineup, 9% and roughly 3% respectively, which for buyers already committed to flagship-tier spending changes the math less than it does for mid-range shoppers. The RTX 5070 Ti’s flat pricing through August also makes it, for the moment, a relative outlier worth watching before the next pricing update.
The mid-range is where the calculus has shifted hardest. A 36-39% increase on the RTX 5070 and RTX 5060 Ti 16GB effectively erases the value proposition that made those cards popular upgrade targets earlier in 2026. Buyers who were planning to wait for a price drop are instead watching prices move the opposite direction, and with AMD’s Radeon lineup also getting more expensive in the same window, there’s no clear escape hatch to a cheaper competing brand this month.
Retailer-tracked pricing also tends to lag official price list changes by a few weeks, meaning further movement, in either direction, is possible before September inventory cycles settle. Buyers with flexibility on timing may want to watch weekly pricing rather than assume August’s numbers are the final word for the quarter.
Predictions: Where GPU Pricing Goes From Here
- Memory costs stay elevated through 2027. With HBM4e and GDDR7 capacity already allocated years ahead to AI infrastructure buyers, consumer GPU pricing is unlikely to meaningfully retreat before new memory fab capacity comes fully online, which analysts generally place in the 2027-2028 window.
- AI server pricing keeps climbing into 2027 shipments. Since Nvidia’s 15%+ increase applies to systems shipping starting early 2027, expect further repricing notices as memory costs and Vera Rubin ramp timelines firm up over the next two quarters.
- Mid-range consumer GPUs remain the most volatile tier. The RTX 5070 and RTX 5060 Ti 16GB saw the sharpest percentage swings this cycle; that volatility is likely to continue as retailers adjust inventory against shifting supplier costs faster than they do at the ultra-high or entry-level tiers.
- AMD holds its relative discount but narrows it. Expect AMD to keep Radeon pricing below Nvidia’s closest competing SKUs to protect its value positioning, even as its own 10% increase closes part of that gap through the rest of 2026.
- Cloud GPU rental rates rise in step. As hardware acquisition costs climb for cloud providers and GPU rental services, expect rental and API pricing for AI compute to reflect a portion of the increase over the coming quarters, rather than providers absorbing it indefinitely.
The Bottom Line for Buyers and Builders
What’s playing out in August 2026 is a single memory supply problem showing up in three different markets at once: consumer GPUs, professional workstation cards, and AI data center systems. The scale of the increases, from a 3% bump on the RTX 5080 to a 110% jump on RTX PRO 6000 pricing versus its original launch, shows how unevenly the cost pressure is landing depending on where a product sits in the memory-intensity spectrum. Cards and chips that lean hardest on the scarcest memory types are absorbing the biggest increases, and that pattern is likely to persist as long as AI infrastructure demand keeps outpacing memory fab capacity.
For most buyers, that means treating 2026 GPU pricing as a moving target rather than a fixed number, and factoring memory market news, not just new product launches, into the timing of the next PC build or AI infrastructure order.
Frequently Asked Questions
Why are Nvidia GPU prices going up in 2026?
Nvidia has attributed the increase primarily to rising memory chip costs, specifically HBM4e for data center chips and GDDR7 for consumer RTX 50-series cards, as AI infrastructure demand consumes a growing share of available memory supply from manufacturers like SK Hynix, Samsung, and Micron.
How much have RTX 50-series GPU prices increased?
Newegg’s median listing prices show increases ranging from 3% on the RTX 5080 up to 39% on the RTX 5060 Ti 16GB between June and August 2026. The RTX 5070 rose 36% and the RTX 5060 rose 27% over the same window, while the RTX 5070 Ti held flat.
Is AMD also raising GPU prices in 2026?
Yes. AMD is implementing roughly 10% price increases on select Radeon GPUs and GDDR6 memory kits in August 2026, citing the same AI-driven memory cost surge that Nvidia referenced for its own increases.
When do Nvidia’s AI server price increases take effect?
The reported 15%+ increase on servers built around Nvidia’s AI chips applies to systems shipping starting in early 2027. Orders placed now for future delivery are already being repriced to reflect the higher costs.
How does this compare to the 2020-2021 GPU shortage?
The 2020-2021 shortage was driven mainly by demand from crypto miners and pandemic-era buyers competing for a limited supply of finished cards. The 2026 increase is a supply-side memory cost problem tied to AI infrastructure absorbing HBM and GDDR capacity, which makes it more structural and less likely to reverse quickly once it eases.
Should I buy a GPU now or wait for prices to drop?
Because the increases are tied to years-long memory supply commitments rather than a temporary shortage, analysts don’t expect a quick reversal. Buyers with flexible timing may want to track weekly retailer pricing, since some tiers, like the RTX 5070 Ti, have stayed flat even as others jumped sharply.
What is the RTX PRO 6000 Blackwell now priced at?
Nvidia’s RTX PRO 6000 Blackwell Workstation Edition is listed at $16,000 on Nvidia’s US marketplace as of mid-August 2026, up 20.8% from its prior $13,250 price and as much as 86-110% above its original launch or preorder pricing.
Will AI cloud and API pricing go up because of this?
Analysts tracking the AI infrastructure market expect at least part of the increase in GPU hardware costs to be reflected in cloud GPU rental rates and AI API pricing over the coming quarters, as providers work higher acquisition costs into their own pricing structures.
Related Coverage
- Nvidia Reportedly Buys Hugging Face for $12.9B [2026]
- NVIDIA Nemotron 3.5: Free 30B Model, $89B GPU Bet [2026]
- RTX 4090 vs RTX 5080: 15% RT Gap, DLSS 5 Snub [2026]
- RX 9070 XT vs RTX 5070 Ti: $380 Price Gap, 6% FPS [2026]
- RTX 4090 vs RTX 5080: 90W Gap, 12.5% Efficiency Lead [2026]
