GPU Pricing Statistics (2026)

Graphics card pricing in 2026 has become one of the defining stories of the PC gaming industry, as an AI-driven memory shortage pushed street prices for GeForce RTX 50-series and Radeon RX 9000-series cards far above their launch MSRPs. This reference page collects the most-cited, verifiable figures from Jon Peddie Research, Tom’s Hardware, TrendForce, Newegg tracking and board-partner disclosures so journalists can quote them with confidence. Free to use with a link back to ShaneTheGamer as the source.
Key Takeaways for Journalists
- The RTX 5090’s median US street price hit $4,699.99 in August 2026 — roughly 135% above its $1,999 launch MSRP, according to Newegg tracking cited by ThinkComputers.
- The RTX 5070 rose about 36% to a $899.99 Newegg median, despite a $549 launch price, per the same tracking.
- GDDR6 spot prices roughly tripled from ~$2.50 to ~$7.50 per GB since late 2025, the core driver of the surge, according to Capital & Compute’s memory model.
- Memory now accounts for around 80% of a high-VRAM GPU’s cost, a structural reversal from when the silicon die dominated, per Chinese channel reporting compiled by TechTimes.
- Nvidia raised its GPU-and-memory kit price to board partners three times in 2026, compounding to roughly 32%–49% versus January while leaving MSRPs untouched.
- AMD’s RX 9070 XT crossed $1,000 in official China Q3 2026 guidance (~$1,038) — 40% above its March 2025 launch price.
- Nvidia held roughly 94% of discrete desktop GPU share in 2025, with AMD at an all-time low near 5%, per Jon Peddie Research.
- Desktop AIB shipments reached ~44.28 million units in 2025, up from 34.7 million in 2024, according to JPR.
- Q1 2026 AIB revenue hit $10.85 billion, up 76% year-over-year, on a 28% unit increase — evidence that dollars rose far faster than volume, per JPR.
- Analysts see no broad relief before 2027, with memory makers warning shortages may persist toward the end of the decade.
The Headline Numbers: Flagship to Entry Tier
The GeForce RTX 5090 is the single most distorted line in the 2026 market. Nvidia set a $1,999 MSRP at CES 2025, but by August 2026 the median US street price had climbed to a level far beyond that. The RTX 5090’s median price reached $4,699.99, about 135% above its $1,999 launch MSRP, according to Newegg data cited by ThinkComputers. During QuakeCon in early August, Tom’s Hardware’s GPU price tracker showed the best price for the RTX 5090 sitting at $4,381 — more than double the $1,999 MSRP that Nvidia set.
The pain is not limited to the flagship. The most affordable RTX 5080, 5070 Ti and 5070 options sat at $1,289 (29% over MSRP), $989 (32% over MSRP) and $629 (over 14% over MSRP), respectively, per Tom’s Hardware’s early-August reading. Newegg medians ran even higher: the RTX 5070 rose 36% to $899.99, while the RTX 5060 climbed 27% to $469.99, according to ThinkComputers.
Mid-range cards illustrate how strange the VRAM-tier pricing has become. Nvidia launched the RTX 5060 Ti 8GB at $379 and the 16GB model at $429, so buyers were originally paying just $50 more to double the VRAM. By August 2026 that gap had blown out: Newegg data put the median RTX 5060 Ti 8GB listing at $529.99 and the 16GB version at $804.99, a difference of $275 — despite both using the same GB206 GPU. The RTX 5060 Ti 16GB recorded the largest increase of any tracked card, reaching a median $804.99.
Regional pricing amplifies the swing. Prices for RTX 50-series GPUs jumped by 30% in South Korea as wafer costs from TSMC increased and GDDR7 modules pushed past $20, Tom’s Hardware reported, and Tech Insider noted South Korean retail RTX 5090 listings running up to about $5,112.
Why Prices Are Rising: The Memory Shock
Unlike the 2020–2021 crunch, the 2026 surge is not being driven by crypto mining or scalping alone — it is a memory-supply shock. The answer traces back to the memory chips soldered around the graphics processor: a worldwide DRAM and GDDR shortage, supercharged by the AI data-center boom, is squeezing supply, inflating costs, and pushing street prices well above MSRP. Nvidia’s RTX 50 series uses GDDR7; AMD’s Radeon RX 9000 cards use GDDR6, and both come from the same three suppliers serving AI servers.
The cost math is stark. GDDR6 spot prices moved from about $2.50 to about $7.50 per GB since autumn 2025, a roughly +200% move, according to Capital & Compute. TrendForce data compiled by other outlets showed blended DRAM contract prices jumping roughly 80% quarter-over-quarter in the first quarter of 2026, with some year-over-year figures cited well above 100%. Because AI accelerators pay far more per wafer, fabs have prioritised HBM, starving consumer GDDR.
That has fundamentally changed a GPU’s bill of materials. On high-VRAM models across the lineup, memory now accounts for 80% of GPU cost, a structural reversal from when the silicon die itself dominated costs, per China channel reporting via TechTimes. Nvidia passed this through to partners quietly: Nvidia repriced the GPU-and-memory kit it charges board partners three times in 2026 while leaving official MSRPs untouched — the RTX 5090 still lists at $1,999 even as street prices run past $4,830.
Analysts do not expect quick relief. Tech Insider cited IDC noting that 2026 DRAM and NAND supply growth of 16% and 17% year-on-year sits below historical norms, leaving price as the only release valve. Memory makers have warned the crunch could persist toward the end of the decade, which suggests elevated GPU pricing is a structural condition rather than a passing spike.
MSRP vs Street: A Widening Gap
One of the most important nuances for reporters is that official MSRPs have not been revised — the divergence lives entirely in wholesale kit pricing and retail markups. Official MSRPs have not been revised; the divergence between MSRP and shelf price reflects wholesale and channel pricing, not a change in the published figure, BuySellRam noted in its August report. As a result, quoting an MSRP in 2026 without a street price alongside it is misleading.
The tier-by-tier premium pattern is instructive. The RTX 5070 at 12GB sat about 15% over MSRP, while the three 16GB cards ran 18% to 31% over, per BuySellRam. Only the smallest cards behaved normally: the RTX 5060 and RTX 5060 Ti 8GB were the only two cards available at something close to their launch price, the same report found — a healthy market that in August 2026 existed only at 8GB.
Trackers using different methodologies broadly agree on direction if not on the exact figure. GPUDrip’s GPU Price Index put the RTX 5090’s average street price at +38.9% above MSRP as of its most recent reading, cooling from a June peak near +50%, while PCPriceWatch’s median tracker had the card at $2,150 in June 2026, up 31.0% from $1,641 in January. The spread across trackers reflects sampling differences — completed sales versus advertised listings — which reporters should flag when citing any single number.
| Model | Launch MSRP | Street price (Aug 2026) | Premium over MSRP |
|---|---|---|---|
| RTX 5090 | $1,999 | ~$4,699.99 (median) | ~135% |
| RTX 5080 | $999 | ~$1,249 new / $1,289 low | ~25–29% |
| RTX 5070 Ti | ~$749 | ~$989 (low) | ~32% |
| RTX 5070 | $549 | ~$899.99 (median) | ~64% |
| RTX 5060 Ti 16GB | $429 | ~$804.99 (median) | ~88% |
| RTX 5060 | $299 | ~$469.99 (median) | ~57% |
AMD Radeon: The Cheaper Camp, Still Rising
AMD’s RDNA 4 cards have risen more gently than Nvidia’s, keeping Radeon the relative value play. AMD launched the RX 9070 at $549, exactly matching the RTX 5070; by late 2025 it sold for roughly $540, and by mid-2026 cost approximately $650, about 20% higher — versus the RTX 5070’s roughly 67% climb over the same period. The flagship RX 9070 XT followed a similar shape: it launched at $599, reached roughly $580 street late in 2025, and current cards sit around $720 to $730, an increase of approximately 26%.
The formal price hikes are now official in some markets. AMD’s Q3 2026 China channel guidance set the RX 9070 XT at ¥6,999 (~$1,038), crossing the $1,000 threshold for the first time and standing 40% above the ¥4,999 (~$741) price at which it launched in China in March 2025, per TechTimes. Wccftech reported the broader notification with increases ranging anywhere from 5% to 20% across the RX 9000 stack.
AMD has been transparent that memory is the trigger. AMD raised its own RDNA 4 distributor pricing by $10 per 8GB of VRAM, adding $20 to the 16GB RX 9070 XT’s cost to channel partners, The FPS Review reported — though it noted that alone doesn’t explain the larger retail swings, where partner markups add the rest. As of publication, AMD had not announced any change to official US or European pricing, and the Q3 China SEP guidance is a China-market document.
For value-focused buyers the lineup still holds appeal. Board-partner listings such as the Gigabyte RX 9070 XT Gaming OC at $719 with 16GB GDDR6 — 33% more VRAM than competing RTX 5070 models — show why Radeon remains competitive despite the increases. ShaneTheGamer’s games-industry research hub tracks how these pricing shifts feed into broader PC gaming affordability.
Market Share, Shipments and Revenue
The pricing story sits atop a market where one vendor dominates. Nvidia grabbed a staggering 94% share of the desktop graphics card market in 2025, while AMD’s Radeon division collapsed to a mere 5% — the lowest market share in the company’s history, according to Jon Peddie Research. That is a dramatic shift from late 2024, when Nvidia held 84% and AMD 15%.
Volume actually grew even as prices climbed. The industry shipped 44.28 million graphics cards in calendar 2025, up from 34.7 million units in 2024, mainly because Nvidia released its GeForce RTX 50-series based on Blackwell, JPR reported. Sales peaked in Q3 2025 at around 12 million units before dipping slightly to 11.48 million in Q4.
Crucially, revenue is rising far faster than units — the clearest quantitative sign that price, not demand, is doing the heavy lifting. Jon Peddie told Tom’s Hardware that in Q1 2025, 9.25 million AIBs were sold, and in Q1 2026, 11.82 million, a 28% unit increase; POS tracking showed Q1 2025 at $6.18 billion and Q1 2026 at $10.85 billion, a 76% increase. That gap between a 28% volume rise and a 76% revenue rise is the datapoint most worth citing.
Looking ahead, JPR sees softness. The firm has forecast a market decline in 2026 as buyers delay upgrades, and Peddie warned that the AIB market is being squeezed from the bottom by powerful new notebooks and CPU integrated graphics, and from the high end by rising pricing due to competition, memory prices, and tariffs. Falling attach rates in prebuilt systems compound the pressure.
| Metric | Q1 2025 | Q1 2026 | Change |
|---|---|---|---|
| AIB units sold | 9.25 million | 11.82 million | +28% |
| AIB revenue (POS) | $6.18 billion | $10.85 billion | +76% |
| Nvidia desktop share | ~92% | ~94% | +2 pts |
The Used Market: A Poor Escape Hatch
Historically, the second-hand market offered relief when new cards spiked. In 2026 that safety valve has largely closed. A healthy GPU market exists only at 8GB, and the three 16GB cards run 18% to 31% over MSRP even before considering used pricing — and used listings now sit unusually close to new. BuySellRam’s August report found used RTX 50 cards listing within roughly 14% of new prices.
Flagship used pricing still shows a discount, but it is small in dollar terms given the elevated new price. BestValueGPU tracked the RTX 5090 at $4,299 new on Amazon versus around $3,590 used on eBay, against a $1,999 launch MSRP, while the RTX 5080 showed $1,249 new versus around $1,100 used, against a $999 launch MSRP. European sold-listing data from PCPrice.Watch put the used RTX 5090 median around €2,517 from tracked completed sales.
Older flagships have held value stubbornly. Tech Insider reported that every August-dated tracker lands in the same $2,100 to $2,500 window for the used RTX 4090, a tighter band than usual for a discontinued two-year-old card — because its 24GB of GDDR6X still commands a premium over anything Nvidia sells new below the RTX 5090. Reporters should treat used figures as advertised or sold-listing medians, not guaranteed transaction prices, given the wide variation by condition and warranty status.
Frequently Asked Questions
How much does an RTX 5090 cost in 2026?
As of August 2026 the RTX 5090’s median US street price reached about $4,699.99, roughly 135% above its $1,999 launch MSRP, according to Newegg tracking cited by ThinkComputers.
Why are GPU prices rising in 2026?
An AI-driven memory shortage is the main cause: GDDR6 spot prices roughly tripled from about $2.50 to $7.50 per GB since late 2025, and memory now makes up around 80% of a high-VRAM GPU’s cost.
Are AMD Radeon cards cheaper than Nvidia in 2026?
Generally yes — the RX 9070 rose about 20% versus the RTX 5070’s roughly 67% climb, though AMD’s Q3 2026 China guidance pushed the RX 9070 XT to about $1,038, some 40% above its launch price.
How much of the GPU market does Nvidia control?
Nvidia held about 94% of the discrete desktop GPU market in 2025 while AMD fell to a record-low ~5%, according to Jon Peddie Research.
Will GPU prices drop in 2026?
Analysts see no broad relief this year; IDC noted 2026 DRAM and NAND supply growth of just 16% and 17%, below historical norms, and memory makers have warned shortages could persist toward 2027 and beyond.
Methodology & Sourcing
This page compiles publicly reported figures from GPU price trackers (Newegg medians via ThinkComputers and PCGamesN, Tom’s Hardware’s price tracker, BestValueGPU, PCPrice.Watch, GPUDrip and Capital & Compute), market analysts (Jon Peddie Research, TrendForce, IDC) and board-partner disclosures (PC Partner, ASUS, plus China channel guidance). Street prices are medians or lowest in-stock listings and will differ from completed-sale values; trackers using different sampling methodologies produce different exact numbers, so where sources disagree we present ranges and label figures as medians, lows or estimates. MSRPs are Nvidia’s and AMD’s announced launch prices and have not been officially revised in the US or Europe as of publication — the gap to street price reflects wholesale kit pricing and retail markups. All figures were current as of the cited dates in mid-to-late 2026 and are subject to rapid change given the ongoing memory-supply volatility.
Sources
- Jon Peddie Research — GPU market share, shipments and AIB revenue (2025–2026)
- Tom’s Hardware — RTX 50 price tracker, QuakeCon pricing, PC Partner analysis (2026)
- ThinkComputers — Newegg median price data (2026)
- TrendForce — DRAM/GDDR contract and spot pricing (2026)
- IDC — 2026 DRAM and NAND supply-growth forecasts
- Capital & Compute — GDDR memory cost model and kit-price tracking (2026)
- TechTimes / Wccftech / Softonic — AMD Radeon RX 9000 China Q3 2026 guidance
- The FPS Review — ASUS Radeon pricing and RDNA 4 distributor pricing (2026)
- ThePCEnthusiast — RTX 50 and RX 9000 street-price comparisons (2026)
- BestValueGPU / PCPrice.Watch — new and used price trackers (2026)
- Tech Insider — regional pricing and used-market analysis (2026)
Spotted a newer figure or an error? Contact the ShaneTheGamer desk and we will update this page.





Join the Conversation