Micron has quietly stopped making the memory chips that sit at the heart of Nvidia’s GeForce RTX 50 series, and the fallout is rippling through an already strained GPU supply chain. According to Tom’s Hardware, Micron’s product catalog now lists both of its 2GB GDDR7 chips as “End of Life,” a designation first flagged by UNIKO’s Hardware on September 22, 2026, and confirmed by multiple outlets including TrendForce, TweakTown, and HWBusters over the following two days.
The two discontinued parts, model numbers MT68A512M32DF-28:A and MT68A512M32DF-32:A, are 28Gbps and 32Gbps variants of the 2GB (16Gb) GDDR7 density that Nvidia has used across most of the RTX 50 desktop lineup, including the GeForce RTX 5060, according to TrendForce. Micron’s 3GB (24Gb) GDDR7 parts remain in its catalog, a density aimed largely at higher-margin professional and AI-focused graphics products rather than mainstream gaming cards, and the price gap behind that preference is already stark: South Korean outlet MT.co.kr reported in July 2026 that 2GB GDDR7 chips were selling for around $20 apiece against $60 to $70 for the 3GB variant, roughly a 3x premium that goes a long way toward explaining why Micron would rather dedicate production lines to the pricier part.
The GDDR7 shortage adds a new front to a memory crisis that has already pushed notebook DDR5 prices up roughly sixfold in twelve months, according to system builder XMG, and forced PC makers like Acer and Schenker to raise prices across their lineups. This piece breaks down exactly what changed, who is still supplying GPU memory, and what it means for anyone shopping for a graphics card heading into 2027.
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: Micron Pulls the Plug on 2GB GDDR7
The discontinuation surfaced through Micron’s own public parts catalog rather than a formal press release. UNIKO’s Hardware, a well-followed hardware tracking account, posted on September 22, 2026 that Micron had marked both of its 2GB GDDR7 SKUs as End of Life. Japanese outlet Gazlog and TechPowerUp picked up the report the same day, and by September 24, Tom’s Hardware, TrendForce, and TweakTown had all independently verified the catalog change and published their own write-ups.
No outlet has published a direct quote from Micron explaining the decision. What the reporting establishes is narrower and more concrete: two specific 2GB GDDR7 part numbers have been pulled from production, while Micron’s 3GB GDDR7 lineup, including a 28Gbps part currently in production and a 32Gbps part available by inquiry, remains listed and active. Tom’s Hardware characterized the move as Micron pivoting toward higher-density memory for AI, but that framing is the publication’s interpretation of the catalog change and broader market conditions, not a confirmed statement from Micron itself.
That distinction matters. Discontinuing a product line is not the same as announcing a strategic pivot, and Micron has not published guidance tying the GDDR7 change to a specific percentage of capacity moving toward AI accelerators. What is verifiable is the supply consequence: GPU makers that depend on 2GB GDDR7 now have one fewer chip supplier to draw from.
The Chips at the Center of the GDDR7 Shortage
GDDR7 memory chips come in different densities measured in gigabits, which translate to gigabytes per chip. A 2GB chip holds 16Gb of data; a 3GB chip holds 24Gb. GPU memory buses are built from multiple chips wired in parallel, so the density of each chip directly shapes how much total VRAM a graphics card can carry. Eight 2GB chips produce a 16GB card. Swap in eight 3GB chips instead, and the same board design yields 24GB.
Micron’s now-discontinued parts were both members of the 2GB density family: one clocked at 28Gbps, the other at a faster 32Gbps. Both have been standard components in the current generation of Nvidia GeForce RTX 50 desktop cards. The GeForce RTX 5060 specifically has shipped with Micron’s 28Gbps 2GB chip, per TrendForce’s reporting, making it one of the most directly exposed products in Nvidia’s current lineup. TrendForce noted that, as of September 2026, 3GB GDDR7 is now the only density left in Micron’s catalog, and VideoCardz pegged those 3GB chips at roughly $60 to $70 apiece that same month, against about $20 for the 2GB part they would be replacing, a cost gap that would hit a budget-tier card like the 5060 hardest if board partners were ever forced to switch densities.
Micron’s GDDR7 Catalog, Before and After
| Part Number | Density | Speed | Status as of Sept 2026 | Primary Use Case |
|---|---|---|---|---|
| MT68A512M32DF-28:A | 2GB (16Gb) | 28Gbps | End of Life | RTX 50 series (e.g. RTX 5060) |
| MT68A512M32DF-32:A | 2GB (16Gb) | 32Gbps | End of Life | RTX 50 series desktop cards |
| 3GB GDDR7 (28Gbps) | 3GB (24Gb) | 28Gbps | Active, in production | Professional / AI-focused GPUs |
| 3GB GDDR7 (32Gbps) | 3GB (24Gb) | 32Gbps | Active, available by inquiry | Professional / AI-focused GPUs |
The available reporting does not identify any confirmed RTX 60-series or AMD Radeon RX 9000-series model using Micron’s 2GB GDDR7 chips. Nvidia has not shipped an RTX 60-series product as of this writing, and AMD’s current Radeon lineup relies on GDDR6 rather than GDDR7, which insulates AMD’s existing cards from this specific supply disruption even as the broader memory crisis touches its own component costs.
Which GPUs Are Actually Affected
The immediate impact lands on Nvidia’s GeForce RTX 50 series, the current desktop lineup built around GDDR7 since it debuted in early 2025. Most RTX 50 boards are designed around the 2GB chip density, and the reporting from TweakTown and HWBusters indicates that Nvidia’s remaining 2GB GDDR7 supply for these cards now comes from just two companies: Samsung and SK hynix.
Two Suppliers Left Standing
Micron had only joined the RTX 50 supply chain in the past year or so, according to TweakTown’s reporting, meaning it exits the 2GB GDDR7 business shortly after ramping up production for it. That leaves Samsung and SK hynix carrying the full weight of 2GB GDDR7 demand across every RTX 50 board built around that density, from budget-tier cards up through higher-end desktop SKUs. Fewer suppliers for a single component generally means less negotiating leverage for the buyer, in this case Nvidia and its add-in board partners, and less redundancy if either remaining supplier faces its own production hiccup.
It’s worth noting what the current reporting does not establish. No outlet has published a confirmed retail price increase for a specific RTX 50 card tied directly to the Micron exit, and no analyst estimate has quantified exactly how much narrower supplier competition could add to a graphics card’s bill of materials. The consequence, for now, is reduced supplier diversity rather than a documented price shock.
Why This Is Happening: AI Memory Demand vs. Gaming GPUs
The GDDR7 shortage did not emerge in isolation. It lands squarely inside a broader memory market where AI infrastructure spending has become the dominant force shaping component pricing and allocation across the entire chip industry. Tom’s Hardware’s RAM coverage this month noted that memory chips are now, on a per-area basis, more expensive than compute chips in some cases, an unusual reversal that reflects how scarce and profitable advanced memory has become as AI accelerator demand competes directly with consumer electronics for the same fabrication capacity.
Higher-density GDDR7, at 3GB per chip, carries a better margin profile and fits more naturally into professional and AI-oriented graphics products where customers are far less price-sensitive than a gamer shopping for a sub-$400 graphics card. That margin advantage shows up directly in per-chip pricing: VideoCardz reported in September 2026 that 3GB GDDR7 was running roughly $60 to $70 per chip, about triple the cost of a standard 2GB module, which MT.co.kr had placed at around $20 just two months earlier. A memory maker facing a finite number of production lines has a straightforward incentive to prioritize the parts that data center customers will pay a premium for, even if that means walking away from a consumer GPU commitment made only months earlier.
Micron underscored that AI-first posture elsewhere in its lineup this month, announcing 512GB DDR5-9200 memory modules built for server and AI workloads, with a reported power draw of 16 watts per module, according to Tom’s Hardware’s RAM news coverage. That kind of high-capacity, high-performance module sits at the opposite end of the product stack from a budget 2GB GDDR7 chip destined for a mainstream gaming card, and it illustrates where memory makers see the more durable demand.
The Wider Memory Crisis: DDR5, DRAM, and the AI Squeeze
The GDDR7 news arrives on top of a memory crisis that has already reshaped PC pricing throughout 2026. System builder XMG told TechRadar that notebook DDR5 memory has climbed close to six times its price from before the current RAM crisis began, a figure serious enough that XMG raised its own laptop prices and explicitly warned that Nvidia GPUs could suffer the same fate. Fellow system builder Schenker also raised laptop prices this month and cautioned that the shortage could continue through 2027, while XMG separately said it expects DRAM and NAND supply to stay tight through 2027 and beyond.
That timeline lines up with what PC makers have already told customers directly. Acer has warned that PC prices could rise as much as 20 percent because of the memory crunch, and Valve’s Steam Deck has seen a price hike reported as high as 46 percent tied to the same DRAM shortage. Those increases predate the specific GDDR7 story, but they establish the pattern the GPU market is now stepping into: components that were priced for a stable supply chain are being repriced for a scarce one.
Export Controls Add Another Layer of Friction
Government policy is compounding the supply squeeze rather than easing it. South Korea’s Ministry of Trade, Industry and Energy added high-performance AI chips and advanced semiconductor manufacturing equipment to the country’s strategic export control list effective September 1, 2026, requiring exporters to secure government approval before shipping covered products overseas. That policy carries outsized weight for the memory market specifically because Samsung and SK hynix, now the sole remaining suppliers of 2GB GDDR7 for Nvidia’s RTX 50 series, are both South Korean companies whose DRAM, NAND, and HBM exports could fall under the new approval requirements.
Separately, a September 23, 2026 legal analysis noted that advanced semiconductors and computing products that incorporate them, including servers, remain subject to export, re-export, and in-country transfer controls for a defined list of restricted destinations, a category that includes China. None of this directly targets GDDR7 gaming chips, but it adds regulatory overhead to the same companies now shouldering all of Nvidia’s 2GB GDDR7 demand, at the exact moment that demand has nowhere else to go.
China’s Memory Makers Are Racing to Fill the Gap
While Samsung, SK hynix, and Micron reposition around AI demand, Chinese memory manufacturers are pushing to close the technology gap. CXMT has reportedly reached a 12nm-class DRAM milestone and begun mass production using its fifth-generation DRAM process technology, while YMTC has been expanding into NAND flash territory typically dominated by Western and South Korean suppliers. Neither company has been named as a GDDR7 supplier for Nvidia’s current cards, but their progress on core DRAM manufacturing signals a longer-term shift in where global memory capacity could come from if the current supplier concentration around Samsung and SK hynix continues.
Market Impact: Data Points From the 2026 Memory Crisis
| Date (2026) | Development | Source |
|---|---|---|
| Throughout the year | Notebook DDR5 prices climb toward 6x their pre-crisis level | XMG, via TechRadar |
| Reported this month | Acer warns PC prices could rise up to 20% due to memory costs | Acer |
| Reported this month | Steam Deck price hike reported as high as 46% | Valve |
| Sept 1 | South Korea adds AI chips and fab equipment to strategic export control list | South Korea Ministry of Trade, Industry and Energy |
| Sept 22 | UNIKO’s Hardware flags Micron’s 2GB GDDR7 chips as End of Life | UNIKO’s Hardware, via Gazlog |
| Sept 24 | Tom’s Hardware, TrendForce, TweakTown confirm Micron GDDR7 discontinuation | Tom’s Hardware, TrendForce, TweakTown |
| Sept 25 | TSMC reportedly plans wafer price hikes of 3-6% starting January 2027 | Digitimes |
That last data point matters beyond memory specifically. Digitimes reported on September 25, 2026 that TSMC plans to adjust wafer prices across different manufacturing processes starting in January 2027, with hikes in the range of 3 to 6 percent, citing tight capacity and strong AI demand. Combined with the GDDR7 supplier squeeze and the broader DRAM crisis, it paints a picture of a chip industry where nearly every layer of the supply chain, from raw wafers to finished memory chips, is being repriced around AI infrastructure demand rather than consumer electronics.
Historical Context: From GDDR6 to an AI-Driven Memory Land Grab
GDDR7 arrived as the successor to GDDR6 alongside Nvidia’s RTX 50 series generation in early 2025, promising higher bandwidth for the same or lower power draw. In its first year and a half on the market, GDDR7 supply looked ordinary by chip industry standards: multiple suppliers, including Micron, Samsung, and SK hynix, all offered comparable 2GB parts to keep Nvidia’s board partners stocked. What has changed is not the technology itself but the economics surrounding it. As AI accelerator demand has pulled memory makers toward higher-density, higher-margin products, the same 2GB density that once had three viable suppliers is down to two, and the underlying reason is not a shortage of raw manufacturing capacity so much as a reallocation of who gets it.
That reallocation echoes a pattern already visible elsewhere in the industry. Samsung and TSMC have both committed to ASML’s High NA EUV lithography machines, with Samsung targeting memory production by 2028 and TSMC targeting advanced logic by 2030, investments aimed squarely at the manufacturing capacity that AI chips will require for the rest of the decade. GPU memory for gaming, by comparison, increasingly looks like a legacy product line competing for factory time against far more lucrative AI silicon.
Competitive Comparison: Micron, Samsung, and SK hynix
The three companies now defining the GDDR7 supply picture are taking noticeably different paths. Micron has exited the 2GB density entirely, keeping only its higher-margin 3GB parts active, a move that reads as a full pivot away from mainstream gaming GPU memory. Samsung and SK hynix, by contrast, remain the last two vendors still supplying 2GB GDDR7 to Nvidia’s RTX 50 lineup, at least for now, even as both companies navigate South Korea’s new export approval requirements and the same AI-driven incentive to prioritize higher-density, higher-margin memory that pushed Micron out.
On the GPU side, Nvidia and AMD are effectively insulated differently from this specific disruption. Nvidia’s current RTX 50 generation is built around GDDR7 and therefore directly exposed to the shrinking supplier base. AMD’s Radeon RX 9000 series, which uses GDDR6, sidesteps the GDDR7-specific shortage even as it remains subject to the same broader DRAM and NAND cost pressures hitting the rest of the PC industry, a dynamic already visible in AMD’s own recent RX 9070 XT software and driver push aimed at squeezing more performance out of existing hardware rather than relying on a new GPU generation. Neither Nvidia nor AMD has issued a public statement addressing Micron’s GDDR7 exit directly, based on the available reporting.
What the Industry Is Saying
Coverage of the GDDR7 shortage has been consistent across outlets even without an official Micron statement to work from. Tom’s Hardware framed the discontinuation as part of a broader pivot toward higher-density, AI-oriented memory. TrendForce described the change as narrowing supply options for Nvidia’s RTX 50 series specifically. TweakTown and HWBusters both emphasized the practical consequence: Nvidia’s 2GB GDDR7 supply for current-generation cards now rests entirely on Samsung and SK hynix. XMG, for its part, extended the warning beyond memory modules and SSDs to graphics cards directly, telling TechRadar that it fears Nvidia GPUs could suffer the same cost pressure already hitting notebooks.
What none of the coverage includes, notably, is a Micron statement contextualizing the move, an Nvidia response addressing supply continuity for RTX 50 cards, or an AMD comment on how the shortage might affect competitive positioning for Radeon. The story, for now, is built entirely from catalog changes and third-party reporting rather than corporate messaging, which is itself telling about how quietly these supply shifts are happening across the memory industry.
What This Means for GPU Shoppers
For anyone shopping for a graphics card in late 2026, the practical takeaway is caution rather than panic. No verified retail price increase has been tied directly to the Micron GDDR7 exit as of this writing, and existing RTX 50 inventory built with Micron’s now-discontinued chips will continue to work exactly as it always has. The concern is forward-looking: as Samsung and SK hynix become Nvidia’s only 2GB GDDR7 suppliers, any production hiccup, allocation shift toward AI customers, or new export friction at either company could translate into tighter card availability or higher wholesale memory costs that eventually reach store shelves.
That risk sits on top of an Nvidia roadmap that is already delayed. Nvidia’s next-generation RTX 60 series has reportedly slipped to 2028, meaning the current RTX 50 lineup, and its narrowing GDDR7 supplier base, will likely remain Nvidia’s primary consumer offering for longer than a typical GPU generation. Buyers hoping a next-generation refresh will reset memory supply and pricing dynamics may be waiting considerably longer than usual, a timeline that also lines up with reports that wafer prices industry-wide could climb as much as 40 percent by 2027.
Predictions: Where the GDDR7 Shortage Goes From Here
Based on the trajectory of the memory market through 2026, a few outcomes look more likely than others heading into 2027.
- Samsung and SK hynix will gain pricing leverage over Nvidia’s remaining 2GB GDDR7 demand, since Nvidia and its board partners no longer have a third supplier to play off against the other two.
- Expect further quiet End of Life notices on lower-margin memory SKUs from Micron and potentially other suppliers, following the same low-visibility playbook used for the 2GB GDDR7 exit, as AI contracts continue to outbid consumer electronics for fab capacity.
- Nvidia’s delayed RTX 60 series, not expected before 2028 according to leaker reporting, will likely launch with 3GB-only GDDR7 as the baseline density, pushing VRAM capacities higher but also raising bill-of-materials costs across the lineup.
- AMD’s GDDR6-based Radeon RX 9000 series could gain a temporary pricing or availability advantage over comparable Nvidia SKUs if GDDR7 supply tightens further, even though AMD remains exposed to the broader DDR5 and NAND cost increases.
- Component cost pressure will keep surfacing in finished-product pricing, following the pattern already visible in Acer’s PC price warnings and the Steam Deck price hike, meaning graphics card MSRPs are a more likely pressure point than an outright shortage of RTX 50 cards on shelves.
The Bigger Picture: AI Is Rewriting Memory Economics
Strip away the specific part numbers, and the GDDR7 story is really about a chip industry reorganizing itself around AI infrastructure spending. GPU gaming memory used to be a straightforward, high-volume, thin-margin business with several reliable suppliers. It is increasingly becoming a secondary priority behind AI accelerators, high-capacity server DRAM, and other products where customers pay a premium and sign longer-term contracts. Micron’s exit from 2GB GDDR7 is a small, specific data point, but it fits a pattern already visible in TSMC’s planned wafer price hikes, South Korea’s export controls on advanced chips, and the sixfold rise in notebook DDR5 pricing that XMG and Schenker have both flagged this year.
None of this means gaming GPUs are about to disappear from Nvidia’s or AMD’s roadmaps. It does mean that the memory inside those cards is being manufactured by companies with increasingly attractive alternatives for their production capacity, and that gamers are, for the first time in years, competing directly with AI data centers for the same silicon.
Frequently Asked Questions
What exactly did Micron discontinue?
Micron marked two 2GB GDDR7 memory chips, MT68A512M32DF-28:A (28Gbps) and MT68A512M32DF-32:A (32Gbps), as End of Life in its product catalog, according to reporting confirmed by Tom’s Hardware, TrendForce, and TweakTown between September 22 and 24, 2026.
Which graphics cards use these Micron GDDR7 chips?
Micron’s 2GB GDDR7 parts have shipped in Nvidia’s GeForce RTX 50 series desktop cards, with the GeForce RTX 5060 specifically identified by TrendForce as using the 28Gbps variant. No confirmed reporting ties these parts to AMD’s Radeon RX 9000 series, which uses GDDR6 rather than GDDR7.
Will this make RTX 50 graphics cards more expensive?
No confirmed retail price increase has been reported as directly tied to Micron’s GDDR7 exit. The concern raised by outlets like TechRadar is about future risk: with fewer suppliers, Nvidia and its board partners have less leverage on memory pricing and less cushion if either remaining supplier faces disruption.
Who still makes 2GB GDDR7 chips for Nvidia’s RTX 50 series?
Based on reporting from TweakTown and HWBusters, Samsung and SK hynix remain the two suppliers still producing 2GB GDDR7 for Nvidia’s current-generation graphics cards after Micron’s exit.
Is this connected to the broader RAM and DRAM shortage?
Yes. The GDDR7 exit lands alongside a separate but related crisis in DDR5 notebook memory, which system builder XMG says has climbed roughly sixfold in price over twelve months, and rising costs across DRAM and NAND generally, driven in large part by AI infrastructure demand for memory capacity.
Did Nvidia or AMD comment on the GDDR7 shortage?
Neither company has issued a public statement addressing Micron’s 2GB GDDR7 discontinuation, based on the reporting available as of this writing.
When will Nvidia’s next GPU generation address this?
Nvidia’s next-generation RTX 60 series has reportedly been delayed to 2028, according to a leaker cited by multiple outlets, meaning the current RTX 50 generation and its narrowing GDDR7 supply base are likely to remain Nvidia’s primary consumer lineup for an extended period.
What is the difference between 2GB and 3GB GDDR7 chips?
A 2GB GDDR7 chip holds 16Gb of data, while a 3GB chip holds 24Gb. Since GPUs use multiple memory chips in parallel, swapping to higher-density chips can increase total VRAM capacity on the same board design, though it typically comes at a higher cost per chip: MT.co.kr reported 2GB chips selling for around $20 in July 2026 versus $60 to $70 for the 3GB variant, and VideoCardz’s September 2026 pricing check found roughly the same 3x gap still holding.
Related Coverage
- Rayman Legends Retold Needs RTX 4080 for 4K/60 [2026]
- Apple vs Nvidia Market Cap: Nvidia Leads by $497B [2026]
- HDMI 2.2 Gaming Setup: 12 Steps, 96Gbps Cables [2026]
- iPhone Air 2 Slips to Spring 2027 After 700K Sales [2026]
- NZXT vs Corsair vs Arctic AIO Coolers: $204 Gap [2026]
