Micron Ends 2GB GDDR7 Production Leaving Nvidia RTX 50 Series With Two Memory Suppliers

Micron Technology has marked both of its 2GB (16Gb) GDDR7 memory chips as End of Life, according to its public product catalog. The two parts are the 28 Gbps MT68A512M32DF-28:A and the 32 Gbps MT68A512M32DF-32:A. With those designations, Samsung and SK hynix become the only remaining suppliers of 2GB GDDR7 for Nvidia's GeForce RTX 50 series. The change was first spotted by UNIKO's Hardware on X and has since been reported by Tom's Hardware, VideoCardz, Hardware Busters, and TrendForce.

The GeForce RTX 5000 series launched on January 30, 2025. At that time, Samsung was reportedly Nvidia's sole GDDR7 supplier. SK hynix joined the supply chain in April, followed later by Micron. Micron's GDDR7 was first spotted on a shipping RTX 50 product in February 2026, when a GALAX GeForce RTX 5060 used four 2GB D8FHL packages for a total of 8GB.

The 2GB density remains the foundation of Nvidia's mainstream desktop GeForce lineup. The RTX 5090 uses 16 chips on a 512 bit bus for 32GB. The RTX 5080 and RTX 5070 Ti use eight chips on 256 bits for 16GB. The RTX 5070 uses six chips on 192 bits for 12GB. Micron's 3GB (24Gb) GDDR7 parts stay in production, and those denser chips are used by RTX PRO Blackwell workstation cards and the GeForce RTX 5090 Laptop GPU.

Micron has not published a notice or explained the decision. Its catalog pages for the 16Gb module norms now return No results found. Micron's 3GB modules are also much more expensive: VideoCardz puts them at around $60 to $70 each, roughly triple the $20 price of standard 2GB modules despite offering just 50% more capacity.

Key Facts

Tom's Hardware reported on September 24, 2026, that Micron has ended production and sales of its 2GB (16Gb) GDDR7 memory in both 28 Gbps and 32 Gbps variants. Micron just began supplying its 2GB GDDR7 to Nvidia earlier this year, and Samsung and SK hynix cover the vast bulk of 2GB GDDR7 modules for the standard RTX 50 retail lineup, so the risk of immediate shortages is minimal.

VideoCardz reported on September 23, 2026, that the D8FHL chip listed as MT68A512M32DF-28:A has reached end of life. The publication noted that Micron had offered 16Gb GDDR7 parts at both 28 Gbps and 32 Gbps. The MT68A512M32DF-28:A operates at 28 Gbps, matching the memory speed used by the RTX 5060. In February 2026, VideoCardz reported the first GeForce RTX 50 graphics card spotted with Micron GDDR7: a GALAX GeForce RTX 5060 that used four D8FHL packages for 8GB of capacity.

Hardware Busters reported on September 23, 2026, that Micron has quietly stopped making the GDDR7 density that almost every GeForce RTX 50 graphics card is built around. The site noted that the change was first spotted by UNIKO's Hardware on X and has since been picked up by TweakTown and VideoCardz. Micron has not published a notice or explained the decision. That leaves Samsung and SK hynix as the only two sources of 2GB (16Gb) GDDR7 for Nvidia's current lineup, while Micron's 3GB (24Gb) parts stay in production.

TrendForce reported on September 24, 2026, that Micron is reportedly phasing out its 2GB GDDR7 chips used in several Nvidia GeForce RTX 50 series graphics cards. The report, citing TechPowerUp and Gazlog, said the products have reached End of Life status. Micron's catalog now lists its 28 Gbps D8FHL chip (MT68A512M32DF-28) as EOL, with the 32 Gbps version carrying the same designation. The move leaves 3GB modules as Micron's only remaining GDDR7 offering.

According to Gazlog, when Nvidia's GeForce RTX 5000 series launched on January 30, 2025, Samsung was reportedly its sole GDDR7 supplier. SK hynix joined the supply chain in April, followed by Micron. Micron's GDDR7 was first spotted on a GALAX GeForce RTX 5060 board in February 2026, which used four 2GB chips for a total of 8GB. TechPowerUp suggests Micron's move is unlikely to prompt Nvidia to shift the existing RTX 50 lineup to 3GB GDDR7 given the steep cost premium. Denser 4GB and 6GB GDDR7 dies are on the way, but not before 2027 or 2028 by current estimates.

Analysis

What this really means is that Micron is exiting a low margin, highly competitive segment to free up fabrication capacity for higher margin, higher density memory aimed at AI accelerators. Micron's uncompetitive position in the 2GB segment appears to have made a fast exit logical. The company was the last of the three DRAM makers to show up on consumer Blackwell cards, and its chips only shipped on a few Nvidia models, such as retail variants of the GeForce RTX 5060. Losing a supplier that shipped little volume is not a supply shock, but it is the wrong direction for a memory market that has already pushed graphics card prices up more than once this year.

The bigger picture here is that Nvidia's RTX 50 series is down to two 2GB GDDR7 suppliers for its mainstream desktop cards. Samsung and SK hynix still provide the bulk of supply, so there is no immediate shortage. However, the reduced supplier base narrows Nvidia's options if either remaining supplier faces production issues or reallocates capacity toward AI memory. The 3GB module is what Nvidia reaches for when it wants more capacity without a wider bus, but board partner leaks indicate 3GB modules are currently much more expensive than 2GB variants.

Micron's 3GB GDDR7 parts allow graphics card makers to increase capacity without increasing the number of memory packages. A four chip design can provide 12GB instead of 8GB, while eight packages provide 24GB instead of 16GB. The 3GB module is also what the long rumoured RTX 50 Super refresh is expected to use. Reports earlier this year said those cards were being held back partly because 3GB chips cost several times as much as 2GB ones. VideoCardz puts the 3GB chips at around $60 to $70 each, roughly triple the $20 price of standard 2GB modules despite offering just 50% more capacity. That cost premium helps explain why Nvidia might keep the existing RTX 50 lineup on 2GB chips rather than switching to 3GB.

It is not very clear what Micron is planning or why it would discontinue only the 2GB modules while keeping the 3GB variants. The 3GB GDDR7 modules are currently used by RTX PRO Blackwell cards and the GeForce RTX 5090 Laptop GPU. That means Micron would effectively stop offering the 2GB GDDR7 memory used by current desktop GeForce models. Amid tightening memory supply, fresh market chatter suggests the GR20x family, expected to underpin the Rubin based RTX 60 series, may have slipped from 2H27 into 2028. If that timing holds, the memory supply picture for consumer graphics could remain constrained for several years.

Why It Matters

For Nvidia, the loss of Micron as a 2GB GDDR7 supplier reduces redundancy in a component that sits at the heart of nearly every mainstream GeForce RTX 50 card. The RTX 5090, RTX 5080, RTX 5070 Ti, and RTX 5070 all rely on 2GB chips in different bus configurations. Samsung and SK hynix now carry that load alone. While both suppliers have been consistent, a two supplier market is more vulnerable to pricing pressure, allocation shifts, or yield problems than a three supplier market.

For consumers, the change adds another layer of uncertainty to graphics card pricing. Memory supply has already tightened, and graphics card prices have risen more than once this year. If Samsung or SK hynix prioritizes higher margin AI memory or long term contracts, the 2GB GDDR7 supply for retail RTX 50 cards could become tighter. The steep premium on 3GB modules also means that any future capacity bump, whether through an RTX 50 Super refresh or a later generation, is unlikely to come cheaply.

For Micron, the decision reflects a broader strategic pivot. The company keeps its 3GB (24Gb) GDDR7 parts in production and is reportedly pivoting toward higher density memory for AI. Denser 4GB and 6GB GDDR7 dies are on the way, but not before 2027 or 2028 by current estimates. Micron's exit from 2GB GDDR7 for gaming GPUs is therefore less about gaming and more about where the company sees the best returns in a market increasingly shaped by AI demand.

Next Up

The immediate question is whether Nvidia will adjust its RTX 50 supply strategy. TechPowerUp suggests Micron's move is unlikely to prompt Nvidia to shift the existing RTX 50 lineup to 3GB GDDR7 given the steep cost premium. If that holds, Samsung and SK hynix will continue to supply 2GB modules for the standard retail lineup. The longer term question is when denser GDDR7 arrives. Denser 4GB and 6GB GDDR7 dies are on the way, but not before 2027 or 2028 by current estimates.

Another item to watch is the RTX 50 Super refresh, which is expected to use 3GB modules. Reports earlier this year said those cards were being held back partly because 3GB chips cost several times as much as 2GB ones. Meanwhile, market chatter suggests the GR20x family, expected to underpin the Rubin based RTX 60 series, may have slipped from 2H27 into 2028. With Micron focused on higher margin AI memory, Samsung and SK hynix remain the two names to track for 2GB GDDR7 supply through the rest of 2026 and beyond.