Nvidia.com

NVIDIA's consumer graphics division is operating under conditions that would have seemed implausible a year ago: GPU supply to its manufacturing partners has been cut by as much as 20 percent, two of its most capable mid-range cards have effectively vanished from shelves, and prices across the RTX 50 series have risen by an average of 15 to 19 percent globally since late 2025. If you are planning a PC build or a GPU upgrade in 2026, the window for buying at anything close to launch pricing has already closed.

In January 2026, hardware leaker MEGAsizeGPU posted on X that NVIDIA had reduced GPU allocations to its add-in card (AIC) partners — the companies that actually manufacture and sell the discrete graphics cards consumers buy — by 15 to 20 percent. According to Tom's Hardware's reporting on the leak, NVIDIA was still bundling its GPU dies with GDDR7 memory for partners — a technically important detail — but no new consumer products appeared to be coming in 2026. That second claim has since evolved: as of early June 2026, leaks suggest a refreshed Super lineup may arrive, with the most credible current timeline pointing to CES 2027 rather than any 2026 window.

RTX 5070 Ti and RTX 5060 Ti 16GB Effectively Sidelined

What happened to the RTX 5070 Ti and the RTX 5060 Ti with 16 gigabytes of video memory is a case study in supply chain reprioritization rather than a formal product cancellation. In January 2026, Hardware Unboxed reported that ASUS — the largest NVIDIA board partner by volume — had told it those two models were in end-of-life status. ASUS subsequently walked that statement back, clarifying that the two cards "have not been discontinued or designated as end-of-life" and attributing the supply problems to memory constraints rather than a production halt. NVIDIA echoed that position, confirming it would "continue to ship all GeForce SKUs" while acknowledging that "memory supply is constrained." Tom's Guide covered the full sequence of statements and retractions.

The practical result, however, has been the same as a discontinuation in many markets. Retailers in Australia reported in January 2026 that they could not source RTX 5070 Ti stock from distributors and did not expect new shipments through at least the end of that quarter. Prices on remaining units climbed from approximately $730 to more than $1,250 in the United States — a 71 percent increase over their lowest available prices.

For the RTX 5060 Ti 16GB, the pattern was similar. The card launched at $429 but was selling for $489 to $530 or more by mid-January 2026 as supply dried up, a 15 to 25 percent jump in just weeks. Its effective disappearance from shelves arrived with particular frustration for builders who had criticized NVIDIA's handling of that card's launch — including limited initial stock and delayed driver review access.

Why GDDR7 Scarcity Drives the Shortage

The supply problem is rooted in a specific engineering and economic reality: GDDR7 is the memory standard used in NVIDIA's Blackwell-based RTX 50 series, and its production is constrained by the same manufacturing capacity pressures affecting all advanced DRAM. GDDR7 delivers up to 60 percent more bandwidth per pin than GDDR6, using PAM-3 signaling technology that allows three data states per transfer cycle rather than two — an architectural change that also supports the on-die error correction features AI accelerator workloads require.

That dual applicability is the core of the problem. The same Samsung, SK Hynix, and Micron fabs that produce GDDR7 modules for consumer graphics cards are also producing other advanced DRAM for AI infrastructure. While the large training accelerators that fill data centers use a separate memory type called HBM (High Bandwidth Memory) rather than GDDR7, the competition for fab capacity is real: every wafer allocated to HBM production is a wafer unavailable for GDDR7, GDDR6, or consumer DDR5. As SK Group Chairman Chey Tae-won noted at Computex in June 2026, the memory supply shortfall was expected to persist at more than 20 percent above demand through 2030, and new fab capacity takes at least four to five years to bring online. TechTimes covered the full scope of the HBM wafer crisis and its downstream effects on consumer memory.

For NVIDIA specifically, the allocation math has clear business logic. A single AI accelerator requires far more memory per unit than a consumer gaming card, and generates far more revenue. In a constrained supply environment, NVIDIA has directed available memory toward its higher-margin commercial products. Consumers are absorbing the remainder.

Why Adding VRAM Requires New Chips, Not Just More of Them

The reason any RTX 50 Super refresh is taking longer than the usual mid-generation cadence is an engineering constraint that goes beyond NVIDIA's supply chain decisions. The RTX 50 series was built around 2-gigabyte GDDR7 memory modules — each module an individual chip soldered to the circuit board around the GPU die. To increase a card's total VRAM without redesigning the board layout or widening the memory bus, NVIDIA needs physically denser chips: 3-gigabyte modules instead of 2-gigabyte modules, using the same footprint on the same 128-bit or 192-bit bus.

The math is straightforward. An RTX 5060 uses a 128-bit memory bus with four 2GB modules, giving it 8GB total. Replace those four chips with 3GB modules on the same bus and the total rises to 12GB — the rumored configuration for the RTX 5060 Super. An RTX 5070 uses a 192-bit bus with six 2GB modules for 12GB total; the Super variant would reportedly use six 3GB chips for 18GB. According to Overclock3D's reporting on the Super specifications, all four Super variants would switch from 2GB to 3GB GDDR7 modules, with higher CUDA core counts and power limits to accompany the memory upgrades.

This constraint — not just NVIDIA's commercial priorities — explains why the RTX 50 Super series has slipped. The cards cannot ship until the 3GB modules are available in volume sufficient to support a product launch, at prices low enough not to make the resulting cards prohibitively expensive. As of June 2026, BenchLife indicated that the most realistic window for the Super series was early 2027, around CES 2027. The same leaker MEGAsizeGPU, who first reported the January supply cuts, said on June 5 that the Super refresh was "back on track" but acknowledged not having a specific release window, per Tom's Hardware's June 5 report on the leaked lineup.

GPU Prices Rise as RTX 50 Series Inventory Shrinks

A global pricing analysis published by TechSpot in February 2026, covering data from November 2025 and February 2026 across ten regional markets, found that the average RTX 50 series price had risen by 19 percent over three months — effectively pushing buyers in each budget tier down by one product generation. A budget of $1,000 that would have covered an RTX 5080 in November 2025 bought only an RTX 5070 Ti by February 2026.

The RTX 5090 — NVIDIA's flagship gaming GPU at launch — had its entry price effectively double from its November 2025 low of approximately $1,999 to more than $3,000 for typical available models by February 2026, with some premium configurations reaching $5,000 or more. Tom's Hardware's current GPU price tracking shows the RTX 5080 and RTX 5070 Ti remain in poor supply at any reasonable price, while 8GB variants of the RTX 5060 and 5060 Ti are available near MSRP. NVIDIA has not updated official MSRP pricing for any RTX 50 series card.

ASUS, MSI, Gigabyte, and ZOTAC — the four major AIC partners affected by NVIDIA's allocation changes — have limited leverage to constrain prices when incoming supply is restricted. With fewer chips flowing through the channel, basic supply and demand has driven retail prices upward in most regions.

AMD's position is worth noting, but does not offer an easy alternative for buyers seeking 16GB or more of VRAM. AMD's Radeon RX 9000 series uses GDDR6 rather than GDDR7, which insulates it somewhat from the most acute shortages. The same TechSpot analysis found that prices on AMD's RX 9070 XT rose by a comparatively modest $50 on average over the same period. For buyers who do not require RTX-exclusive features such as DLSS 4.5 with Multi-Frame Generation, AMD's current mid-range lineup offers more stable pricing and better availability.

What This Means for Builders and Upgraders in 2026

No new desktop GPU architecture from either NVIDIA or AMD is confirmed for 2026. AMD's next-generation RDNA 5 is broadly expected in the second half of 2027 by board partners who spoke at Computex in early June, with some indicating the schedule could slip to early 2028, according to Tom's Hardware's Computex reporting. NVIDIA's next full generation, codenamed Rubin, has been discussed publicly only in the context of data center accelerators; there is no announced timeline for a consumer Rubin GPU.

The RTX 50 Super series remains the only near-term candidate for new consumer NVIDIA hardware, but its timeline is genuinely uncertain. If 3GB GDDR7 modules become available in sufficient volume in the second half of 2026, a late-year announcement would be possible — though the cards themselves might not reach retail until early 2027. If module supply remains constrained, CES 2027 in January becomes the earliest credible window. Either way, a buyer waiting for the Super refresh is facing at minimum six months of additional scarcity and price pressure, with no guarantee the resulting cards will be affordable given the same memory market conditions that delayed them.

The practical recommendation for anyone who needs a GPU in 2026 has not changed significantly since January: the RTX 5060 Ti 8GB and RTX 5060 8GB remain available at or near MSRP and represent the most accessible entry points into the Blackwell architecture. The RTX 5070 with 12GB VRAM has maintained relatively consistent availability at elevated but not extreme prices and is the highest-VRAM model that NVIDIA reportedly prioritized for continued volume allocation. For buyers who specifically require 16GB or more of VRAM in a GeForce product, the supply picture is poor and the pricing premium is real. Waiting for the Super refresh may address that gap — but it is not certain to arrive before 2027, and nothing about its pricing has been confirmed.

Frequently Asked Questions

Why are NVIDIA RTX 50 series GPU prices rising in 2026?

GDDR7 memory — the type used in all RTX 50 series graphics cards — is manufactured on the same production lines as memory used in AI data center infrastructure. As demand for AI hardware has surged, manufacturers including Samsung, SK Hynix, and Micron have directed available fab capacity toward higher-margin products, constraining the supply of memory available for consumer GPUs. NVIDIA cut shipments to its manufacturing partners by 15 to 20 percent in January 2026, reducing the number of cards reaching retail shelves and pushing prices upward. A global price analysis found average RTX 50 series prices had risen by approximately 19 percent in the three months ending February 2026.

Is the RTX 5070 Ti discontinued?

Officially, no. NVIDIA stated in January 2026 that it would continue to ship all GeForce SKUs, and ASUS walked back an initial end-of-life designation for both the RTX 5070 Ti and RTX 5060 Ti 16GB, attributing the issue to memory supply constraints rather than a production halt. In practice, the RTX 5070 Ti has been effectively absent from retail at reasonable prices since January 2026, with available units selling at significant premiums above launch MSRP. Whether NVIDIA restores meaningful supply to these models depends on whether GDDR7 availability improves — a condition that memory manufacturers have said may not resolve before late 2026 or 2027.

When will the RTX 50 Super GPU series come out?

No RTX 50 Super release date has been confirmed by NVIDIA. The company has not officially acknowledged the Super lineup exists. Leaks from multiple sources in June 2026 suggest the cards are back in active development, with the rumored lineup including a 12GB RTX 5060 Super, an 18GB RTX 5070 Super, and 24GB variants of the RTX 5070 Ti Super and RTX 5080 Super. BenchLife, a Taiwanese publication with supply chain sources, reported in June 2026 that CES 2027 — held in early January — represented the earliest realistic launch window. Other leaks have suggested a possible 2026 release, but no board partner or manufacturer has publicly confirmed either timeline.

Should I buy a graphics card now or wait for the RTX 50 Super refresh?

If you need a GPU for a current build, the RTX 5060 8GB and RTX 5060 Ti 8GB remain available near their original prices and deliver solid Blackwell-architecture performance for 1080p and 1440p gaming. The RTX 5070 is the highest-VRAM model with reasonably consistent availability. Waiting for the Super series means accepting at minimum six more months of the current price environment — and no guarantee the Super cards will launch at accessible prices given the same memory market conditions responsible for today's shortage. If your use case specifically requires 16GB or more of VRAM, the current options are limited and expensive; the Super refresh may address that gap, but its arrival is not certain before 2027.