TL;DR: NVIDIA has reportedly paused all RTX 50 series production until at least Q3 2026. The anticipated RTX 50 SUPER refresh is canceled. The RTX 60 "Rubin" series is delayed to late 2027 at the earliest. The root cause — a global DRAM memory shortage driven by AI datacenter demand — shows no signs of easing. This is the most significant disruption to consumer GPU availability in NVIDIA's 30+ year history.
---
The Escalation
Six weeks ago, coverage of the RTX 5090 shortage began with supply constraints and scalper pricing. Four weeks ago, RTX 5090 pricing hit $5,000. Two weeks ago, RTX 5070 Ti EOL rumors signaled the mid-range was in trouble.
Now the situation has escalated beyond individual SKU shortages. According to reports from TweakTown, Tom's Guide, and Tom's Hardware, NVIDIA has effectively paused production of the entire RTX 50 series lineup. Multiple AIB partners have confirmed production halts on specific models.
This isn't a supply constraint. It's a production shutdown.
The Numbers
The scale of this disruption is worth quantifying:
| Metric | Current Status |
|---|---|
| RTX 5090 street price | $6,000+ (190% above $1,999 MSRP) |
| RTX 5090 price increase (3 months) | +79% |
| RTX 5080 price increase (3 months) | +35% |
| Production cut (GeForce line) | 30-40% initially, now effectively paused |
| RTX 5090 major retailer availability (US) | Zero in-stock at Amazon, Best Buy, Newegg, Micro Center |
| New gaming GPU releases planned for 2026 | Zero |
According to Tom's Hardware, this marks the first year in three decades that NVIDIA will not release a new gaming GPU.
What's Dead, What's Delayed
Confirmed or Highly Likely
- RTX 50 SUPER refresh: Canceled. NVIDIA did not announce any new GPUs at CES 2026 — a first in five years — and reports indicate the SUPER series will not launch in 2026
- RTX 5070 Ti 16GB: Production halted by ASUS and likely other AIB partners, per WCCFTech
- RTX 5060 Ti 16GB: Production ended to focus on 8GB variant
- RTX 5060 8GB: Production reportedly halted
Delayed
- RTX 60 "Rubin" series: Originally expected 2026-2027, now pushed to late 2027 or 2028, creating a nearly three-year gap between major GPU generations
NVIDIA's Official Position
NVIDIA maintains that all RTX 50 series GPUs will "continue to ship" and that Founders Edition cards have not been discontinued. But as Tom's Guide notes, "stock and supply tell a different story." Saying products will "continue to ship" is technically compatible with shipping very small quantities. The FE cards were delisted from NVIDIA's site because they were out of stock.
The Root Cause: DRAM Memory Crisis
This isn't a temporary supply hiccup. It's a structural shift in global memory allocation.
According to TrendForce, AI will consume 20% of total DRAM wafer capacity in 2026. HBM consumes roughly 4x the wafer capacity per gigabyte compared to standard DRAM. GDDR7 requires 1.7x the capacity of standard DRAM.
The math is simple: AI companies pay 500%+ premiums for memory priority. SK hynix's advanced packaging lines are at capacity through 2026. Samsung and Micron face identical constraints. Consumer GPUs that need GDDR7 memory are competing for wafer capacity against datacenter GPUs that need HBM3 — and the datacenter GPUs pay far more.
As covered in The AI Memory Crisis, this is a problem that gets worse before it gets better. IDC's analysis projects the shortage extending through 2026-2027.
What This Means for AI Hardware Buyers
Consumer/Prosumer GPU Buyers
The window to buy RTX 50 series cards at reasonable prices has closed. Remaining strategies from 4 Ways to Get an RTX 5090 at MSRP still apply, but availability is tighter than when that was published:
- Prebuilt systems remain the most reliable path to RTX 5090 cards — integrators like Bizon and EmpoweredPC received direct GPU allocations
- European markets still have better availability than the US
- RTX 4090 and 4080 SUPER cards, while previous-gen, are increasingly attractive as RTX 50 prices diverge further from MSRP
Enterprise AI Buyers
Enterprise GPU supply (H100, H200, L40S) is a separate allocation from consumer GeForce, but the same memory crisis affects both. The takeaway:
- Lock in orders now if you're planning 2026 infrastructure buildout — lead times will extend
- Consider alternatives: AMD MI300X and Intel Gaudi 3 face their own supply constraints, but diversifying vendor risk is prudent. See 3 NVIDIA Alternatives for Enterprise AI
- Cloud GPU pricing will likely increase as on-premise supply tightens — compare current rates before they adjust
The Silver Lining
If there is one: this shortage is forcing the market to mature. AMD's ROCm ecosystem improves with every enterprise customer that can't get H100s. Intel's Gaudi 3 is gaining deployment traction. Tenstorrent's open hardware approach looks more strategic when the dominant vendor can't ship product.
Competition in AI accelerators has been theoretical for years. A sustained NVIDIA supply crisis makes it practical.
Timeline: What to Expect
| Period | Expected Development |
|---|---|
| Q1-Q2 2026 | RTX 50 production remains paused; existing inventory depletes |
| Q3 2026 | Possible production resumption at reduced volumes |
| Q4 2026 | Earliest window for meaningful RTX 50 restocking |
| H2 2027 | RTX 60 "Rubin" earliest possible launch (optimistic) |
| 2028 | More realistic RTX 60 launch window |
None of these dates are confirmed. The timeline depends entirely on DRAM supply recovery, which depends on AI infrastructure buildout pace — which shows no signs of slowing.
---
Previous coverage:
