TL;DR: Four weeks into the RTX 50-series launch, prices continue climbing. RTX 5090 cards now routinely sell above $4,000, some approaching $5,000. NVIDIA has reportedly shifted production priority to RTX 5060 series, and the underlying memory crisis may persist until late 2027. Here's what's changed since our initial coverage.
---
What's New This Week
Since our original coverage on January 22nd, several developments have escalated the situation:
Prices Keep Climbing
| Card | Week 1 Price | Week 4 Price | Change |
|---|---|---|---|
| RTX 5090 | $3,000-3,500 | $3,500-5,000 | +25-40% |
| RTX 5080 | $1,200-1,400 | $1,400-1,600 | +15-20% |
| RTX 5070 Ti | $900-1,000 | $1,000-1,200 | +15-20% |
At Newegg, RTX 5090 listings have pushed past $4,000, with some models approaching $5,000. Best Buy's pricing shows similar trends, with the ASUS Prime RTX 5080 at $1,599 - $600 over MSRP.
NVIDIA's Production Shift
Reports from China indicate NVIDIA has adjusted its RTX 50-series supply plan:
- Prioritized: RTX 5060 and RTX 5060 Ti 8GB
- Reduced: RTX 5060 Ti 16GB and RTX 5070 Ti shipments
- Rationale: Lower VRAM models are easier to produce given memory constraints
This explains the RTX 5070 Ti EOL speculation - NVIDIA may not be discontinuing it, but production volumes are likely minimal.
Memory Crisis Timeline Extended
Industry analysts now project the GDDR7/HBM shortage may persist until late 2027 or early 2028. Contributing factors:
- AI datacenter demand: Hyperscalers consuming vast majority of HBM production
- GDDR7 ramp: New process with lower initial yields
- Capacity limits: SK Hynix and Samsung can't expand fast enough
European Market Paralysis
A German parts seller shared distributor communications revealing they cannot sell RTX 5090, 5080, or 5070 Ti at all. RTX 5070 orders are capped at 5 units per model. The European market appears even more constrained than North America.
---
Updated Timeline
Based on current data, here's the revised outlook:
| Period | RTX 5090/5080 | RTX 5070/5070 Ti | RTX 5060 |
|---|---|---|---|
| Feb 2026 | Severe shortage | Severe shortage | Limited availability |
| Mar-Apr 2026 | Very constrained | Very constrained | Improving |
| May-Jun 2026 | Constrained | Gradual improvement | Normal |
| Jul-Aug 2026 | Meaningful improvement | Near normal | Normal |
| Sep+ 2026 | Normalizing | Normal | Normal |
The RTX 5060 series will likely see normal availability first, followed by 5070, then 5080/5090.
---
What This Means for AI Buyers
For Those Who Can Wait
If your current hardware is adequate, waiting until Q3 2026 makes sense. By then:
- RTX 5090 should approach MSRP
- RTX 5080 should be available at reasonable premiums
- Used RTX 4090 prices may drop as supply normalizes
For Those Who Need Hardware Now
The calculus hasn't changed much, but our previous recommendations are reinforced:
- RTX 4090: Still the best value for 24GB VRAM
- RTX 4080 Super: Best value for 16GB VRAM
- RTX 6000 Ada / L40S: 48GB professional options outside consumer shortage
- Cloud GPU: Immediate access, no hardware hassles
RTX 5090 at Current Prices?
Paying $4,000+ for an RTX 5090 only makes sense if:
- You specifically need 32GB VRAM that 4090's 24GB can't provide
- The time value of having hardware now exceeds the $2,000 premium
- You're building for production work where professional cards are more appropriate anyway
For most AI developers, the RTX 4090 at $1,800-2,200 delivers 90% of the practical utility at 50% of the current RTX 5090 price.
---
Available RTX 5090 Options
If you do decide to buy at current prices, here's what's in stock:
- MSI RTX 5090 Gaming Trio - Viperatech
- NVIDIA RTX 5090 Founders Edition - Viperatech
- ASUS ROG Astral RTX 5090 White - Central Computer
- ASUS TUF Gaming RTX 5090 - Central Computer
Check current prices before purchasing - they change daily.
---
The Broader Context
This shortage isn't just a consumer GPU problem. NVIDIA's Vera Rubin platform announcement at CES revealed a 40% production cut expectation due to memory shortages affecting datacenter GPUs too.
The entire AI hardware stack - from consumer cards to datacenter accelerators - is constrained by the same memory supply bottleneck. This systemic issue explains why the shortage won't resolve quickly.
---
What We're Watching
- NVIDIA earnings call (Feb 2026): Official guidance on supply trajectory
- Samsung GDDR7 ramp: Second major supplier could ease constraints
- AMD Radeon RX 9000 launch: Alternative if AMD can secure memory supply
- Used market prices: RTX 4090 secondary market as indicator
---
Related:
- The RTX 5090 Shortage: What AI Hardware Buyers Need to Know (Original coverage)
- The AI Memory Crisis: Why Your Next GPU Server Will Cost More
- 5 GPUs You Can Actually Buy for AI in 2026
---
