By 
In a dramatic reallocation of high-performance silicon, Nvidia has reportedly halted the supply of its flagship GB202 graphics processors to the consumer GeForce RTX 5090. Supply chain disclosures and board-partner notices indicate that Nvidia is redirecting all future GB202 wafer allocations exclusively to its high-margin workstation portfolio, specifically the RTX PRO 6000 and the newly revealed RTX PRO 5500.
With remaining RTX 5090 inventories expected to deplete rapidly, hardware leaks also indicate that a 24 GB variant of the RTX 5080—built on the smaller GB203 die—will step in as the top-tier consumer GeForce offering.
The Silicon Pivot: Prioritizing RTX PRO Over Consumer Gaming
The GB202 processor represents the largest monolithic graphics die in Nvidia’s "Blackwell" generation. While the consumer RTX 5090 utilizes a cut-down configuration featuring 21,760 CUDA cores and 32 GB of GDDR7 memory, the full die is deployed in enterprise and professional visualization hardware:
- RTX PRO 6000: Features the fully unlocked GB202 die with 24,064 CUDA cores paired with a massive 96 GB of GDDR7 memory, commanding enterprise price tags upwards of $16,000.
- RTX PRO 5500: Mirrors the RTX 5090’s core count of 21,760 CUDA cores but integrates 84 GB of GDDR7 memory, explicitly targeting local AI fine-tuning and heavy generative model inference.
Industry leakers MEGAsizeGPU and hongxing2020 revealed factory notices indicating an effective end-of-life (EOL) status for all RTX 5090 product lines, including regional China-specific "D" variants. Given the enormous margin disparity between retail gaming graphics cards and professional AI accelerators, directing scarce foundry capacity toward enterprise channels maximizes revenue per silicon wafer.
| Specification / Model | GeForce RTX 5090 (Consumer) | RTX PRO 5500 (Workstation) | RTX PRO 6000 (Enterprise) |
|---|---|---|---|
| Silicon Architecture | GB202-300 (Blackwell) | GB202-250 (Blackwell) | GB202-400 (Blackwell) |
| CUDA Core Count | 21,760 | 21,760 | 24,064 |
| VRAM Capacity | 32 GB GDDR7 | 84 GB GDDR7 | 96 GB GDDR7 |
| Memory Bus Width | 512-bit | 448-bit | 512-bit |
| Primary Target Market | Enthusiast Gaming / Creators | AI Development & Workstations | Data Center / Visual Computing |
| Price Point | $1,999 MSRP ($4,000+ Retail) | $8,000–$10,000 (Est.) | $15,000–$16,000 |
Escalating Retail Costs and Market Supply Pressure
The RTX 5090 was originally launched in January 2025 at an MSRP of $1,999. However, intense commercial demand from artificial intelligence startups and automated compute farms led to rampant grey-market procurement. AI startups frequently purchased bulk consumer cards directly from distributors, starving retail channels.
As a result, street prices surged dramatically throughout 2026. European retail listings routinely crossed €5,000, while major US retailers like Walmart experienced instant sell-outs at prices topping $4,299. The formal reallocation of GB202 chips will likely trigger further speculative inflation for remaining existing stock held by add-in-board (AIB) partners such as ASUS, MSI, and Gigabyte.
The Shift to RTX 5080 24 GB as the Consumer Flagship
To fill the void at the top of the consumer product stack, factory documents point to an upcoming GeForce RTX 5080 24 GB. Built on the GB203 processor, this model will replace both the original 16 GB RTX 5080 and absorb the role originally anticipated for an "RTX 5080 SUPER."
- VRAM Expansion: Upgrades the memory pool from 16 GB to 24 GB of high-speed GDDR7, alleviating memory bottlenecks for 4K ray-traced gaming and prosumer workflows.
- Extended Flagship Longevity: With rumors indicating that Nvidia's next-generation "Rubin" architecture (GeForce RTX 60-series) has slipped to 2028, the RTX 5080 24 GB is poised to remain Nvidia’s flagship consumer graphics card for an extended multi-year cycle.
Why This Matters: The Enterprise AI Preemption
From an industry economics perspective, this development underlines the structural decoupling of high-end graphics processing from traditional PC gaming:
- Margin Optimization: A single GB202 wafer yielded into enterprise RTX PRO boards generates three to four times the revenue of consumer GeForce packaging.
- AI Scarcity Realities: As commercial agentic coding and local model training demand explodes, enterprise buyers willingly absorb premium workstation price points for expanded VRAM buffers.
- Consumer Ceiling: The consumer market has demonstrated resistance to multi-thousand-dollar MSRP tiers, making mid-die options like the GB203 more sustainable for volume retail channels.
While existing channel inventory of the RTX 5090 will continue to trickle through third-party vendors for several weeks, enthusiasts should prepare for the 24 GB RTX 5080 to become the definitive ceiling of the GeForce gaming ecosystem.
Slug: nvidia-ending-rtx-5090-gb202-silicon-workstation-gpus
Frequently Asked Questions
Why is Nvidia reportedly discontinuing the RTX 5090?
Nvidia is reallocating its flagship GB202 silicon exclusively to higher-margin workstation cards like the RTX PRO 6000 and RTX PRO 5500, which command significantly higher profit margins from AI and enterprise customers.
What will replace the RTX 5090 as the top GeForce card?
An upcoming GeForce RTX 5080 24 GB (built on the GB203 chip) is expected to become Nvidia’s top-tier consumer gaming graphics card.
Will existing RTX 5090 cards still be supported?
Yes, drivers and software support for the RTX 5090 will continue through standard Game Ready and Studio driver branches.
