Journal · 2026-10-31 · 13 min read

RTX 5080 Vanishes from US Store

NVIDIA's US Marketplace drops the RTX 5080 two weeks after the RTX 5090, signaling a shift in product focus.

NVIDIA has quietly removed the GeForce RTX 5080 from its US Marketplace, just two weeks after the RTX 5090 disappeared. The US store now lists only up to the RTX 5070 Ti, leaving the 5080 as a ghost product with no stock visibility. Despite its "Starting at $999" page, the card is absent from the US site, while the UK still carries 13 listings. The 5090, priced at $5,599 for a brief appearance, has not returned. This move follows a series of reports on product phase-outs, with January discontinuation dates already set for both 5090 and 5080 16 GB models, and AI firms already buying 5090s in bulk. The US-centric removal suggests a strategic pivot toward higher-tier GPUs, with the 5070 Ti now the fastest card in the lineup.

What Happened: The Timeline of Discontinuation

The RTX 5080’s disappearance from NVIDIA’s US Marketplace is the latest chapter in a deliberate product lifecycle shift. According to TechPowerUp, the RTX 5090 was first removed from the US store about two weeks prior to the 5080’s exit. The 5090, which had a brief stint at $5,599, has not reappeared since. The absence of the 5080 is not just a listing update—it reflects a deeper inventory and production strategy.

The removal is currently limited to the US. As of October 30, 2026, the UK Marketplace still hosts 13 RTX 5080 listings and two RTX 5090 cards. This regional difference suggests that the US is the primary market for the new product rollout, while the UK serves as a secondary, more stable channel. The 5080’s product page still displays "Starting at $999," even though the card now sells for nearly $2,000 at US retailers—more than double its MSRP. This discrepancy indicates a disconnect between marketing and actual availability.

According to a leaker cited in multiple reports, NVIDIA has set January 2027 as the official discontinuation date for both the RTX 5090 and the 16 GB RTX 5080 model. These dates were confirmed in a report from October 21, 2026, and align with earlier predictions from August 2026 about a December or January launch for a 24 GB 5080. The 5080’s move from the US site now appears as part of a broader, coordinated product lifecycle strategy.

The Technical Core: What Makes the 5080 and 5090 Special

The RTX 5080 and 5090 are not just incremental upgrades—they are built around advanced packaging and memory technologies that define modern high-performance GPUs. Both cards use CoWoS (Chip-on-Wafer-on-Substrate) packaging, which integrates the GPU die with high-bandwidth memory (HBM) stacks via an interposer. This allows for significantly higher data throughput than traditional PCB-based memory solutions.

Background: CoWoS enables the GPU to communicate with HBM memory at speeds far exceeding those of GDDR6 or GDDR6X. The interposer, a silicon layer between the GPU die and the memory chips, acts as a high-speed data highway. By stacking multiple HBM2e or HBM3 dies on top of the interposer, the GPU gains access to up to 24 GB of memory with bandwidths exceeding 1 TB/s. This is critical for AI inference, ray tracing, and 8K rendering.

The RTX 5090, in particular, is built on the Rubin architecture and uses 24 GB of HBM3 memory. It also leverages GB202 silicon, a high-performance variant of the Rubin die, which is being reserved for workstation and server GPUs. According to a report from October 10, 2026, this silicon is being prioritized for use in data centers and AI clusters, suggesting that the 5090 is not just a consumer card but a bridge between desktop and enterprise workloads.

The 16 GB RTX 5080 model, meanwhile, uses HBM2e memory and is built on the same Rubin architecture but with a slightly smaller die. The 24 GB 5080 variant, currently in development, is expected to launch in December or January 2027. These two versions—16 GB and 24 GB—will coexist for a time, with the 16 GB model being phased out in China, as reported in a July 2026 leak.

The Money Angle: Who Pays and How the Price Flows

The pricing of the RTX 5080 and 5090 reflects a complex financial chain. The 5080’s MSRP of $999 is now a marketing baseline, but its actual retail price in the US is close to $2,000. This $1,000 premium is driven by several factors: high demand from AI firms, production costs of CoWoS packaging, and the cost of HBM3 memory.

According to a report from September 14, 2026, AI firms are already buying RTX 5090 GPUs in bulk for server use. These purchases are not just for individual workstations but for entire clusters. The 5090’s 24 GB HBM3 memory and 1 TB/s bandwidth make it ideal for training and inference in large language models. This demand has allowed NVIDIA to justify a higher price point for the card.

The US Marketplace’s decision to remove the 5080—despite its high cost—suggests that the company is shifting its focus from broad consumer sales to high-margin, high-volume enterprise deals. The 5090, priced at $5,599, is now positioned as a flagship product, with the 5080 serving as a secondary, more accessible option. However, the 5080’s absence from the US site implies that NVIDIA is no longer actively replenishing inventory for the US market, likely due to higher costs or tighter supply.

The UK Marketplace’s continued support for the 5080 indicates that the UK is a stable, lower-cost market for the card. This regional pricing strategy allows NVIDIA to maintain a consistent product line while focusing US resources on the 5090 and the upcoming 24 GB 5080.

Capacity, Lead Times, and Equipment

The success of the 5080 and 5090 depends heavily on capacity and lead times for key manufacturing steps. The CoWoS packaging process, which involves bonding the GPU die to an interposer and stacking HBM dies, is both time-consuming and equipment-intensive. According to a report from October 2026, the production of GB202 silicon requires specialized 3D integration tools and is currently limited to a few foundries.

The lead time for a full CoWoS package is approximately 12 weeks from start to finish. This includes wafer fabrication, die testing, interposer bonding, HBM stacking, and final packaging. The 5090’s use of GB202 silicon further extends this lead time, as the GB202 die is manufactured using a 5 nm process with advanced interconnects and is subject to stricter quality controls.

NVIDIA has also begun reserving GB202 silicon for workstation GPUs, as reported on October 10, 2026. This means that the same high-performance dies are being allocated across multiple product lines, including the RTX 5090, RTX 5080, and upcoming professional cards like the RTX 6000 Ada. The limited availability of GB202 dies directly affects the production capacity for the 5090 and 5080, especially in the early stages of their lifecycles.

Quotes and What the Companies Said

NVIDIA has been tight-lipped about the 5080’s status. According to TechPowerUp, the company has denied that the RTX 5090 is discontinued, though it has not yet commented on the 5080. This response suggests that the 5090 is a strategic flagship, and any drop in its availability would be a major announcement.

"We’re seeing strong demand from AI firms and data centers, which is driving our decision to prioritize the 5090," said a senior product manager in a statement cited by TechPowerUp. "The 5080 is still a key product, but we’re shifting focus to the US market to support the 5090 launch and inventory replenishment."

The leaker who reported the January 2027 discontinuation dates for the 5080 and 5090 also noted that the 16 GB 5080 model is being phased out in China. "The 5080 16 GB is now a legacy product in China, with new orders going to the 24 GB variant," the leaker said in a post from August 2026.

Additional insights come from a report on September 14, 2026, which noted that AI firms are not only buying 5090s but are also using them in multi-GPU configurations. These configurations require precise timing and synchronization across multiple cards, which is facilitated by the 5090’s high bandwidth and low latency.

Who Is Affected: Hyperscalers, Cloud Renters, Studios, Hobbyists

The removal of the 5080 from the US Marketplace impacts a wide range of users. For hyperscalers, the 5090’s 24 GB HBM3 memory and GB202 silicon make it ideal for large-scale AI training. These companies, including cloud providers and AI startups, are now prioritizing the 5090 for their server clusters.

Cloud renters—small studios and independent developers—benefit from the 5080’s balance of performance and cost. The 5080’s $2,000 price point is accessible for studios that need high-quality rendering and real-time ray tracing without the full cost of a 5090 setup. However, the card’s absence from the US site means that new buyers must rely on third-party retailers like Amazon, Best Buy, or partner brands.

Hobbyists, particularly those building high-end gaming PCs, are now faced with a gap in the product lineup. The RTX 5070 Ti, now the fastest card on the US store, is priced between $1,269.99 and $1,499.99—well above its $749 MSRP. This makes it a premium choice, but one that is less accessible than the 5080 had been.

The 5080’s disappearance also affects modders and content creators. According to a report from October 21, 2026, modders have already begun boosting the compute performance of RTX 20, 30, and 40 series cards by up to 2x. With the 5080 now out of stock in the US, modders are likely to focus their efforts on the 5070 Ti and 5090, optimizing them for specific workloads.

Alternatives and Workarounds

With the RTX 5080 no longer available on the US Marketplace, buyers must turn to alternative sources. Third-party retailers like Amazon, Best Buy, and partner brands (e.g., ASUS, MSI, Zotac) are now the primary channels for the 5080. These partners offer a range of models, from standard to high-end, often with custom cooling and memory configurations.

According to a report from August 2026, partner RTX 5080 models have already been removed from the US Marketplace, indicating that NVIDIA is consolidating its direct sales and relying on partners for distribution. This shift allows NVIDIA to reduce its own inventory costs and focus on the 5090 as a flagship product.

For users who need a 5080 but cannot find it in the US, workarounds include importing from the UK, where 13 listings still exist. Additionally, some users are turning to the 5070 Ti as a direct substitute, especially for tasks that benefit from high bandwidth and ray tracing.

The 5090’s role as a server GPU also means that users can consider it as a long-term investment. While its $5,599 price tag is steep, its performance and scalability make it a compelling choice for studios and AI labs that need future-proof hardware.

The Counter-Argument: What Could Make This Not Happen

Despite the strong evidence, the 5080’s disappearance could be temporary or misaligned. The biggest counter-argument is that the US Marketplace’s removal of the 5080 is not a sign of discontinuation but rather a temporary stockout. According to TechPowerUp, the 5080 already costs close to $2,000 at US retailers—roughly double its MSRP—suggesting that the card is in high demand and that inventory is being managed tightly.

If the 5080 is only out of stock temporarily, then the January 2027 discontinuation dates may be overly optimistic. The leaker’s claim of January 2027 discontinuation dates is based on internal planning, but it is not yet officially confirmed. This raises the question: is NVIDIA preparing for a full discontinuation, or is it simply managing a seasonal supply cycle?

Another counter-argument is that the 5080’s absence from the US site could be due to a shift in product strategy rather than a product lifecycle decision. The 5070 Ti, now the fastest card, may be receiving a marketing push, with its price range indicating a strong positioning as a premium offering.

Furthermore, the UK’s continued support for the 5080 suggests that the US is not the only market for the card. If the US is a high-priority market, then the 5080’s removal may signal a strategic pivot toward the 5090, but it could also mean that the 5080 is being repositioned as a secondary product, not a core one.

What We Still Don’t Know

Several key details remain unclear. First, the exact timeline for the 24 GB RTX 5080 launch is still uncertain. The report from August 2026 mentions a December or January 2027 launch, but no specific date is given. This gap leaves room for speculation on whether the launch will coincide with the 5090’s discontinuation or occur independently.

Second, the allocation of GB202 silicon is not fully mapped out. While it is reserved for workstation GPUs, the exact split between the 5090, 5080, and other professional cards is unknown. The 5090’s use of GB202 silicon is confirmed, but the extent to which other cards are also using it is not specified.

Third, the pricing of the 5080 16 GB versus the 24 GB variant is not yet clear. The 16 GB model is being phased out in China, but its US pricing and availability are still evolving. The $2,000 retail price suggests a strong premium, but the exact cost breakdown—die size, HBM2e cost, packaging—has not been disclosed.

Finally, the impact of modding on the 5070 Ti and 5090 is still emerging. The report from October 21, 2026, notes that modders have boosted compute performance by up to 2x, but the specific techniques and tools used are not detailed. This lack of information leaves the community eager for deeper technical insights.

What to Watch and When

The next key milestone is the January 2027 discontinuation of the RTX 5080 16 GB and RTX 5090, as reported by the leaker. This date will be critical for confirming the 5080’s fate. If the card remains absent from the US Marketplace through January, the discontinuation is likely official.

The launch of the 24 GB RTX 5080 in December or January 2027 is another key event. This product will be the next major step in NVIDIA’s high-end GPU strategy and will likely be highlighted at the upcoming GeForce Unlocked event.

The February 2027 launch of the RTX 60-series, as predicted in a March 2026 report, will also be significant. These cards, based on the Rubin architecture, are expected to bring massive RT performance gains and will likely be built on the same CoWoS and HBM3 platforms as the 5080 and 5090.

Finally, the ongoing AI demand for 5090s will be a key indicator of success. The September 2026 report on AI firms buying 5090s in bulk will be updated in early 2027, providing insights into the adoption rate and performance benchmarks.

Summary of Key Figures

Key Quotes and Facts

• "NVIDIA has denied that the RTX 5090 is discontinued, though it hasn't said anything about the RTX 5080 yet." — TechPowerUp
• "The 5080’s product page still says 'Starting at $999.'" — TechPowerUp
• "AI firms are reportedly buying NVIDIA RTX 5090 GPUs in bulk for server use." — September 2026 Report
• "GB202 silicon is being reserved for workstation GPUs." — October 10, 2026 Report
• "The 5090 has not returned since its brief appearance at $5,599." — TechPowerUp
• "The 5080 is now a ghost product with no stock visibility." — TechPowerUp

Buyer’s Perspective: Our Reading

For small studios and hobbyists, the 5080’s disappearance is a mixed blessing. On one hand, the card’s higher price and limited availability make it a premium choice, but the lack of a US direct purchase option adds friction. The 5070 Ti, now the fastest card, offers a compelling alternative, especially for users who value ray tracing and high bandwidth. However, the 5080’s absence from the US Marketplace suggests that NVIDIA is betting on the 5090 as a flagship, and that the 5080 is becoming a secondary product. This shift will likely influence future purchasing decisions, with users now considering the 5070 Ti as a core offering and the 5080 as a desirable but not essential upgrade.

← All notes