Nvidia and AMD Accelerators: Roadmaps, Pricing, Supply — 2026-09-17
AMD has overtaken Nvidia in consumer GPU sales share at key retailers as RTX 5090 prices surge to $9,500, driven by a GDDR7 shortage and AI demand. Meanwhile, LG CNS confirmed a massive $280 million investment in Nvidia's next-generation Vera Rubin accelerators, signaling continued hyperscaler spending despite rising hardware costs. Reports indicate both AMD and Nvidia may delay their next-generation consumer GPU launches to 2028 to prioritize data center production.
Nvidia and AMD Accelerators: Roadmaps, Pricing, Supply — 2026-09-17
Top developments
AMD Overtakes Nvidia in Retail GPU Sales Amid RTX 5090 Price Surge
In September 2026, AMD captured 56% of GPU sales at major retailer Mindfactory, surpassing Nvidia for the first time in recent months. This shift is attributed to the Nvidia RTX 5090 hitting a market price of $9,500 due to a severe GDDR7 memory shortage and high AI-related demand. The price spike highlights the tension between consumer availability and data center allocation, as Nvidia prioritizes higher-margin AI chips.
LG CNS Invests 381.4 Billion KRW in Nvidia Vera Rubin Accelerators
LG CNS announced on September 17, 2026, that it will invest 381.4 billion KRW (approx. $280 million) to secure Nvidia's next-generation Vera Rubin AI accelerators. This move aims to bolster LG Group's "AI Factory" capabilities and expand its GPU-as-a-Service offerings. The substantial commitment underscores the continued dominance of Nvidia's roadmap in enterprise AI infrastructure, even as competitors like AMD gain traction in specific segments.
Next-Gen Consumer GPUs Delayed to 2028 for Both AMD and Nvidia
Reports from September 17, 2026, indicate that both AMD and Nvidia plan to postpone their next-generation consumer GPU launches (RDNA 5 and RTX 60 series equivalents) to 2028. Only one mid-range RDNA 5 card is expected in 2027, suggesting a strategic pivot to maximize data center revenue through Blackwell, Rubin, and Instinct MI400 lines. This delay impacts the consumer upgrade cycle and keeps prices elevated for current-gen cards.
Nebius Raises GPU Rental Prices by Up to 21%
Cloud provider Nebius announced plans to increase GPU rental prices by up to 21%, citing supply constraints and increased demand for AI computing resources. This move reflects broader industry trends where supply-side pricing power is strengthening, potentially impacting the total cost of ownership for AI startups and enterprises relying on cloud-based acceleration.
Local view
Korea: ZDNet Korea highlights LG CNS's massive Vera Rubin investment as a critical step for national AI competitiveness, noting the strategic importance of securing next-gen hardware early. Global Economic News reports on the "scarcity economy" of the RTX 5090, with black market prices in Korea soaring to over 12.8 million KRW ($9,500+), driven by US retail stockouts.
China: Sina Finance notes that Nvidia B200 GPUs are retaining value exceptionally well, with residual values 58% higher than initial launch prices, indicating strong secondary market demand despite new product cycles. Chinese media also report on Sharp and Foxconn partnering to launch AI servers using Nvidia RTX PRO 6000 Blackwell GPUs, targeting Japanese government and enterprise sectors.
Context & numbers
- Retail Prices: Nvidia RTX 5090 listed at ~$9,500; RTX 5060 Ti 16GB up 39% since June 2026.
- Data Center Costs: A single Vera Rubin chip line item accounts for ~70% of the $4 million cost of Nvidia's newest rack-scale supercomputer.
- Lead Times: B200 GPU lead times remain at 6–16 weeks for allocated OEM orders.
- Market Share: AMD holds ~56% of Mindfactory's GPU sales unit share in Sept 2026, reversing previous Nvidia dominance in that channel.
On the radar
- Vera Rubin Deployment: Monitor for actual shipment volumes of Vera Rubin systems following LG CNS's order and other enterprise commitments.
- GDDR7 Supply Chain: Watch for resolution of the GDDR7 shortage which is currently inflating consumer GPU prices and aiding AMD's retail sales share.
- AI Energy Alliance: Nvidia and Google's new "AI Energy Management Alliance" may influence future data center power procurement strategies and site selection.
This content was collected, curated, and summarized entirely by AI — including how and what to gather. It may contain inaccuracies. Crew does not guarantee the accuracy of any information presented here. Always verify facts on your own before acting on them. Crew assumes no legal liability for any consequences arising from reliance on this content.
Powered by
