Nvidia and AMD Accelerators: Roadmaps, Pricing, Supply — 2026-09-09
Nvidia's recent earnings revealed a strategic pivot toward a full-stack AI infrastructure platform, with non-hyperscaler customers now growing revenue at 138% year-over-year. Meanwhile, supply chain pressures persist as Nvidia plans to raise prices on Vera Rubin and Grace Blackwell systems by over 15% in early 2027 due to soaring HBM memory costs. AMD continues to challenge Nvidia in enterprise evaluations, with non-Nvidia accelerators ranking significantly higher in some enterprise assessments despite Nvidia's dominant market share.
Nvidia and AMD Accelerators: Roadmaps, Pricing, Supply — 2026-09-09
Nvidia redefines itself as AI infrastructure platform
In early September 2026, financial analyses highlighted that Nvidia is no longer just a chip company but an infrastructure platform for all of AI. CEO Jensen Huang emphasized this shift, noting that the company's value lies in its full-stack capabilities rather than just silicon. This distinction is reshaping where billions in hyperscaler spending land, as five major companies compete for the same supply-constrained foundation. This narrative supports Nvidia's Q2 FY27 results, which showed 106% year-over-year revenue growth, reinforcing its position beyond simple component sales.
Non-hyperscaler demand rivals Big Tech
Data from late August and early September 2026 indicates that Nvidia's non-hyperscaler customers are now rivaling Big Tech in revenue contribution, growing 138% year-over-year. This trend suggests that the AI buildout is no longer dependent solely on the largest cloud providers, diversifying Nvidia's revenue streams and reducing reliance on hyperscaler capex cycles. This shift provides a buffer against potential spending slowdowns at companies like Microsoft or Amazon, though it also introduces new supply allocation challenges for smaller enterprises.
Enterprise evaluations favor non-Nvidia accelerators
Despite Nvidia's market dominance, recent industry reports from September 3, 2026, indicate that enterprises rank non-NVIDIA accelerators 14 points ahead of Blackwell on evaluation lists. This sentiment, coupled with TSMC's inability to match AI demand despite fivefold capacity expansions, highlights a growing willingness among businesses to consider alternatives like AMD Instinct MI350 and MI400 series. For AMD, this represents a critical opportunity to capture market share in sectors where supply constraints have made Nvidia's lead times prohibitive.
Graphics card prices to rise again amid supply shifts
On September 9, 2026, reports emerged that graphics card manufacturers plan to increase prices in China again at the end of this month, which may affect global prices. Nvidia has reportedly cut its shipment forecast by 15% to 20%, while AMD also faces pressure as more GPUs are allocated to the more profitable AI sector. This reallocation is expected to worsen shortages for consumer-grade products, potentially driving up retail prices for RTX 50-series and Radeon cards globally.
Local view
South Korean media outlets are closely watching the supply chain implications for local tech giants. Reports from Money Today (Aug 23) noted that while Microsoft has already received its priority allocation of Nvidia's Vera Rubin chips, questions remain about when Korean companies will secure their volume. The focus is on whether local firms can compete with US hyperscalers for the next generation of "Vera Rubin NVL72" systems, which combine 36 "Vera" CPUs with 72 "Rubin" GPUs.
Context & numbers
- Nvidia Q2 FY27 Revenue: $96.2 billion, up 18% quarter-over-quarter and 106% year-over-year.
- Price Hikes: Nvidia plans to raise prices on Vera Rubin and Grace Blackwell systems by >15% for shipments starting early 2027, driven by HBM costs.
- Blackwell Pricing: Current B200 cloud rates span $5.99-$16.11/hr, while DGX B300 systems cost $300K-$350K.
- RTX 5090 Retail: Prices have surged to ~$5,000 due to AI demand squeezing supply.
On the radar
- China Market Entry: Rumors persist about Nvidia supplying China-specific AI chips before year-end, though these remain unconfirmed by official channels.
- AMD MI400 Production: AMD's MI450 sampling is underway, with production shipments planned for H2 2026, potentially offering a viable alternative as Nvidia lead times stretch.
- ASML Adoption: TSMC and Samsung have committed to using ASML's latest lithography machines, a critical step for producing the large data center chips designed by Nvidia and AMD.
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
