Nvidia and AMD Accelerators: Roadmaps, Pricing, Supply — 2026-09-02
Nvidia reported a record $96.2 billion revenue quarter, prompting CEO Jensen Huang to forecast 70% fiscal 2028 growth, while simultaneously notifying customers of >15% price hikes on Vera Rubin and Grace Blackwell systems due to soaring memory costs. Meanwhile, AMD continues its MI400 roadmap progression with 432GB HBM4 configurations targeted for late 2026 production, positioning it as a direct challenger to Nvidia's Blackwell and Rubin architectures.
Nvidia and AMD Accelerators: Roadmaps, Pricing, Supply — 2026-09-02
Top developments
Nvidia’s Record Quarter and Aggressive Price Hikes
On August 26, 2026, Nvidia reported Q2 FY2027 revenue of $96.2 billion, up 106% year-over-year, leading to a market cap increase of over $400 billion. However, this success comes with a cost increase: Nvidia has notified largest customers of price hikes exceeding 15% on AI server systems, specifically impacting Vera Rubin and Grace Blackwell configurations shipped in early 2027. This move is explicitly attributed to soaring HBM and DRAM costs, squeezing margins despite strong demand.
AMD MI400 Roadmap and Helios Platform Progress
AMD is advancing its Instinct MI400 series, which features the CDNA-Next architecture and 432GB of HBM4 memory, with production shipments planned for H2 2026. The MI400 is projected to deliver up to 40 petaflops FP4 and 19.6TB/s bandwidth, aiming to compete directly with Nvidia’s Blackwell and upcoming Rubin chips. Recent reports indicate the Helios platform, designed to challenge Nvidia in data centers, is seeing increased traction as customers seek alternatives to Nvidia's rising prices and supply constraints.
AWS Orders Additional 2 Million GPUs Amid Supply Tightness
Amazon Web Services (AWS) has reportedly ordered an additional two million GPUs from Nvidia, signaling sustained hyperscaler demand despite price increases. This massive order aligns with reports that Nvidia’s gross margin guidance was cut to 71-72% due to memory costs, highlighting the tension between volume growth and component inflation. The order underscores the critical role of supply allocation, where major cloud providers secure priority access to Blackwell and Rubin systems, potentially delaying availability for smaller enterprises and neoclouds.
Nvidia’s Strategic Shift Beyond Chips
Nvidia is increasingly emphasizing software and networking efficiency over raw chip cycles, with new data center systems leveraging smarter traffic control to enhance performance. This strategic pivot aims to maintain its competitive moat against AMD’s hardware-focused improvements and custom silicon from hyperscalers like Microsoft Maia 300. By optimizing system-level efficiency, Nvidia seeks to justify its premium pricing and lock in customers who rely on its full-stack ecosystem rather than just GPU compute power.
Local view
South Korean media outlets are closely tracking the impact of Nvidia's price hikes on local tech giants and global supply chains. Seoul Economic Daily reported that Nvidia will raise prices on Grace Blackwell 300 and Vera Rubin 200 racks by up to 17%, affecting key customers like Microsoft, Google, and AWS. The report highlights that memory semiconductor price surges are the primary driver, creating pressure on Korean memory manufacturers like Samsung and SK Hynix to balance profitability with customer relationships. Additionally, Newspim noted that Nvidia agreed to supply Amazon with an additional 2 million GPUs, reflecting the intense competition for limited high-end accelerator supply.
Context & numbers
- Nvidia Revenue: $96.2 billion in Q2 FY2027, up 106% YoY.
- Price Hikes: >15% increase on AI servers (Vera Rubin, Grace Blackwell) effective early 2027.
- GPU Pricing: B200 street prices remain firm near $40,000 per unit, with rental rates at $4.50–$7.00/hr.
- AMD MI400 Specs: 432GB HBM4, 40 petaflops FP4, 19.6TB/s bandwidth.
- Market Reaction: Nvidia shares added >$400 billion in value post-earnings.
On the radar
- Rubin Delivery: Monitor for initial Vera Rubin system deliveries to hyperscalers in early 2027, which will test the market's tolerance for the announced price hikes.
- AMD MI450 Sampling: Watch for updates on AMD’s MI450 sampling progress, which precedes full production shipments and could shift supply dynamics in H2 2026.
- Memory Cost Trends: Continued volatility in HBM and DRAM contract prices remains the key variable affecting Nvidia’s margin guidance and final retail pricing for AI accelerators.
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
