TLDR
- Nvidia customers are facing server price hikes of more than 15% due to rising memory chip costs
- The increases will apply to systems shipped early next year, including Vera Rubin and Grace Blackwell chip systems
- Samsung, SK Hynix and Micron have gained leverage as demand for AI infrastructure surges
- Nvidia reports Q2 earnings this week, with revenue expected to hit $92 billion, up 96% year-over-year
- NVDA stock dropped from $227 to around $214.75 ahead of the earnings report
Nvidia customers are being told to expect server price hikes of more than 15%, driven by surging memory chip costs. The increases will hit systems shipped early next year.
The affected systems include those built around Nvidia’s Vera Rubin and Grace Blackwell chips. The size of each price increase will depend on the chip generation and memory configuration involved.
Contract server builders have already notified major data center operators of the coming increases. Those customers include Microsoft, Google and Oracle, according to Bloomberg.
The pricing pressure comes from memory chipmakers Samsung, SK Hynix and Micron, which together control most of the world’s DRAM production. Demand for AI infrastructure has given them rare pricing power.
Nvidia’s AI accelerators are highly dependent on DRAM capacity, which makes the company especially exposed to memory cost swings.
Nvidia currently holds a gross margin of around 75% and charges tens of thousands of dollars per chip. Supply from contract manufacturer TSMC still cannot keep up with demand.
The company has also raised prices on gaming-focused PC graphics cards, according to industry site Tom’s Hardware.
Customers Building Their Own Chips
Amazon, Microsoft, Google and Meta Platforms remain heavily dependent on Nvidia for their data center buildouts, even as they each develop in-house alternatives.
Amazon is working to expand its Trainium chips, which power much of Anthropic’s AI training. Microsoft has its Maia chip. Google and Meta are developing TPU and MTIA chips respectively.
Last week, Google expanded its partnership with Marvell Technology to speed up production of its TPU chips. Other firms including Cerebras, Etched and AMD are also scaling up AI chip production.
The price hikes could push some customers to accelerate their shift toward in-house silicon, though meaningful independence from Nvidia remains a long-term project for all of them.
Earnings in Focus This Week
NVDA stock dropped from $227 to around $214.75 in recent sessions as traders positioned ahead of the Q2 earnings report due this week.
The stock has pulled back to near its 50-period Exponential Moving Average and is sitting at a key support level of $214, which was the high from July 25.
Wall Street expects Nvidia to report revenue of around $92 billion for Q2, representing 96% growth year-over-year. Q3 guidance is expected to come in near $103 billion.
Nvidia has a track record of beating forecasts, so some analysts expect the real Q2 and Q3 numbers to land closer to $96 billion and $112 billion respectively.
One potential stock catalyst beyond earnings would be an expanded share buyback program. Nvidia announced an $80 billion repurchase earlier this year, and a larger announcement could move the stock.
The stock has technical support at $214, with resistance at $227. A break below $200 would signal a shift in the current trend.
