Technology newsroom
Nvidia to Significantly Raise AI Server Prices by Over 15% Due to Unstoppable Memory Price Surge! 🚀
Nvidia has reportedly informed major customers of an over 15% price increase for its new AI server systems, Grace Blackwell and Vera Rubin, which are set to begin shipping in early 2027. The primary reason cited is the continuous surge in HBM memory costs.
📌 Key Highlights:
- Nvidia has notified major customers that prices for new AI server models will increase by more than 15% in many cases.
- This price increase will affect Grace Blackwell and Vera Rubin systems, scheduled for delivery starting in early 2027.
- The primary reason is the massive surge in memory costs (both DRAM and HBM), a phenomenon dubbed RAMageddon.
Bloomberg has reported that Nvidia has informed some of its largest customers about price adjustments for its AI chip-powered servers, which could increase by over 15%. This price adjustment will apply to new systems like Grace Blackwell and Vera Rubin, scheduled for delivery starting early next year. The rate of increase will vary depending on the chip model and chosen memory specifications.
Contract server manufacturers for major data center providers such as Microsoft, Google, and Oracle have also begun notifying their customers of the upcoming price increases. This is another clear example of the crisis known as RAMageddon, which is severely impacting the DRAM market, with contract prices soaring at an unprecedented rate this year.
📈 Memory Bottleneck Issues
The main reason for the soaring prices is the massive increase in demand for High Bandwidth Memory (HBM) for AI systems. Modern AI systems require immense amounts of memory. For example, Nvidia's Rubin GPU will come with up to 288GB of HBM4 per package, and the rack-scale NVL72 system will incorporate 72 GPUs, resulting in over 20TB of HBM in a single rack. HBM production uses approximately four times more wafer space than standard DRAM, making memory one of the most expensive components in an AI server's bill of materials.
🔥 Ripple Effect
The supply shortage crisis currently driving up the prices of Nvidia's systems is a direct result of the demand Nvidia itself helped create. The three major memory manufacturers have spent this year and last year shifting production capacity to focus on HBM and high-capacity server DRAM, leading to shortages in the general consumer market. This is not the first time Nvidia has passed on costs to consumers; previously, GeForce graphics card prices were adjusted, and now the pressure has extended to the company's top-tier products.
Although Nvidia boasts a high gross margin of approximately 75%, one of the highest in the semiconductor industry, the company has chosen to pass on the increased memory costs to customers rather than absorb them itself. This is because the demand for their processing chips continues to exceed TSMC's production capacity, leaving buyers with limited negotiating power. However, this situation might push large customers to consider alternatives like AMD chips or to develop their own chips. Ultimately, though, all players still rely on the same three HBM suppliers, whose production capacity remains constrained.
💬 How do you think this Nvidia price increase will affect the AI industry and the cloud services we use in the future?