NVIDIA has told some of its largest customers that server prices will rise by more than 15% in many casesCITE:E1CITE:E5. The increase will apply to systems shipping early next year, covering products built around NVIDIA's flagship Vera Rubin and Grace Blackwell chipsCITE:E2CITE:E6. The exact size of each increase will depend on the chip generation involved and the specific memory configuration of the systemCITE:E2CITE:E6.
Contract manufacturers that assemble servers for Microsoft, Google and Oracle have already notified their customers of the coming increase, while NVIDIA has declined to comment on the matterCITE:E3.
Samsung, SK Hynix and Micron together account for most of the world's memory production, yet even as these companies keep raising output they still cannot keep up with sharply higher demandCITE:E4. SK Hynix said in October last year that it had already sold out its entire 2026 memory production capacity, and Samsung and SK Hynix raised their 2026 HBM3E supply prices by close to 20% before that year even beganCITE:E8.
The price pressure spans everything from bulk DRAM contracts to retail memory kits.
| Product | Metric | Value |
|---|---|---|
| Conventional DRAM contracts | Q1 2026 quarter-over-quarter change | +90% to +95% |
| Conventional DRAM contracts | Q2 2026 projected quarter-over-quarter change | +58% to +63% |
| HBM3E supply price | Increase before 2026 began | ~20% |
| Consumer 32GB DDR5-6000 kit | Price a year earlier vs. August 2026 | $110–$140 → ~$392 |
Analysts projected that conventional DRAM contract prices would climb 58% to 63% quarter-over-quarter in the second quarter of 2026, following a first-quarter surge of 90% to 95%, as suppliers reallocated capacity toward HBM and server productsCITE:E7. On the consumer side, DDR5 pricing has more than doubled since late 2025, with a mainstream 32GB DDR5-6000 kit selling for around $392 in August against $110 to $140 a year earlierCITE:E10.
NVIDIA's Rubin GPU ships with up to 288GB of HBM4 per package, and its NVL72 rack-scale system combines 72 of those GPUsCITE:E9. That configuration puts more than 20TB of HBM into a single rack, before even accounting for the LPDDR memory attached to the system's Vera CPUsCITE:E9.
NVIDIA runs a non-GAAP gross margin of roughly 75%, among the highest in the semiconductor industryCITE:E11. The reported price increases indicate that the company intends to pass memory cost inflation on to customers rather than absorb itCITE:E11.
The chain of evidence lines up end to end: memory suppliers Samsung, SK Hynix and Micron are unable to keep pace with demandCITE:E4, SK Hynix sold out its 2026 capacity and HBM3E prices rose close to 20% before the year startedCITE:E8, DRAM contract prices are tracking increases of 90–95% and 58–63% in successive quartersCITE:E7, and consumer DDR5 kits have already jumped from $110–$140 to about $392CITE:E10. Against that backdrop, NVIDIA's more-than-15% server price increaseCITE:E1CITE:E5 is being layered on top of a business already running a roughly 75% non-GAAP gross marginCITE:E11, while the contract manufacturers building the affected systems have moved to notify customers even as NVIDIA itself has stayed silentCITE:E3.
NVIDIA has told some of its largest customers that server prices will rise by more than 15% in many casesCITE:E1CITE:E5.
Contract manufacturers that assemble servers for Microsoft, Google and Oracle have already notified their customers of the coming increase, while NVIDIA has dec…
Samsung, SK Hynix and Micron together account for most of the world's memory production, yet even as these companies keep raising output they still cannot keep …
The price pressure spans everything from bulk DRAM contracts to retail memory kits.
The numbers in this story point to a choice, not just a pass-through. NVIDIA's roughly 75% non-GAAP gross margin sits well above the cost pressure it cites, and choosing to add more than 15% to server prices rather than absorb any of the DRAM and HBM3E increases suggests the margin cushion is being preserved, not spent. The Rubin generation compounds the exposure: at up to 288GB of HBM4 per GPU and 72 GPUs per NVL72 rack, each system carries over 20TB of memory, so any further supply tightening from SK Hynix, Samsung or Micron flows straight into the next round of pricing. The metric worth watching is whether NVIDIA's non-GAAP gross margin holds near 75% once these price increases take effect on Grace Blackwell and Vera Rubin shipments next year — if it does, the increase was about protecting margin; if it slips anyway, the memory market is tighter than even a 15%+ hike can offset.
Japan's Ministry of Economy, Trade and Industry plans to request an additional ¥150 billion (~NT$30 billion) for Rapidus in next year's budget. If approved, it would be the government's third capital injection, lifting cumulative state capital in the company to ¥400 billion, while total government support is projected to reach roughly ¥3 trillion by March 2028.
The US government is preparing to push the Dutch government to bar ASML from selling almost all chipmaking equipment, including DUV lithography systems, to China. A bipartisan bill in Congress, the MATCH Act, would also ban maintenance on DUV machines already delivered to China and threatens sanctions on the Netherlands if it refuses to comply, while Chinese toolmaker Shanghai Yuliangsheng Technology has begun limited domestic DUV production.
NVIDIA reports fiscal Q2 earnings on August 26, with analysts projecting record revenue of $92 billion, above the $78 billion initial-year estimate. After 14 straight quarters of beating forecasts, analysts say NVIDIA needs 95% profit growth and $51.5 billion in net income to clear the bar again, while options price a 5.3% post-earnings swing and its shares have fallen after each of the last four reports.