Nvidia is implementing a price increase of up to 17% on its AI server portfolio, driven by escalating costs for advanced packaging, memory modules, and liquid cooling components.
Nvidia will increase prices on its latest high-performance artificial intelligence server racks—the “Grace Blackwell” and “Vera Rubin” lines—by up to 17% early next year. The adjustment, driven by surging memory chip costs, has been communicated to major cloud providers including Microsoft, Google, and Amazon Web Services (AWS). This marks the extension of recent pricing increases into the enterprise sector, following three separate hikes to Nvidia’s consumer graphics cards earlier this year.
According to Bloomberg, Reuters, and The Information, supply prices for rack-level systems like the Grace Blackwell 300 and Vera Rubin 200 will rise by 17%. The Information reported that some customers received direct notifications from system manufacturers regarding the adjustments. Prior to the increase, the Grace Blackwell 300 was priced between $3.7 million and $4 million per rack, while its successor, the Vera Rubin 200, averaged around $7.8 million. With the hike, the Vera Rubin 200 is now projected to exceed $8 million per unit.
If the increases are finalized, constructing a 1-gigawatt (GW) data center could see costs climb by at least $5 billion (approximately 6.94 trillion won), according to The Information. A facility of this scale represents a massive campus requiring dedicated power generation and hundreds of thousands of Nvidia’s rack-level AI servers.
The Grace Blackwell 300 and Vera Rubin 200 are ultra-high-performance AI server architectures deployed across large-scale data centers. Each system integrates graphics processing units (GPUs), central processing units (CPUs), and NVLink interconnect technology. Their overall performance is heavily dependent on the high-bandwidth memory (HBM) modules bundled within them. Consequently, ongoing supply constraints among the three primary HBM producers—Samsung Electronics, SK hynix, and Micron Technology—are continuing to exert upward pressure on pricing.
The memory chip shortage is simultaneously driving price increases across consumer electronics reliant on general-purpose semiconductors. Amazon has raised prices on several devices, including the U.S. market-leading Echo smart speaker and Fire TV streaming products. The entry-level Echo Dot saw the steepest jump, increasing 60% from $49.99 to $79.99. The Echo Show 11 rose from $219.99 to $249.99, while the Kindle Paperwhite increased from $159.99 to $199.99. Additionally, both the Fire TV Stick HD and the eero 7 wireless mesh networking device were raised by $30 each, bringing their prices to $100. Similar adjustments have been implemented by Dell for its January PC lineup, alongside broader price increases from Google and Apple on key products.
Amid surging demand for semiconductor back-end production, including packaging and testing, major Taiwanese manufacturers are aggressively scaling operations. An analysis published on August 23 by Taiwan’s Economic Daily News examined the capital expenditure plans of five leading back-end firms—ASE Holdings, PTI, KYEC, Sigurd, and ChipMOS—and found their combined capacity-expansion investments exceed NT$460 billion (approximately 20 trillion won). ASE Holdings dominates the spending, accounting for 70% of the total. The company has already raised its investment allocation twice this year, increasing its annual capex from an initial level of approximately $7 billion to $10.5 billion.