Nvidia to raise AI server prices by more than 15% as memory supply tightens: Bloomberg

Nvidia to raise AI server prices by more than 15% as memory supply tightens: Bloomberg As Nvidia enjoys a gross margin of about 75 percent, its decision to pass those costs on underscores the growing pricing power of memory suppliers, according to Bloomberg. News Team News Team Published August 23, 2026 - 8:39 p.m. Nvidia CEO Jensen Huang, right, and SK Group Chairman Chey Tae-won attend a briefing in Seoul on June 8. AP/YONHAP Samsung Electronics and SK hynix are gaining outsize pricing power as soaring memory costs force even Nvidia — the dominant supplier of AI chips — to raise prices on its servers by more than 15 percent. But while the AI boom has handed memory makers such as Samsung Electronics and SK hynix greater l everage over their customers, that power is a double-edged sword: More expensive memory now drives up the cost of the servers and data centers that power the boom itself. “Some of Nvidia’s biggest customers have been told that the prices of servers containing its AI chips are going up more than 15 percent,” Bloomberg reported on Saturday . “The price hikes will go into effect on systems shipped early next year and will impact systems including those with the flagship Vera Rubin and Grace Blackwell chips.” Related Article SK hynix to mark groundbreaking of Indiana fab, possible Chey-Huang meeting draws attention Korea to stash chip tax windfall in new fund for AI, future tech Samsung’s cash pile raises stakes for shareholder returns after SK hynix buyback The price increase reflects a tightening supply of memory chips, a critical component of AI systems. Nvidia’s AI accelerators rely on high bandwidth memory (HBM), which stacks layers of dynamic random-access memory (DRAM) to shuttle large amounts of data to GPUs. Demand for such memory is rising as AI models become more sophisticated and require more data to run. Samsung Electronics, SK hynix and U.S.-based Micron produce most of the world’s DRAM. All three are ramping up production, but demand for HBM and server memory is outpacing supply as companies pour money into AI data centers. Contract prices for server DRAM jumped 53 to 58 percent in the second quarter of this year from the previous quarter. They are expected to rise another 13 to 18 percent in the third quarter, according to market researcher TrendForce. Nvidia CEO Jensen Huang, second from right, poses for a photo with Samsung Electronics executives at Nvidia's global AI conference in San Jose, California, on March 16. REUTERS/YONHAP The race to expand AI data centers, coupled with more long-term supply agreements by major tech companies, has further tightened the amount of memory available in the market. Price increases by Apple and Qualcomm — and now Nvidia — underscore the shift in the market. Nvidia enjoys a gross margin of about 75 percent, meaning it keeps roughly three-quarters of its revenue after accounting for the cost of its products. Its decision to pass those costs on underscores the growing pricing power of memory suppliers, Bloomberg said. Additionally, tech firms’ push to loosen Nvidia’s grip could further strengthen memory makers. Amazon, Microsoft, Google and Meta Platforms are developing their own AI chips, but those processors still need high-performance memory such as HBM. That leaves memory suppliers in an increasingly powerful position regardless of Nvidia’s dominance of the AI chip market. Analysts note that the supply-demand imbalance could work in SK hynix’s and Samsung Electronics’ favor for years to come. Advanced memory for AI is technically difficult to make, and qualifying new chips with customers and expanding production capacity take time. The more that AI infrastructure investment grows, the more that both Korean companies stand to benefit. But elsewhere in the AI ecosystem, the cost burden is beginning to mount. Some server makers have told customers that prices for Nvidia’s major AI server systems could rise by about 17 percent, according to U.S. technology publication The Information. The Nvidia logo REUTERS/YONHAP At that rate, building a 1-gigawatt data center could cost at least $5 billion or more. The AI boom is, in effect, starting to pay the price for its own success. Soaring demand has driven up memory prices, which are now pushing AI infrastructure costs higher. Nvidia is scheduled to report its second quarter earnings on Wednesday. A key question will be whether demand for AI accelerators can maintain its growth despite rising costs. Its outlook will also gauge whether major tech firms’ appetite for AI infrastructure investment can remain as strong as it has been. “If the cost of building AI infrastructure continues to rise, even major tech companies may find it difficult to maintain their current pace of investment,” said Shin Joong-ho, head of LS Securities’ research center. “If higher memory costs continue to be passed on to customers through Nvidia’s system prices, they could also squeeze returns on AI data center investments.” BY PARK YOUNG-WOO [lee.soojung1@joongang.co.kr] This article was originally written in Korean and translated by a bilingual reporter with the help of generative AI tools. It was then edited by a native English-speaking editor. All AI-assisted translations are reviewed and refined by our newsroom. industry business sk hynix tech samsung electronics nvidia