惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

A
About on SuperTechFans
博客园 - 聂微东
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
博客园 - 司徒正美
宝玉的分享
宝玉的分享
美团技术团队
量子位
The Cloudflare Blog
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
IT之家
IT之家
爱范儿
爱范儿
J
Java Code Geeks
博客园 - Franky
Last Week in AI
Last Week in AI
B
Blog
H
Hackread – Cybersecurity News, Data Breaches, AI and More
I
InfoQ
GbyAI
GbyAI
Recent Announcements
Recent Announcements
小众软件
小众软件
H
Help Net Security
Microsoft Azure Blog
Microsoft Azure Blog
MyScale Blog
MyScale Blog

Latest from Tom's Hardware in News

Analytics group signals possible delays at 40% of AI data center construction sites — companies deny schedule holdups, but satellite imagery indicates otherwise Intel hires tenured Samsung exec to lead Foundry Services — signals company focus on winning business from potential Foundry suitors Elon Musk pushing forward with Terafab at Microsoft's April patch puts Windows domain controllers into reboot loops — third known issue from KB5082063… AMD Ryzen 9 9950X3D2 appears on Amazon with $1,000 pre-order price — AMD confirms recommended pricing is still set… AMD's market cap hits all-time high, Intel hits 25-year high on Agentic AI's insatiable demand for CPUs Meta raising Quest headset prices due to AI-driven RAM shortage — Quest 3 to cost $600, Quest 3S $350 from April… Elegoo announces the Jupiter 2 resin 3D printer for $949, early bird price of $849 — new model offers massive print volume but is still physically smaller than previous models Pragmata PC performance tested: 18 GPUs take us to the Moon Chinese fabs import record volumes of US chipmaking equipment via Singapore and Malaysia — homegrown tool makers booked record 2025 revenues as price competition squeezes margins US appeals court restarts $3 billion patent infringement lawsuit against Intel — VLSI case from 2017 returns after… Two US citizens get combined 16 years in prison for running North Korean laptop farms — fake remote IT work scheme netted DPRK $5 million in around three years Intel launches Wildcat Lake as Core Series 3 for value laptops and edge systems — six consumer SKUs built on 18A promise Legendary Qualcomm, Apple, and Nuvia alumni form new CPU startup — Nuvacore promises to Bambu updates its 3D printers to print unique hues or gradients using two or three filaments — company acknowledges OrcaSlicer-FullSpectrum fork as the basis for the color prediction part of the new feature Broadcom to supply Meta with custom silicon through 2029 — Broadom CEO Hock Tan departs Meta Anonymous perps behind 86 million files scraped from Spotify hit with $322 million court judgement — Anna Non-functioning counterfeit Samsung 990 Pro SSDs are circulating in Europe — Despite convincing packaging, blue… Oklahoma farmer arrested and jailed for trespassing during AI data center town hall — removed by officers after going a few seconds over allotted speaking time, trying to hand paperwork to counselors Virginia voter support for new data centers collapses from 69% in 2023 to 35% in new poll — Multi-gigawatt, 37-building Digital Gateway project abandoned Struggling shoemaker and apparel brand Albird pivots to AI data centers, stock jumps 580% in a single day — sells core business and leveraging $50 million in financing to become a GPU-as-a-Service and AI cloud solutions provider IPv6 usage reaches historic 50% across Google services, matching IPv4 — increased usage eases pressure on the IPv4 address market as 'new' protocol designed in 1998 finally hits its stride Engineer open-sources DIY radar system that's 95% cheaper than $250,000 commercial offerings, has 20 kilometer range — Moroccan engineer designs Aeris-10 radar, shares it on GitHub Elon Musk demonstrates first sample of Tesla AI5 processor, accidentally thanks TSC rather than TSMC  — claims 40X performance boost over the predecessor Valve might be adding a 30-day price tracker to Steam — feature is already available in some EU countries to spoof… Memory cards and flash drives prices rocket 124%, some products peak at 261% jump — increases from 2025 driven by AI chip shortage across a range of formats and capacities Netgear secures conditional approval from the FCC following router ban — company can continue importing foreign-made routers through October 2027 China tests deep-sea electro-hydrostatic actuator that can cut undersea cables at a depth of 3,500 meters — state hails successful trial and hints at deployment readiness Iran reportedly bought an in-orbit Chinese satellite to target US military sites in the Middle East — purchase agreement included ongoing ground control services based in China Our lifestyle tech colleagues at Tom's Guide have overhauled their site for smarter shopping — more video and access to experts make it 'the biggest relaunch in our history'
Nvidia's memory costs soar 485%, latest AI systems now co...
Anton Shilov · 2026-05-21 · via Latest from Tom's Hardware in News
Nvidia
(Image credit: Nvidia)

As prices of components are increasing rapidly due to high demand from the AI sector, the cost of these machines is also increasing significantly. Morgan Stanley Research estimates that a next-generation Vera Rubin-based VR200 NVL72 rack will cost major hyperscale cloud service providers (CSPs) around $7.8 million per unit (via @Aaronwei3n), which is tangibly more than about $4 million per GB300 NVL72. Furthermore, because every VR200 NVL72 rack packs plenty of DRAM and NAND, memory now accounts for around 25% of the total cost.

Nvidia plans to charge $55,000 per Rubin GPU and $5,000 per Vera CPU when selling them in volume inside VR200 NVL72 chassis to hyperscalers, according to Morgan Stanley. Although the upcoming VR200 NVL72 racks use the already familiar Oberon chassis, they use more sophisticated switching, networking, printed circuit board (PCB), cooling, power supply, and even chip packaging components, which increases bill-of-material (BOM) costs and eventually the price of the systems. As a result, each VR200 NVL72 will cost hyperscalers around $7.8 million, according to Morgan Stanley, which is higher than around $7 million we were told by one of our sources in late March. Meanwhile, the cost of memory within a VR200 NVL72 rack will be about $2 million, up 435% from the memory cost in GB300 NVL72, according to the same figures.

Sheesh.$NVDA VR200 Bom Analysis from MS. pic.twitter.com/sutjttSkyWMay 21, 2026

There are several reasons why the cost of memory is expected to account for 25% of the cost of a VR200 NVL72 system and why the system carries $2 million worth of memory.

First up, each of such racks now contains 54 TB of LPDDR5X memory, up from 17 TB of LPDDR5X in the case of a GB200 NVL72, a threefold increase. SemiAnalysis estimates that Nvidia paid $8 per GB per GB of LPDDR5X in Q1, though that price may increase as demand rises in the coming quarters, especially if we are talking about SOCAMM2 modules that are expensive to make and test. In any case, even at $8 per GB, each GB200 NVL72 machine carries $136,000 worth of LPDDR5X memory, whereas each VR200 NVL72 system will contain $408,000 worth of LPDDR5X content. If the price rises to $10, we are talking about $540,000 for LPDDR5X alone. Note that even $10 per GB may be an underestimate* as Nvidia adds its own markup.

Secondly, each VR200 NVL72 rack carries about $1 million or more of 3D NAND storage, up from virtually zero inside GB200 NVL72.

As a result, $2 million of memory content per Vera Rubin NVL72 rack is not something completely unexpected: the system uses a lot of LPDDR5X and 3D NAND memory (not to mention HBM4 memory onboard of Rubin GPUs), and memory now comes at massive prices.

*Contract price of DDR5 memory is now between $12 and $16 per GB, depending on various factors and luck, according to Framework. Spot price for DDR5 was about $20 per GB on average at press time, according to DRAMeXchange. LPDDR5X is more expensive than DDR5. When installed on SOCAMM2 modules (which are exclusively used by Nvidia's Vera CPUs), it will get even more costly, especially when Nvidia's markup is added.

Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.

Google Preferred Source

Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.

Anton Shilov is a contributing writer at Tom’s Hardware. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends.