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

Blog — PlanetScale
Blog — PlanetScale
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC
Vercel News
Vercel News
B
Blog
腾讯CDC
P
Proofpoint News Feed
Google DeepMind News
Google DeepMind News
N
Netflix TechBlog - Medium
L
LangChain Blog
F
Fortinet All Blogs
T
The Blog of Author Tim Ferriss
人人都是产品经理
人人都是产品经理
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
I
InfoQ
IT之家
IT之家
酷 壳 – CoolShell
酷 壳 – CoolShell
aimingoo的专栏
aimingoo的专栏
D
DataBreaches.Net
Stack Overflow Blog
Stack Overflow Blog
The Cloudflare Blog
Last Week in AI
Last Week in AI
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 - 三生石上(FineUI控件)
T
Tailwind CSS Blog

Latest from Tom's Hardware

Our experts review your astonishing PC builds and setups in Rig Rundown — from wall-mounted setups to a system… News outlets are blocking Wayback Machine from archiving their pages — 23 outlets concerned AI companies might abuse fair use and use it to train their models Mark Zuckerberg reportedly working on AI clone of himself — Meta insiders claim 3D photoreal animated Zuck will be able to engage with employees on his behalf Score a massive $700 off this 4K-ready Lenovo gaming PC with an RTX 5070 Ti, now just $1,899 — epic Legion Tower 5i pre-built ships with a 20-core Intel CPU, 32GB DDR5 and a 2TB SSD Pay $1,349.99 for Gigabyte's Aero X16 laptop and save $300 on this 32GB beast with RTX 5070 graphics —… Veteran Windows dev shows off AI running on 47-year-old PDP11 with 6 MHz CPU and 64KB of RAM — 'gloriously absurd' project runs transformer model written in PDP-11 assembly language Half of all US employees now use artificial intelligence at work, crossing landmark threshold for first time — Gallup data shows daily and weekly usage hitting all-time high of 28% in Q1 2026, with 65% feeling positive about its impact on productivity China has spent 3.6 times more than the US on chipmaking subsidies over the past decade — $142 billion and counting, easily outweighs CHIPS Act FAA approves military use of drone-killing laser weapons in US airspace — decision comes after it was decided ‘systems do not present an increased risk to the flying public’ Nvidia says AI cuts 10-month, eight-engineer GPU design task to overnight job — company is still 'a long way' from AI designing chips without human input Small Missouri town ousts half its city council after $6 billion AI data center approval — petition calls for mayor's removal as frustration (and violence) over AI data centers mounts New tech can see a CPU's transistors in action — terahertz radiation can potentially steal data as a chip is… Intel's Nova Lake CPUs gear up to seize AMD’s 3D V-Cache gaming throne — early leak points to up to 52 cores, blazing DDR5-8000 support, and massive 175W TDP Acer Predator GX850 SFX power supply review: Solid electrical performance with good efficiency NZXT to cough up $3.45 million over 'predatory' Flex PC rental scheme in RICO class-action settlement — in-debt customers to get up to $5,000 of relief, eligible renters to be granted ownership Bulbous 15x fan PC case side panel dubbed the ‘Superdome’ lowers temps by 20 degrees —  $600 worth of Noctua fans arrayed in 3D-printed structure Approvals for Nvidia and AMD AI chip exports to China stall under government bottleneck —  20% staff turnover… Espresso Lite 15 Review: An entry-level portable monitor with a splash of color Save a massive $700 on this 4K-ready HP gaming PC with a 9800X3D and RTX 5070 Ti, now just $2,499 — discounted HP Omen 35L pre-built powerhouse ships with 32GB DDR5 RAM and a 1TB SSD 'CopprLink' destroys every eGPU standard in new test, achieves near-native-level performance with an RTX 5090 — setup requires $2,300 worth of additional hardware Website backup crippled by 1.6MB Friends GIF that was replicated 246,173 times, breaking Linux's EXT4 filesystem limit — Jennifer Aniston's 'happy dance' animation ate up 377 gigabytes of data due to security policy Why we spent 50+ hours retesting Intel’s Core Ultra 270K Plus and 250K Plus Just $284.99 for 32GB of Team T-Create Classic DDR5-6000 RAM is the cheapest going right now — this double-dipping… Grab MSI’s RTX 5080 gaming laptop for just over $2,000 — offers fast 240 Hz QHD+ display, dual storage slots, and expandable DDR5 memory Lenovo hikes Legion Go 2 handheld gaming PC to almost $3,000 for 2 TB model — Handheld now costs more than AMD's Strix Halo devices despite relatively weaker Z2 Extreme chip Iran's forced nationwide internet blackout becomes second-longest on record as it passes 1,000 hours offline — possessing Starlink terminals punishable by death, country using 'military-grade jamming' against service Tiny 3-inch cube PCs bring a splash of color to the passive PC market with red, orange, green and blue options — Intel Twin Lake-powered Kubb Mini PCs start at $500 Veteran Microsoft engineer says original Task Manager was only 80KB so it could run smoothly on 90s computers — original utility used a smart technique to determine whether it was the only running instance Tech enthusiast gets Doom to run on a 40-year-old printer controller — ancient Agfa Compugraphic 9000PS came with a Motorola 68020 onboard for fast processing Keychron Q6 Ultra 8K Review: 660 hours of battery life at 8 KHz
Nvidia's memory costs soar 485%, latest AI systems now co...
Anton Shilov · 2026-05-21 · via Latest from Tom's Hardware
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.