Nvidia’s Vera Rubin Chips Enter Production as $950 Billion South Korea Deal Reshapes AI Supply Chains
Published on 07/26/2026 at 22:31 | Redaktion boerse-global.de
Jensen Huang has two messages for Nvidia investors this week: the next-generation Vera Rubin chips are already rolling off production lines, and the company is planting a $950 billion flag in South Korea’s semiconductor ecosystem. The twin announcements, delivered over the weekend, come as three of Nvidia’s biggest customers prepare to report earnings in a session that will test whether the artificial intelligence spending spree has further to run.
A $500 Billion Memory Alliance Takes Shape
The centerpiece of the Korean offensive is a partnership with the SK Group valued at more than $500 billion. SK Hynix will supply long-term memory shipments and co-develop HBM4 chips, the high-bandwidth memory modules that are becoming the bottleneck in next-generation AI systems. Building on that foundation, SK Telecom plans to construct a two-gigawatt data center by 2027 powered by Vera Rubin chips and HBM4 memory.
Nvidia is also injecting $1 billion into Naver, the Korean internet conglomerate, to expand its AI data center in Sejong. The facility, which currently runs at 55 megawatts, is slated to reach 200 megawatts by 2028 with an infusion of up to $9 billion from Brookfield. The buildout will use both Vera Rubin and Blackwell chips, with Naver targeting a long-term capacity of one gigawatt. Samsung, meanwhile, signed a $200 billion memorandum with Broadcom covering memory and foundry capacity. Huang described the moment as a “golden era” for Korea’s chip industry.
Huang Pushes Back on Delay Rumors
The CEO used a developer conference to directly refute reports from research firm SemiAnalysis that the Kyber NVL144 rack — a key component of the Rubin Ultra platform — would not be available until 2028. Huang stated that Vera Rubin is “already in production” and promised “enormous volumes” ahead. He projected combined revenue from Blackwell and Vera Rubin systems will hit roughly $1 trillion by 2027.
Should investors sell immediately? Or is it worth buying Nvidia?
The technical specifications are striking: the new Rubin GPU packs 336 billion transistors and 288 gigabytes of HBM4 memory, while the accompanying Vera CPU is designed to outperform traditional x86 processors on agentic AI workloads. During training of the DeepSeek-V3 model, a GB300 NVL72 system achieved a company-record 1,648 TFLOPs per GPU.
Memory Costs Force a Trade-Off
The boom has a sting in its tail. Rising component costs are forcing Nvidia to pare back memory on the Vera Rubin NV72 rack. SOCAMM memory is being halved from 192 to 96 gigabytes, and CPU memory per rack drops from roughly 55 to about 28 terabytes. GPU memory via HBM4 stays unchanged at 20.7 terabytes per rack. The culprit: HBM4 prices could nearly triple to $53 per gigabyte by 2027. Despite the cuts, a full rack will still cost around $9.1 million, with Nvidia capping memory costs at one-fifth of total materials.
The Bubble Question and Insider Selling
Huang addressed growing fears of an AI bubble, arguing that a near-term burst is unlikely because demand is structural rather than seasonal. He said the semiconductor industry needs to grow five- to tenfold within a decade, and that supply constraints in chips, land, energy, and talent are actually helpful because they delay overcapacity.
Yet insider activity tells a more cautious story. Over the past three months, Nvidia board members and executives sold shares worth roughly $410.6 million with no corresponding purchases. The stock closed Friday in German trading at €182.00, down 0.80 percent on the day. That leaves it about 10 percent below its 52-week high of €202.50 set in mid-May, though still roughly 10 percent above its 200-day moving average of €166.11 — a sign the medium-term uptrend remains intact despite recent consolidation.
Nvidia at a turning point? This analysis reveals what investors need to know now.
Hyperscaler Earnings in the Spotlight
This week brings earnings from Microsoft, Meta, and Amazon, three of Nvidia’s largest customers. For the chipmaker, the critical metric is capital expenditure on AI infrastructure — the number that has been the most reliable demand signal for H200 and Blackwell chips in recent quarters. Alphabet already sent positive signals in its latest update. Whether the other hyperscalers follow suit will likely determine whether the sector pushes to new highs or continues to consolidate.
Analysts at UBS see no threat to Nvidia from the shift toward open-source AI models. If anything, they argue, companies optimizing internal AI applications will demand more powerful GPUs. KeyBanc recently raised its price target to $330, while 24/7 Wall St. maintains a buy rating with a $258 target, implying roughly 27 percent upside. The RSI on a 14-day basis sits at 52.6 — neutral territory that the coming earnings reports could quickly break.
Ad
Nvidia Stock: New Analysis - 26 July
Fresh Nvidia information released. What's the impact for investors? Our latest independent report examines recent figures and market trends.
Disclaimer regarding our articles: No investment advice, no buy or sell recommendation. Information on prices, companies, and markets is provided without guarantee; changes are possible at any time. Stock market transactions can lead to substantial losses. Our articles are created and reviewed in whole or in part automatically with the support of AI.
