rubin

According to Money Today, the South Korean government has budgeted 27.5 billion Korean won per rack for NVIDIA's next-generation AI chip Vera Rubin, which is higher than previous estimates by foreign media. The Ministry of Science and ICT's budget for GPU server expansion next year is 3.85 trillion Korean won, all allocated for the procurement of 140 Vera Rubin NVL72 server racks. Each NVL72 rack is equipped with 72 Rubin GPUs and 36 Vera CPUs, totaling 10,080 Rubin GPUs. This batch is independent of the 52,000 advanced GPUs announced by the ministry to be acquired by 2028, and the actual supply may be adjusted according to NVIDIA's delivery situation.The budget of 27.5 billion Korean won per rack is significantly higher than foreign media predictions. U.S. investment bank Bernstein's June report estimated the cost of a Vera Rubin NVL72 rack at 9.1 million dollars (approximately 12.3 billion Korean won), while Morgan Stanley estimated it at about 7.8 million dollars (approximately 10.5 billion Korean won). The government stated that the budget has taken into account potential price increases due to GPU shortages and the costs of supporting facilities. This batch of chips will be used for the development of next-generation cutting-edge models such as general artificial intelligence. Deputy Prime Minister and Minister of Science and ICT Lee Kyung-sook mentioned in June that creating a model on par with Anthropic's large model Mythos would require about 10,000 Vera Rubins, and this scale of 140 units is comparable.

NVIDIA's Chief Financial Officer expects Vera Rubin to contribute approximately 20% of data center business revenue in the third quarter.

NVIDIA stated that it will ship both Blackwell and Rubin architecture systems in the future, and is currently facing certain supply constraints.

NVIDIA (NVDA.O) announced the full production of Vera Rubin.

Indian AI infrastructure company AM Intelligence has ordered 9,000 sets of NVIDIA Vera Rubin systems, planning to deploy them in southern India next year. The company's clients include large cloud service providers, AI laboratories, and institutions developing local AI models.

SemiAnalysis reported that the Nvidia Rubin Ultra will feature 192GB of HBM4 8-hi memory. This is not only a significant reduction from the initially previewed 1TB HBM but is even lower than the 288GB capacity of the standard Rubin.The reasons for the HBM configuration being cut down to about one-fifth of the original include: the cancellation of the four-die Rubin Ultra scheme, reducing the number of cubes to 8; the stacking height being lowered from 16-hi to 8-hi; and the use of HBM4 with 24Gb dies instead of HBM4E with 32Gb dies. An upgraded version with 8-hi HBM4E may be released later.

Elon Musk stated that the Nvidia Vera Rubin chip platform optimized for space will reach significant scale by 2028. The Vera Rubin NVL72 system is set to launch into orbit in the fourth quarter of next year.

NVIDIA officially announced that SpaceXAI will deploy the NVIDIA Vera CPU to accelerate its next-generation agent-based AI applications. Vera is the first CPU specifically built for AI agents, featuring 88 NVIDIA self-developed Olympus cores, utilizing spatial multithreading technology and LPDDR5X high-bandwidth memory, with a bandwidth of up to 1.2TB/s. In scenarios such as agent-based AI, reinforcement learning, and data processing, the task completion speed can be improved by up to 1.8 times compared to x86 CPUs.Mike Nicolls, President of SpaceXAI, stated that the CPU performance and memory bandwidth of Vera are sufficient to support large-scale task orchestration, code execution, and data processing, allowing GPUs to operate at peak efficiency, thus gaining more effective output per watt of computation.SpaceXAI also plans to expand the AI infrastructure behind Grok based on the NVIDIA Vera Rubin platform, moving towards gigawatt-level computing capacity. Vera Rubin integrates NVIDIA accelerated computing, NVLink interconnect, Spectrum-X networking, and BlueField data processing technologies, aiming to reduce the cost per token and improve the overall energy efficiency of AI factories.More notably, SpaceXAI will send an optimized version of the Vera Rubin NVL72 system into space, with its first-generation Starmind AI satellite adopting this rack-level architecture, becoming the first iconic deployment of NVIDIA accelerated computing extending from ground data centers to orbital computing.

Market news indicates that Hon Hai will begin shipping NVIDIA's Vera Rubin platform in the fourth quarter.

According to UBS's analysis of the Vera Rubin superchip Bill of Materials (BOM), the share of memory semiconductors in the total cost of the platform has risen to about 62%, a significant increase from the 53% of the previous generation Blackwell Ultra (GB300). The total cost of the Vera Rubin superchip (which includes 1 Vera CPU and 2 Rubin GPUs) is approximately $38,902, with memory-related costs reaching $24,297. The cost of HBM4 is about $4,943 (accounting for 12.7% of the total cost), while the cost of the SOCAMM2 connecting the CPU is about $19,355 (accounting for 49.8% of the total cost), making it the core driver of rising memory costs.SOCAMM2 is a server memory module based on low-power DRAM LPDDR5X, supporting a maximum capacity of 1.5TB per CPU. The Vera Rubin NVL72 cabinet (which includes 72 Rubin GPUs and 36 Vera CPUs) is equipped with a total of 74.7TB DRAM, of which HBM4 accounts for 20.7TB, and LPDDR5X for the CPUs reaches 54TB, with the LPDDR usage in a single cabinet equivalent to about 4,500 high-end smartphones. However, due to uncertainties in the supply of LPDDR5X and HBM4E, NVIDIA is considering reducing the HBM configuration for the next generation Rubin Ultra.

According to The Information, Nvidia is weighing a significant adjustment measure, considering using less high-bandwidth memory for the Rubin Ultra GPU to address the shortage of advanced high-bandwidth memory chips.

According to Jinshi reports, Musk stated that the Starmind AI satellite will be optimized for the Vera Rubin computer.

According to the latest research by TrendForce, the tight supply of DRAM is expected to continue until 2027, and the uncertainty of the HBM4e verification schedule is prompting AI chip manufacturers to consider reducing the HBM configuration of their next-generation products. Starting from Q3 2026, NVIDIA has begun evaluating the HBM solution for Rubin Ultra, expanding from the original 12-Hi HBM4e to alternative options such as 8-Hi HBM4e, 12-Hi HBM4, and 8-Hi HBM4, with the final specifications yet to be determined. Previously, due to the expected supply bottleneck of LPDDR5X lasting until 2027, NVIDIA decided to halve the SOCAMM capacity of the next-generation Vera Rubin Superchip module.TrendForce pointed out that the main reasons for the downgrade of Rubin Ultra from 12-Hi HBM4e include: the overall DRAM shortage in 2027 will limit the allocation of wafer capacity to HBM, and there is still uncertainty regarding the mass production schedule and yield ramp-up of 12-Hi HBM4e. It is expected that the HBM bit shipment volume in 2027 will increase by approximately 50-60% year-on-year, which will still be insufficient to meet the growth in demand. HBM suppliers will maintain pricing power throughout 2027, and HBM prices are expected to rise significantly. The final configuration will depend on the wafer allocation decisions of memory suppliers.

Citrini analyst Jukan posted on platform X that he received news from an expert conference call stating that the release date of Nvidia's next-generation Rubin architecture products may be delayed again, to the point where he "can't even take a break on Saturday."However, as of now, Nvidia has not made a formal response to the rumors regarding the delay of the Rubin project. Rubin is seen as Nvidia's new generation AI chip architecture following Blackwell, and its research and production progress has been highly monitored by the market. The above news remains an unverified market rumor, and the specific release date will need to wait for Nvidia's official disclosure.

The semiconductor research organization SemiAnalysis published an article stating that Nvidia is expected to experience a significant ramp-up in production capacity in the second half of this year (i.e., the second half of fiscal year 2027), with its data center computing business revenue projected to exceed current market consensus expectations by 20%.The organization noted that the issues previously delaying Nvidia's next-generation Rubin architecture chips due to technical bottlenecks with HBM4 have now been resolved, and initial reserves of front-end wafer supply have been completed, with large-scale mass production about to begin.In addition, SemiAnalysis emphasized that its forecast data is based on in-depth research across the entire industry chain, covering dynamics from upstream materials, wafer foundries, components, and server integrators, to downstream hyperscale cloud service providers and cutting-edge AI laboratories. Its tracking model also includes major AI chip ecosystem participants such as Broadcom, AMD, MediaTek, and Marvell.

NVIDIA (NVDA.O) announced that the highly anticipated new "Rubin" data center product will be launched this year, allowing customers to trial the technology and accelerate the development of artificial intelligence. NVIDIA stated that six Rubin chips have been delivered from manufacturing partners and have passed several key tests, indicating that customer deployments are progressing as planned. The newly launched Rubin accelerator improves training performance by 3.5 times compared to the previous generation Blackwell, and speeds up the running of AI software by 5 times.The company emphasized that the operating costs of Rubin-based systems will be lower than those of Blackwell-based systems, as they achieve the same results with fewer components. It is reported that Microsoft and other large remote computing service providers will be the first to adopt this new hardware in the second half of this year. (Jin Shi)

ChainCatcher news, according to Bloomberg, Ethereum co-founder and Sharplink Gaming Inc. Chairman Joe Lubin stated in an interview, "Sharplink is not using leverage, and we may take some measures to bring some leverage to the company, such as convertible equity. We will issue long-term bonds at low interest rates. We will not put this strategy at risk."Previously, Sharplink recently completed a $425 million private placement, purchasing Ether and transforming into an Ether financial company. However, after registering shares for a group of investors, its stock price plummeted by 69%.

ChainCatcher news, according to Securities Times, Nvidia founder and CEO Jensen Huang delivered a keynote speech at the 2024 Taipei International Computer Show, stating that Nvidia's Blackwell chips have now begun production. Nvidia will launch the Blackwell Ultra AI chip in 2025. The next-generation Rubin platform is under development and will be released in 2026, with the Rubin AI platform adopting HBM4 memory chips.