nvidia

ASUS and Gigabyte, among other graphics card manufacturers, plan to raise product prices in the mainland Chinese market again in September, by up to 500 yuan (approximately 2,300 New Taiwan dollars). The price increase effect will also boost the performance of ASUS, Gigabyte, MSI, Leadtek, and Chengqi.According to reports from mainland Chinese media IT Home, ASUS will raise the net cost price of the NVIDIA RTX 5070 series by 150 to 300 yuan in September, while prices for other series models will remain unchanged. Gigabyte will increase the price of all models in the NVIDIA RTX 5060TI 16G series by 200 yuan in September, while other models will remain the same. AMD's RX 9070 XT series will increase by 200 yuan; the RX 9060 XT 16G/8G series will increase by 200 to 250 yuan; and the RX 7650 GRE series will increase by 400 yuan.The main reason for the price increase is the shortage of memory, and production capacity has also shifted towards manufacturing AI computing power GPUs. NVIDIA has again lowered the overall shipment volume of consumer GPUs by 15% to 20% for the third quarter, and the estimated supply for various brands in September is similar to that in August, with supply still tight. The high-end models such as the NVIDIA RTX 5080 and 5070 TI remain among the most out-of-stock models. This year, NVIDIA has raised graphics card prices multiple times, and September marks another increase.

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 CEO Jensen Huang stated that GPT-6 Astra was trained using over 100,000 NVIDIA Grace Blackwell NVLink72 chips. It took only 4 years from ChatGPT to o1, and then to Astra. AGI has arrived. Congratulations to the OpenAI team.

According to Reuters, AI data center optical interconnect company iPronics announced the completion of a $125 million Series B financing round, co-led by Maverick Silicon and Light Street Capital, with participation from NVIDIA, Triatomic Capital, Bosch Ventures, Catalight Capital, and the European Innovation Council Fund. The company's total financing has now reached $177 million.iPronics primarily provides rack-mounted optical switching equipment for AI infrastructure, allowing AI clusters to dynamically adjust network connections in real-time for training and inference workloads through a programmable optical layer. The new funds will be used to expand operational scale and accelerate commercial deployment.

According to a report by Bloomberg on September 4, the artificial intelligence company DeepSeek plans to deploy at least 160,000 top accelerators from Huawei Technologies Co. in a large data center being built in Inner Mongolia, potentially forming one of the largest known clusters of Huawei AI chips and advancing China's efforts to replace Nvidia.DeepSeek intends to use Huawei's next-generation Ascend 950DT chips to run its models, with the installation timeline depending on Huawei's production capacity. Sources say that the company currently does not plan to use the 950DT chips for training, although Huawei has designed and marketed these processors for more demanding training functions. DeepSeek had previously attempted to train models using Huawei chips but has so far relied on Nvidia accelerators to complete this critical step.

JPMorgan believes that the core issue currently facing Nvidia is not a lack of demand, but rather supply constraints limiting its growth potential. Nvidia's management recently informed JPMorgan that if supply constraints can be lifted, the company's business scale could even double year-on-year.JPMorgan analyst Harlan Sur maintained an "Overweight" rating on Nvidia after meeting with the Nvidia investor relations team and kept the target price at $320. Nvidia's management expects the company's business to grow by about 70% year-on-year in fiscal year 2028, but Sur believes that, based on the current customer demand situation, the actual potential growth capacity may be even stronger. Meanwhile, Nvidia's visibility into future demand is also improving. Sur stated that the reason the company's management provided an early growth outlook for fiscal year 2028 is due to a significant gap between their internal demand assessment and Wall Street forecasts. Releasing this expectation early also helps suppliers and other partners plan capacity in advance for higher future demand levels.

According to Bloomberg, after NVIDIA acquired the AI platform Hugging Face for about $13 billion, the three French co-founders—38-year-old Clément Delangue, 42-year-old Julien Chaumond, and 41-year-old Thomas Wolf—each saw their net worth rise to about $1.8 billion, marking their debut on the Bloomberg Billionaires Index.The deal is expected to be completed in the first half of 2027, and the three co-founders will continue to serve under a six-year retention agreement. NVIDIA has committed that Hugging Face will remain an open platform, allowing users to freely upload and download AI models and datasets. Hugging Face was founded in 2016, initially starting in Paris and now headquartered in Brooklyn.

According to Electronic News on September 3, Micron plans to double its high bandwidth memory (HBM) production capacity compared to last year, with a maximum addition of 60,000 wafers per month by the end of this year. Insiders say that Micron's HBM output was about 40,000 to 50,000 wafers per month last year, and it is expected to reach about 100,000 wafers per month by the end of the year, narrowing the gap with Samsung Electronics and SK Hynix.Micron is increasing procurement orders from multiple HBM equipment suppliers, with production bases mainly in Taiwan and Singapore, continuously introducing equipment. Currently, the output is mainly HBM3E 12-layer, with HBM4 12-layer accounting for about 20% to 30% at the beginning of this year, which may rise to as high as 50% by the end of the year. HBM4 12-layer will be used in NVIDIA's latest AI accelerator, Vera Rubin, and Micron has started mass production of this product since the second quarter of this year.Micron CEO Mehrotra stated in the Q3 earnings call for fiscal year 2026 in June this year that the ramp-up speed of HBM4 12-layer mass production is about twice that of HBM3E 12-layer, and cumulative revenue from HBM4 shipments has exceeded $1 billion. The industry believes that Samsung and SK Hynix each have HBM production capacities of about 150,000 to 200,000 wafers per month. Counterpoint data shows that in Q2 of this year, the global HBM market shares were SK Hynix 50%, Samsung 32%, and Micron 18%. Micron is also preparing related investments in Hiroshima, Japan.

NVIDIA announced that the RTX Spark devices equipped with the N1X chip will begin shipping in October this year. The N1X supports up to 128GB of unified memory, and the Blackwell GPU provides up to 10 petaflops of floating-point performance per second. The full version is equipped with a 6144 CUDA core Blackwell GPU and a 20-core Grace CPU, supporting 24GB to 128GB of unified memory; another version has 5120 GPU cores and 18 CPU cores, supporting only 24GB to 32GB of unified memory.NVIDIA launched the open-source tool NVIDIA PAIR, which can connect multiple RTX devices, DGX Spark, and even Apple devices in a home network to schedule idle computing power for collaborative processing of AI agent tasks. Compatible devices include computers with NVIDIA graphics cards (RTX 20 series and later, RTX Pro GPU, DGX Spark) as well as devices with Apple M4 or newer chips. PAIR prioritizes the use of idle computing power and automatically adjusts as devices join or leave the network.In a media briefing example, a household had approximately 165 TFLOPS of underutilized computing power. The Qianwen 3.6 35B A3B model completed agent tasks in an average of 18 minutes on a single Spark notebook, while a three-device PAIR cluster averaged 8 minutes and 48 seconds. The AI agent applications Perplexity Portable Computer, Hermes Agent, and OpenClaw will receive a more simplified local deployment method.

NVIDIA (NVDA.O) has agreed to acquire Hugging Face for $12.9303 billion.Hugging Face will continue to operate as an open platform for the entire AI ecosystem. The equity retention plan for Hugging Face employees joining NVIDIA is up to $1 billion.

Nikkei Chinese reported on September 3 that demand for AI servers from Nvidia is driving the semiconductor memory DRAM market to accelerate upward. The price of general DRAM products used in personal computers and other devices increased by 5% month-on-month, while products for industrial equipment saw double-digit increases. Memory manufacturers are prioritizing the supply of AI server-specific memory, and its ripple effect is further spreading.In addition to being used in personal computers, cameras, and smartphones, DRAM is also installed in AI servers and other devices in data centers, serving as temporary data storage. As sellers, memory manufacturers negotiate monthly or quarterly with equipment manufacturers and module manufacturers, who act as buyers, to determine bulk transaction prices. The bulk transaction price for the 16Gb specification of the general model DDR5 used in PCs, game consoles, and servers was set at approximately $40.2 in July, an increase of 5% month-on-month. Prices have risen for seven consecutive months, and executives from electronics trading companies stated that they have reached a historic high.

On Wednesday, local time in the United States, NVIDIA CEO Jensen Huang appeared at the G20 Innovation Ministerial Meeting and engaged in a fireside chat with U.S. Secretary of Commerce Gina Raimondo. Huang stated that AI is evolving into a national economic infrastructure similar to electricity and the internet, and the biggest risk countries face is not sufficiently investing in and adopting AI, ultimately being left behind by the industrial revolution.Huang mentioned that currently, building 1 gigawatt of AI infrastructure requires an investment of about $50 billion to $60 billion, and he expects that from now until the end of this decade, the scale of related construction will reach approximately 100 gigawatts. He also referenced the "five-layer cake" model of energy, chips, infrastructure, models, and applications, indicating that every country needs to build AI infrastructure and decide which aspects they wish to participate in.Huang anticipates that in the coming years, AI will essentially achieve AGI, and the next phase will transition from large language models to intelligent agents and embodied intelligence. He believes that AI is more likely to replace tasks rather than completely replace jobs, with operational tasks such as writing and information processing potentially becoming gradually automated.

According to GMGN market data, the Meme coin microduck, based on the open-source robot from Hugging Face, briefly surpassed a market capitalization of 28 million USD, setting a new high, with a 24-hour trading volume of 16.7 million USD and a 24-hour increase of 17%.This Meme coin is inspired by the real open-source mini robot "Microduck" (approximately 399 USD for a 25cm duck-shaped RL robot) released by Hugging Face on August 27, 2026, and is also combined with rumors of NVIDIA-related acquisitions. It is completely unrelated to Hugging Face, Pollen Robotics, or NVIDIA; it is merely an independent Meme token launched by the community leveraging the trend.

According to CoinDesk, AI company Anthropic has reached a $35 billion computing power procurement agreement with Lambda, an AI cloud service provider supported by NVIDIA, part of which will be provided through Bitcoin mining company Hut 8's Beacon Point facility located in Nueces County, Texas.Hut 8 previously disclosed that the facility has signed two long-term leases for 15 years, covering 704 MW of IT capacity, with a total contract value of $19.6 billion, but the identity of the tenants had not been disclosed earlier. The facility spans 525 acres and has a maximum power access capacity of 1 GW with ready grid connections.According to the currently disclosed transaction structure, NVIDIA holds the lease for the facility, Lambda will deploy NVIDIA chips, and Anthropic will procure the computing power generated from this. Hut 8 has not confirmed how much of the 704 MW capacity is related to Lambda and Anthropic, nor has it confirmed whether NVIDIA is the previously undisclosed tenant.As the demand for power and computing resources in AI data centers surges, Bitcoin mining companies with ready power and grid access capabilities are accelerating their transformation into AI infrastructure. Mining companies including Hut 8, TeraWulf, IREN, and Bitdeer have all entered the AI data center market by selling or leasing computing power and power infrastructure.Hut 8's revenue for the second quarter of this year was $74.93 million, a year-on-year increase of 81%; its two leases at Beacon Point have a total value of $19.6 billion, and the facility is expected to generate approximately $1.31 billion in operating profit annually once fully operational.

Venture capitalist Tomasz Tunguz posted on X that the cutting-edge AI market is splitting into closed camps, with labs choosing partners, cutting off competitors, and rationing access to the strongest models. Salesforce selected Anthropic as its dedicated AI partner, making Claude the default model for the world's largest CRM and Slack.OpenAI cut off its API access on November 12 after SpaceX acquired Cursor. Z.ai placed its flagship model under a security review after cumulative revenue exceeded $10 billion. Anthropic rationed Mythos 5 through Project Glasswing and restricted Fable to inference only in the United States. OpenAI first launched the GPT-5.6 variant to trusted partners.Nvidia is investing in an open ecosystem, including $13 billion for Hugging Face, $7 billion for Poolside, and a $26 billion commitment to Nemotron. Tunguz stated that access, rather than price, has become the new scarcity at the forefront.

MediaTek announced on the 31st that it has completed the pricing of $3.9 billion (nearly NT$125 billion) in overseas convertible bonds, bringing in funds from two AI giants, Nvidia and Google's parent company Alphabet. Nvidia accounted for $3.5 billion (about NT$110 billion), approximately 90% of the total amount, and announced that the two companies will deepen their cooperation in the future. This is the largest overseas convertible bond issuance in the history of Taiwan's capital market and also Nvidia's first investment in a Taiwanese company.According to a joint statement from Nvidia and MediaTek, Nvidia will purchase MediaTek's overseas convertible bonds. This investment expands the cooperation between the two parties, and MediaTek will use Nvidia's NVLink Fusion and Nvidia's recently released NVHBM technology to facilitate smoother communication of components in data centers. Nvidia's CEO Jensen Huang stated that the cooperation with MediaTek has "expanded significantly," making it one of the top partnerships; MediaTek's processors will become the foundation for the next generation of Windows PCs. MediaTek stated that this overseas fundraising attracted enthusiastic participation and subscriptions from long-term important partners such as Nvidia and Alphabet, as well as renowned international investors.

NVIDIA (NVDA.O) will invest $3.5 billion in MediaTek, further deepening the cooperation between the two parties. Currently, NVIDIA is working to persuade more companies to develop chips that can connect to its dominant data center ecosystem. The two companies stated in a joint announcement that NVIDIA will purchase bonds that can be converted into MediaTek stock. This investment will expand the collaboration between the two chip designers.According to the cooperation arrangement, MediaTek will adopt NVLink Fusion and the newly announced NVHBM technology, which are part of NVIDIA's technology suite aimed at facilitating smoother communication among various components in data centers. MediaTek is trying to challenge the market positions of Broadcom (AVGO.O) and Marvell Technology (MRVL.O) by helping data center operators create customized components.Previously, Amazon (AMZN.O) also announced a similar cooperation agreement, agreeing to deploy an additional 2 million NVIDIA components. More importantly, Amazon also committed to using NVIDIA's connectivity technology in its self-developed chips.

According to South Korean media reports, Samsung Electronics is cooperating with Nvidia (NVDA.O) to develop 8-layer seventh-generation high bandwidth memory (HBM4E), significantly reducing the stacking height compared to the original plans of 12 layers and 16 layers. The speed specification proposed by Nvidia is 17-18Gbps, which is about 20% higher than Samsung's initial HBM4E sample speed of 14.4Gbps. This product is reported to be used for NVHBM.

The Taiwanese prosecutors stated that they are investigating the global leading chip substrate and printed circuit board manufacturer, Unimicron Technology Corporation, which is a key supplier for NVIDIA, Intel, Google, and Amazon. The investigation targets the company for allegedly mislabeling parts made in China as being produced in Taiwan, which may violate laws prohibiting the laundering of origin.Unimicron Technology Corporation is one of the world's top printed circuit board manufacturers, and this investigation occurs against the backdrop of the United States cracking down on transshipment.