chip

Amazon AWS today announced the widespread availability of Trn2 instances based on its in-house team-developed AI training chip, Trainium2, with the launch of the Trn2 UltraServer large-scale AI training system, as well as the release of the next generation of the even more advanced 3nm-processed Trainium3 chip. A single Trn2 instance consists of 16 Trainium2 chips interconnected by an ultra-fast, high-bandwidth, low-latency NeuronLink interconnect, providing 20.8 petaflo...

Bloomberg yesterday (November 25) released a blog post, reported that Amazon in order to reduce dependence on Nvidia, has launched a "moon landing" (moonshot) plan, hoping to deploy 100,000 second-generation self-developed chips, improve data processing efficiency and reduce the cost of AI chip procurement. 1AI cited sources reported that Amazon set up an engineering laboratory in Austin, Texas, and set up a specialized team of engineers dedicated to the development of AI chips. The source said the lab atmosphere is similar to a startup, with engineers focusing on circuit board and cooling system improvements to optimize un...

Silicon Valley startup d-Matrix on Tuesday announced the start of shipments of its first artificial intelligence chip aimed at providing services such as chatbots and video generation. d-Matrix has raised more than $160 million (currently about Rs 1,158 million) in funding so far, including an investment from Microsoft's venture capital arm. Early customers are testing sample chips and full shipments are expected next year, the company said. Santa Clara, California-based d-Matrix did not name specific customers, but said Super Micro Co.

Nov. 4 (Bloomberg) -- NVIDIA CEO Jen-Hsun Huang has asked SK Hynix to supply next-generation, high-bandwidth memory chips known as HBM4 six months ahead of schedule, SK Group Chairman Choi Tae-won said, according to a Reuters report today. SK Hynix plans to launch the first HBM4 products with 12-layer DRAM stacks in the second half of 2025, with 16-layer stacked HBMs launching slightly later in 2026. SK Hynix and TSMC signed a Memorandum of Understanding (MOU) in April this year, announcing that they will collaborate on HBM in...

Microsoft's official Azure X account posted last night to "show off" its new gear: the company has gotten its hands on an AI server powered by NVIDIA's GB200 superchip, making it the first of the world's cloud service providers to use the Blackwell architecture. In March, NVIDIA unveiled its most powerful AI acceleration card, the GB200, at the GTC 2024 developer conference, which utilizes the next-generation AI graphics processor architecture, Blackwell, etched using TSMC's 4-nanometer (4NP) process. bla...

Foxconn Senior Vice President Benjamin Ting announced at the 2024 Hon Hai Technology Day that Foxconn plans to build the world's largest NVIDIA GB200 chip manufacturing plant in Mexico, although he did not disclose exactly where the facility will be built, Reuters reported today. Foxconn, currently a major supplier to Apple, is expanding its business to make other electronics. As demand soars for AI startups to train big models, which require a lot of computing power, Foxconn is looking to compete for new markets, and hitching a ride on NVIDIA's giant wheel is a natural first choice. This year ...

Research consulting firm Omdia said in an article published on the 1st of this month that the market demand for AI data center chips will reach $151 billion (currently about 1.08 trillion yuan) in 2029, but growth will slow significantly after 2026. According to Omdia's "Cloud Computing and Data Center Artificial Intelligence Processor Forecast" report, the market size of AI data center chips will be less than $10 billion in 2022, and has now grown to $78 billion this year (currently about $5562.87 billion)...

According to official news from Tsinghua University, the research team of Professor Fang Lu from the Department of Electronic Engineering of Tsinghua University and the research team of Academician Dai Qionghai from the Department of Automation have taken a different approach and pioneered a fully forward intelligent optical computing training architecture, developed the "Tai Chi-II" optical training chip, and achieved efficient and accurate training of large-scale neural networks in optical computing systems. The research results were published online in the journal Nature on the evening of August 7, Beijing time, under the title "Full Forward Training of Optical Neural Networks". According to inquiries, the Department of Electronic Engineering of Tsinghua University is the first unit of the paper, Professor Fang Lu and Professor Dai Qionghai are the corresponding authors of the paper, and Xue Zhiwei, a doctoral student in the Department of Electronic Engineering of Tsinghua University...

The U.S. Department of Commerce has just announced a massive $400 million investment to support domestic semiconductor wafer production and advance U.S. technology. The investment is part of the CHIPS Act, which aims to boost domestic semiconductor wafer production and enhance U.S. technological capabilities. This time, the lucky winner is GlobalWafers, which reached a preliminary non-binding memorandum of understanding (PMT) with the Department of Commerce and received this huge amount of money. Source Note: The image is generated by AI, and the image licensing service provider Midjou…

Kneron, an AI chip startup, launched its next-generation AI products, the KNEO 330 server and PC devices equipped with the third-generation NPU chip KL830, at the 2024 Taipei International Computer Show on Wednesday. It is reported that the company was founded in 2015 by CEO Liu Juncheng and Zhang Maozhong. It is a company that provides edge computing artificial intelligence (edge AI) technology. Its investors mainly include well-known groups such as Li Ka-shing's Horizons Ventures, Qualcomm, Foxconn Group, Alibaba Entrepreneurs Fund, and China Development Capital. …

British chip design company Arm Holdings predicts that by the end of 2025, there will be 100 billion Arm devices ready for artificial intelligence applications worldwide. Arm CEO Rene Haas announced the news at the Computex forum in Taipei. The news means that chip devices designed by Arm will promote the development of artificial intelligence worldwide. Source Note: Image generated by AI, image licensing service provider Midjourney Arm is a leading chip design company, and their products are widely used in...

Intel recently announced the release window for its new Lunar Lake chip laptop processor, which is expected to be officially launched in the third quarter of this year. It is designed to bring a new AI experience to Copilot Plus PC. Intel's announcement coincides with Microsoft's Surface and AI event, where Microsoft adopted Qualcomm's Snapdragon X Elite and Plus chips to provide high performance and longer battery life. However, Intel claims that its Lunar Lake processors are more powerful than Stabl…

Amazon's cloud computing division AWS has reportedly suspended orders for Nvidia's most advanced "superchip" Grace Hopper in anticipation of a new, more powerful product, Grace Blackwell (GH200). The move comes as investors worry that Nvidia will see a demand slump between product cycles. In March, Nvidia released a new generation of processors called Blackwell, less than a year after its predecessor, Hopper, began shipping to customers. Nvidia CEO Jensen Huang said the new product is effective in training large language...

On the 24th, the Beijing Municipal Bureau of Economy and Information Technology and the Beijing Municipal Communications Administration issued the "Beijing Computing Infrastructure Construction Implementation Plan (2024-2027)". The "Implementation Plan" proposes that by 2027, the quality and scale of computing power supply in Beijing, Tianjin, Hebei and Mongolia will be optimized, and efforts will be made to ensure that independent and controllable computing power meets the needs of large model training, and computing power energy consumption standards will reach the leading domestic level. Key tasks include promoting independent innovation in the computing power industry, building an efficient computing power supply system, promoting the integrated construction of computing power in Beijing, Tianjin, Hebei and Mongolia, improving the green and low-carbon level of the intelligent computing center, deepening computing power empowerment industry applications, and ensuring computing power foundations...

LG recently released a press release, showcasing its self-developed on-device AI chip DQ-C, which is planned to be deployed in 46 products in 8 categories. This DQ-C chip is mainly used in the internal system of home appliances, supporting AI control, driving LCD screens, recognizing voice, etc. The chip is produced by TSMC's 28nm process technology. LG spent three years in-depth research and development of the DQ-C chip, which was first released in July 2023 and used in LG washing machines, dryers, air conditioners and other 5…

Meta Platforms released the latest version of its Training and Inference Accelerator (MTIA) project on the 10th local time. MTIA is a custom chip series designed by Meta specifically for AI workloads. It is reported that the new generation of MTIA released this time has significantly improved performance compared to the first generation of MTIA, and helps to strengthen content ranking and recommendation advertising models. Its architecture fundamentally focuses on providing the right balance of computing, memory bandwidth and memory capacity. The chip can also help improve training efficiency and make inference (that is, actual reasoning tasks) more…

One of the beneficiaries of the boom in demand for high-bandwidth memory driven by artificial intelligence is a Kyoto-based company that controls a small but crucial part of the chipmaking process. Towa has two-thirds of the global market for chip molding equipment, according to research firm TechInsights. It’s a crucial step that encases chip dies and wires in resin, protecting them from dust, moisture and shock so they can be safely stacked on top of each other, giving graphics processors such as Nvidia Corp. more power to better train artificial intelligence. The…

Elon Musk's Tesla and his mysterious AI-focused company xAI have stockpiled a large number of Nvidia H100 series chips. Tesla intends to use this to overcome the ultimate challenge of autonomous driving - Level 5 autonomous driving, while xAI is responsible for realizing Musk's vision of "ultimate truth artificial intelligence". X platform user "The Technology Brother" recently announced that Meta has stockpiled the world's largest number of H100 GPUs, reaching an astonishing 350,000. However, Musk is not interested in this...

Well-known Bloomberg reporter Mark Gurman recently revealed that Apple is going all out to develop a new MacBook Pro with the M4 chip. At the same time, a striking roadmap released by Canalys shows that the much-anticipated M4 series chips are expected to be officially unveiled in the first quarter of 2025. This AI processor roadmap carefully drawn by Canalys not only reveals the launch time of the M4 series chips, but also points out that the series of chips will focus on AI functions. This means that the future MacBook Pro will usher in a major breakthrough in the field of artificial intelligence, which will bring…

OpenAI CEO said the company is interested in working with major Korean chipmakers such as Samsung Electronics and SK Hynix. Sam Altman revealed that he visited South Korea twice in the past six months and had fruitful talks with Samsung and SK Hynix during his most recent visit. At a meeting at the company's headquarters in San Francisco, when asked by Korean reporters whether there were any plans for semiconductor cooperation with Korean chipmakers, Altman said he had visited South Korea in late February 2024 and met with executives from Samsung Electronics and SK Hynix...

Nvidia will hold a GTC 2024 keynote tomorrow, and Jensen Huang is expected to announce the next-generation GPU architecture called Blackwell. According to XpeaGPU, the B100 GPU to be launched tomorrow will use two chips based on TSMC's CoWoS-L packaging technology. CoWoS (chip on wafer substrate) is an advanced 2.5D packaging technology that involves stacking chips together to increase processing power while saving space and reducing power consumption. XpeaGPU revealed that the two computing...

Wafer-level chip innovation company Cerebras has launched its third-generation chip WSE-3, claiming that it has doubled the performance of its previous generation WSE-2 with the same power consumption. The parameters of WSE-3 are as follows: TSMC 5nm process; 4 trillion transistors; 900,000 AI cores; 44GB on-chip SRAM cache; 1.5TB/12TB/1.2PB optional off-chip memory capacity; 125 PFLOPS peak AI computing power. Cerebras claims that based on WS…

According to Bloomberg, AMD tried to approve an artificial intelligence chip specifically for the Chinese market, but was rejected by US officials. People familiar with the matter revealed that AMD hopes to obtain approval from the US Department of Commerce to sell the chip to Chinese customers, but US officials believe that the chip is still highly advanced technology, so AMD must obtain permission from the Bureau of Industry and Security of the Department of Commerce before it can be sold. This stricter control has also affected the sales of processors designed specifically for China by Nvidia, another leading artificial intelligence chip manufacturer, which has complied with the 2022 version of the export regulations. For…

Source @Kepler_L2 revealed that AMD will launch a revised version of the AI accelerator MI300 using HBM3e to compete with Nvidia B100 at a low price, and AMD's next-generation Instinct MI400 accelerator will be released in 2025. AMD announced two Instinct MI300 accelerators in 2023, namely the pure GPU MI300X and the APU architecture MI300A, both of which use HBM3 memory with a capacity of 19…