AI Chips

There is another big news in the field of AI computing chips! Domestic AI chip design company Suiyuan Technology (full name: Shanghai Suiyuan Technology Co., Ltd.) has officially started the A-share listing guidance process. According to the latest announcement of the China Securities Regulatory Commission, Suiyuan Technology signed a listing guidance agreement with CICC on August 23, officially starting the A-share IPO process. According to the information, Suiyuan Technology was established in 2018 and focuses on the research and development of cloud computing products in the field of artificial intelligence. Their products cover AI acceleration cards, intelligent computing clusters, and related software and platforms. The company's founders Zhao Lidong and Zhang Yalin...

According to reports from Korean media Chosun Ilbo and ZDNET Korea, Park Kyung, president of SK Hynix's system architecture, said at an academic seminar held yesterday in South Korea that storage products have transformed from simple components to solutions in the AI era. Park Kyung said: "It is not enough to just have DRAM design skills. We need to look at how the system is changing, and what opportunities and challenges exist, from the perspective of "operations and systems," not just from the perspective of semiconductor technology. The semiconductor ecosystem used to be dominated by supply and demand, but it has not...

Preferred Networks (PFN), a Japanese AI chip "unicorn" company, announced today that it has formed a capital and business alliance with Japanese financial giant SBI Holdings for the development and productization of PFN's next-generation AI semiconductors. According to the agreement signed between PFN and SBI Holdings, the two companies plan to jointly develop PFN's next-generation AI semiconductors and work together to promote chip productization, and cooperate in the packaging and testing process of the chip. In addition, SBI Holdings' parent company...

Recently, a Silicon Valley company called Etched is attracting attention. It is a startup founded by two Harvard dropouts. They launched an AI chip called Sohu, claiming that it is 20 times faster than Nvidia's current most powerful H100 chip when running AI models like ChatGPT. This news makes people wonder what kind of magical product Sohu is that can have such a huge impact on the AI hardware market. Source Note: The picture is generated by AI, and the picture authorization service provider Midjourne…

AI chip startup Groq announced on Monday that it had recently completed a $640 million (currently about RMB 4.567 billion) financing, with the goal of creating a new chip that can handle AI reasoning and other tasks faster. After completing this round of financing, Groq has raised more than $1 billion in total, and the company's valuation has reached $2.8 billion (currently about RMB 19.983 billion). It was reported that the company initially planned to raise $300 million at a valuation of $2.5 billion, but the company's AI chip was recognized by investors, so...

Recently, according to The Information, the well-known generative AI application giant OpenAI is in talks with several chip design companies, including Broadcom. The focus of the talks is on developing new AI chips. OpenAI is exploring the possibility of making its own AI chips, which can not only achieve efficient integration of software and hardware, but also alleviate the current shortage of AI chips. Source Note: The image is generated by AI, and the image is authorized by Midjourney. In addition, OpenAI is actively recruiting former Google employees, intending to use their experience in developing...

According to Graphcore's official website, the troubled British AI chip company has been acquired by SoftBank. According to the agreement, Graphcore will become a wholly-owned subsidiary of SoftBank and continue to operate under its current name. The two parties did not disclose the specific transaction value. Graphcore was founded in Bristol, a city in southwest England in 2016, and has launched several generations of AI workload accelerators called Intelligence Processing Unit (IPU for short). Graphcore was once regarded as "…

Etched announced that it has completed a $120 million (currently about 873 million RMB) Series A financing, which will be used to develop and sell the world's first Transformer application-specific integrated circuit (ASIC) chip Sohu. According to public information, Etched was founded by two Harvard dropouts, Gavin Uberti and Chris Zhu, and was established less than two years ago. The biggest highlight of the Sohu chip is that it directly uses the Transformer architecture…

According to Sina Finance, there is market news today that ByteDance will cooperate with Broadcom, an American chip design company, to develop AI chips. In response, ByteDance said: The news is untrue. Earlier today, there were reports that ByteDance is working with Broadcom of the United States to develop AI processors to ensure that there are enough high-end chips. People familiar with the matter have said that this AI processor has a 5nm process and will be manufactured by TSMC. Although the design work is progressing smoothly, the "tapeout" that marks the end of the design phase and the beginning of manufacturing has not yet begun. It is reported that ByteDance and Broadcom have always been business partners, and Broadcom has been...

Elon Musk has ordered thousands of Nvidia-made AI chips to be shipped to Tesla, which will be transferred to his social media company X, according to emails from the chipmaker obtained by CNBC. The move has the potential to delay Tesla's acquisition of $500 million worth of processors by months, the outlet reported. Source note: Image generated by AI, image licensing service provider Midjourney According to Musk, Tesla should stockpile Nvidia's H100 artificial intelligence chip to fuel its transformation into a "leader in artificial intelligence and robotics." In…

At the annual Computex technology conference, GPU manufacturers showcased their latest technological achievements. Among them, Nvidia and AMD performed particularly well. Nvidia showcased the mass-produced Blackwell chip at the conference and announced its product roadmap for the next three years, including the next-generation Rubin AI platform. AMD CEO Lisa Su showcased AMD's CPU, GPU products and future roadmap, including the desktop Ryzen 9000 series CPUs, AI PC chips, data center chips and GPUs based on the new Zen 5 architecture. …

According to CNBC, citing internal emails from senior Nvidia employees, Musk "exaggerated" Tesla's chip purchases to shareholders and transferred a large number of AI chips originally planned for Tesla to the X (Twitter) platform. The report said that Musk ordered Nvidia to allow X to "jump the queue" and give it priority over Tesla, which delayed Tesla's receipt of processors worth more than $500 million (currently about 3.625 billion yuan) by several months, which may delay Tesla's progress in building a supercomputer. The report also mentioned a British...

Market research firm Gartner believes that the total global AI chip revenue will reach US$71.252 billion (currently about RMB 517.29 billion) in 2024, a year-on-year increase of 33%. Image source Pexels 2024 will be the year with the most prominent growth in AI chip revenue. The growth rate will decline in the next few years, but will continue to maintain a double-digit growth percentage. In 2025, the total global AI chip revenue is expected to reach US$91.955 billion (currently about RMB 667.593 billion), which is 1.3% higher than the previous year.

According to Reuters, citing people familiar with the matter, Nvidia's AI chips developed for the Chinese market had a weak start and were in ample supply. Under competition from Huawei, they began to cut prices, and in some cases, they were priced lower than Huawei chips. The report said that the Chinese market accounted for 17% of Nvidia's revenue in fiscal 2024, and the growing competitive pressure has also sounded the alarm for Nvidia's investors. Nvidia launched three chips tailored for the Chinese market at the end of last year, among which the most powerful H20 chip has attracted much attention. Three supply chain sources said that the chip is in ample supply in the market, indicating weak demand. The three…

SoftBank plans to launch its own AI chip in 2025 and invest $64 billion (10 billion yen) in AI chips, robots, data centers and other fields. According to the Nikkei, SoftBank's subsidiary Arm, known for its smartphone chip design, will establish its own AI chip division. Source Note: Image generated by AI, image licensing service provider Midjourney expects to launch a prototype in spring 2025 and mass-produce it by contract manufacturers. Arm will initially bear hundreds of billions of yen in development costs and receive SoftBank's...

Taiwan Semiconductor Manufacturing Co. (TSMC), a major producer of advanced chips for artificial intelligence applications, forecast a 51% rise in first-quarter profit on Thursday, thanks to strong demand. The world's largest foundry chipmaker, whose clients include Apple and Nvidia, has benefited from a surge in artificial intelligence, helping it weather the weakening demand for electronics caused by the pandemic, pushing TSMC shares to a record high. Source note: Image generated by AI, image licensing service provider Midjourney Analysts expect TSMC's net profit to reach NT$217.2 billion (67.1 billion) in the quarter ending March 31...

Marvell announced at an investor event on AI infrastructure last week that it had recently received an AI chip order from a large technology company. Marvell claimed that, including this order, it has won custom chip orders from three of the "four largest US companies", including custom AI training accelerators and AI inference accelerators for customer A, custom Arm architecture CPUs for customer B, and recently custom AI accelerators for customer C. ▲ Image source Marvell…

On March 28, according to a report by the Nikkei today, Nvidia's cutting-edge image processing semiconductor (GPU) H200 is now available. H200 is a semiconductor for the AI field, and its performance exceeds the current flagship H100. According to the performance evaluation results released by Nvidia, taking the processing speed of Meta's large language model Llama 2 as an example, the processing speed of generative AI derived answers of H200 is up to 45% higher than that of H100. Market research organization Omdia once said that in 2022…

According to The Information, Meta is taking a series of "unconventional measures" to gain a greater advantage in the competition for artificial intelligence researchers, further increasing the gunpowder smell of the AI talent war. Image source Pexels reported that this series of measures include offering job opportunities to candidates without interviewing them, and increasing the salaries of employees who previously threatened to leave - before this, Meta's "long-term practice" was not to increase salaries. According to two people who have viewed Meta's related emails, the company's CEO...

A Meta spokesperson told Reuters on the 19th Eastern Time that the company expects to receive Nvidia's latest flagship AI chip, the B200, later this year, which will be Nvidia's first shipment. Nvidia's Chief Financial Officer Colette Kress told financial analysts, "We think (the product) will be available later this year," but she also said that shipments of the new GPU will not increase until 2025. It is reported that social media giant Meta is one of Nvidia's largest customers. The company has purchased hundreds of thousands of previous-generation accelerator cards (H100...

The newly released world's most powerful AI chip WSE-3 has attracted the attention of the industry. The chip uses an advanced 4 trillion transistor 5nm process and has amazing performance. A single machine can train a model with 24 trillion parameters, far exceeding Nvidia's H100 GPU. The technical advantage of WSE-3 lies in its 900,000 AI cores and 44GB on-chip SRAM storage, which enables its peak performance to reach 125FP16PetaFLOPS, achieving single-chip cluster-level performance. Cerebras' CS-3 supercomputer further improves the efficiency of large-scale AI model training, and can train more than GP...

According to foreign media SiliconANGLE, Taalas, an AI chip startup led by Tenstorrent founder Ljubisa Bajic, has successfully completed two rounds of financing totaling US$50 million, aiming to create customized dedicated chips for specific AI models. Source note: The image is generated by AI, and the image licensing service provider Midjourney Ljubisa Bajic has worked at AMD and Nvidia and is the founder and first CEO of Tenstorrent. He moved to T…

On March 10, three American writers filed a class action lawsuit against chipmaker Nvidia in the San Francisco federal court, accusing the company of using their copyrighted works to train the artificial intelligence (AI) system NeMo without authorization. According to the lawsuit, writers Brian Keene, Abdi Nazemian and Stewart O'Nan claimed that their works were part of a data set of nearly 200,000 books that were used to train NeMo to simulate...

According to a recent survey, a large number of artificial intelligence professionals are planning to switch from Nvidia to AMD Instinct MI300X GPUs. Jeff Tatarchuk, co-founder of TensorWave, revealed that they conducted an independent survey covering 82 engineers and artificial intelligence professionals. Among them, about 50% of respondents said they were confident in the AMD Instinct MI300X GPU because compared to Nvidia's H100 series products, the MI300X …