share_log

软银(SFTBY.US)拟投资近10亿美元布局AI算力,将采用英伟达(NVDA.US)芯片

SoftBank (SFTBY.US) plans to invest nearly 1 billion US dollars to deploy AI computing power and will use Nvidia (NVDA.US) chips

Zhitong Finance ·  Apr 23 22:48

Japanese tech giant SoftBank (SFTBY.US) is seeking to develop a “world-class” Japanese-specific generative artificial intelligence model and plans to invest $960 million over the next two years to strengthen its computing facilities.

The Zhitong Finance App learned that according to foreign media reports, Japanese tech giant SoftBank (SFTBY.US) is seeking to develop a “world-class” Japanese-specific generative artificial intelligence model and plans to invest 960 million US dollars to strengthen its computing facilities over the next two years.

The report quoted people familiar with the matter as saying that big language models (LLM, such as OpenAI's ChatGPT) require advanced graphics processing units (GPUs) for training, and SoftBank plans to buy them from US chip giant Nvidia (NVDA.US). The report said that this investment will be carried out in 2024 and 2025, and will cost 150 billion yen (about 960 million US dollars), which is a further supplement to SoftBank's 20 billion yen investment in computing infrastructure last year.

This investment is considered to be the largest investment of its kind by a Japanese company. At that time, SoftBank is expected to have the strongest computing power in Japan.

Notably, despite growing interest in LLM technology, Japan lacks private companies with the high-performance supercomputers needed to build LLMs. SoftBank's investment may change this situation. At a time when international companies are trying to enter the Japanese market, Japan will have strong domestic players in the field of generative artificial intelligence.

Just last week, OpenAI opened its first office in Tokyo as part of a global expansion plan. Meanwhile, Microsoft (MSFT.YS) said it will invest $2.9 billion over two years to increase cloud computing and artificial intelligence infrastructure in Japan.

According to reports, SoftBank expects to complete the first model in the 2024 fiscal year. The model will have 390 billion parameters, which fully demonstrates the complexity of LLM. SoftBank will also begin developing a high-performance model with 1 trillion parameters as soon as 2025.

In addition, other local companies such as Japan Telecom Corporation NTT have also announced plans to develop LLM this fiscal year. NTT said it will also invest 8 trillion yen (about 51.7 billion US dollars) in growth fields such as data centers and artificial intelligence over the next five years.

According to data from Statista Market Insights, Japan's artificial intelligence market is expected to grow to about 13 billion US dollars by 2030, an increase of 17 times compared to 2023.

As SoftBank shifts its focus to artificial intelligence, the company's stock price is trending well and has risen about 20% so far this year. SoftBank is also a major shareholder of the chip company Arm (ARM.US). Benefiting from the artificial intelligence boom, Arm's revenue expectations have also been boosted.

According to reports, SoftBank is building artificial intelligence data centers across Japan, and recently joined a 65 billion yen Hokkaido data center construction project.

Disclaimer: This content is for informational and educational purposes only and does not constitute a recommendation or endorsement of any specific investment or investment strategy. Read more
    Write a comment