On February 3, local time, Sam Altman, co-founder and CEO of OpenAI, the developer of ChatGPT, said in an interview with the media in Tokyo, Japan that OpenAI has “no plans” to sue Chinese artificial intelligence (AI) startup DeepSeek and will continue to win with excellent products and technical strength.
DeepSeek has not affected major U.S. cloud service companies to increase AI capital expenditures
Recently, the two open source AI models V3 and R1 launched by DeepSeek, a Chinese AI technology manufacturer, have subverted the industry's view of "AI computing power requirements."
In particular, the large-scale inference model DeepSeek-R1 launched by DeepSeek is not only comparable in performance to OpenAIo1, but its training cost may be only about 1/20 of the latter (only 2048 H800 GPUs were used, it took two months to complete the training, and only cost about 5.58 million US dollars), and the API pricing is only about 1/28 of the latter, which is equivalent to a reduction of about 97% in usage costs.
In other words, DeepSeek uses less advanced AI chips, with lower computing power requirements and lower costs, to achieve the effect of top AI large models from American AI technology manufacturers such as OpenAI.
This achievement is considered a threat to the United States' leadership in artificial intelligence. It not only triggered panic among many large model manufacturers such as OpenAI, Meta, and Google, but also triggered concerns about the decline in demand for AI chips in the future. The value of AI chip companies such as Nvidia was revalued and their stock prices plummeted.
but,Judging from the current situation, US cloud service vendors have not cut back on AI infrastructure spending in the short term.
When Microsoft and Meta held a financial report meeting on January 29, they both emphasized the need to invest huge amounts of money in the field of AI. Capital expenditures in 2025 were not affected at all by the AI "dark horse" DeepSeek.
Meta CEO Zuckerberg reiterated that capital expenditures of US$60 billion to US$65 billion will be invested in AI development in 2025. He believes that over time, large investments in capital expenditures and infrastructure will become strategic advantages.
Microsoft also reiterated that capital expenditures will continue to increase in 2025, and it is expected to invest US$80 billion in building AI data centers. Microsoft CEO Nadella said that as AI becomes more efficient and accessible, demand will grow rapidly. Microsoft will continue to expand computing resources globally and strike an appropriate balance between training and inference resource allocation.
On February 3, Japan's SoftBank Group and OpenAI held a joint press conference in Tokyo, Japan. SoftBank Group CEO Masayoshi Son announced that they had reached a cooperation agreement with OpenAI CEO Altman to establish a joint venture SBOpenAIJapan in Japan to provide artificial intelligence services to corporate customers.
The joint venture SBOpenAIJapan will be jointly owned by a company established by SoftBank Group, OpenAI and Japan’s domestic telecommunications sector. SoftBank Group will pay $3 billion annually for SoftBank Group companies to use OpenAI's technology.
Prior to this, companies such as SoftBank Group, OpenAI, and Oracle also announced the establishment of a joint venture called Stargate, which will invest US$500 billion in the next four years, and will now invest US$100 billion immediately to build AI infrastructure.
Regarding concerns about declining demand for computing power caused by DeepSeek, Altman said: "The world is going to need so much computing."
A recent research report from market research organization Gartner also stated that driven by the demand for artificial intelligence, global semiconductor revenue is expected to increase by 12.6% year-on-year to US$705 billion by 2025.
OpenAI CEO: Will not sue DeepSeek and will win with strength!
After the DeepSeek-R1 inference model became popular on the Internet,OpenAI officials have stated that they are investigating signs that DeepSeek may have improperly "distilled" its model.Distillation is a technique for transferring knowledge from a large model to a smaller model.
An OpenAI spokesperson said distillation does not expose the inner workings of the model and developers can use it to improve their applications. However, OpenAI’s terms of service prohibit users from using data obtained through distillation to build competing AI products.
Related reports stated that OpenAI researchers had previously noticed that individuals possibly associated with DeepSeek had extracted large amounts of data from the company’s application programming interface last fall.
OpenAI also warned that Chinese companies are actively trying to copy its advanced artificial intelligence models.
However, OpenAI's attitude seems to have changed recently. OpenAI CEO Altman publicly stated after releasing the latest inference model o3-mini on February 1:"It (DeepSeek) is indeed a very good model, and OpenAI will develop better models, but we will not maintain as big a lead as we did in previous years."
Altman also made a rare admission of his mistakes, saying that OpenAI is discussing a new open source strategy. “I personally think we’re on the wrong side of this issue and need to come up with a different open source strategy; not everyone at OpenAI shares that view, and it’s not our highest priority right now.”
At the joint press conference between SoftBank Group and OpenAI in Tokyo, Japan, Altman also told reporters:"We have no plans to sue DeepSeek at the moment. We will only continue to build excellent products and lead the world with our model strength."
"DeepSeek is indeed an impressive model, but we believe that we will continue to push forward and launch excellent products, so we are happy to see competitors." Altman said: "We have faced many opponents in the past, but I think it is in everyone's interest to promote technological progress and stay ahead."
It is worth mentioning that OpenAI itself has also faced multiple accusations of intellectual property infringement. These accusations are mainly related to the use of copyrighted content to train its generative AI models.