Source: Hard AI, Author: Li Dan.

Sometimes, the more idealistic the dream, the more harsh the reality appears. In February of this year, OpenAI CEO Sam Altman's fundraising news of $7 trillion to build a chip empire shocked the world. Recent reports show that OpenAI has chosen a more practical path, starting by bringing in partners to create their own chip, rather than establishing a network of chip factories.

On Tuesday, October 29th in the Eastern United States time, Reuters reported, citing sources familiar with the matter, that due to the cost and time required to establish a network of chip factories, OpenAI has now abandoned its ambitious chip factory plans and instead plans to focus on internal chip design work.

The report stated that OpenAI has formed a chip team of about 20 people, led by top engineers who previously worked for Google to create the Tensor Processing Unit (TPU), in collaboration with Broadcom and Taiwan Semiconductor, to create their first self-developed chip aimed to support their artificial intelligence (AI) system, with a focus on inference.

The report mentioned that OpenAI has been cooperating with Broadcom for several months, and through Broadcom, has already reserved the production capacity of Taiwan Semiconductor, planning for Taiwan Semiconductor to manufacture OpenAI's first custom-designed chip in 2026. OpenAI has yet to determine whether to develop or acquire other elements of their chip design, and may collaborate with more partners in creating the chip.

At the same time, the report mentioned that to meet the increasing demand for infrastructure, OpenAI will not only continue to use Nvidia chips but also adopt AMD chips.

This is not the first time that OpenAI has reported collaborating with Broadcom to develop ai chips. In July of this year, the Wall Street Journal mentioned that the media reported that OpenAI has been recruiting former Google employees who were involved in producing Google's TPU ai chips, and seeking to develop an ai server chip, and has been in talks with chip design companies including Broadcom about developing new ai chips.

At the time, the media mentioned that Altman discussed with executives from Taiwan Semiconductor whether Taiwan Semiconductor could increase production capacity to produce more Nvidia chips or the new ai chips proposed by Altman. Taiwan Semiconductor's executives told Altman that if he or OpenAI could commit to a large number of new chip orders, Taiwan Semiconductor would be willing to expand chip production.

Developing new chips is only part of Altman's plan. Altman reportedly also plans to establish one or more companies with external investors to cover the costs of property, electrical utilities, datacenters, and setting up dedicated AI chip servers.

The media believes that the possibility of OpenAI developing server chips comparable to Nvidia is slim, it may take years for results to show, and may bring the risk of annoying Nvidia, which is OpenAI's most important chip supplier. However, it may also provide leverage for future price negotiations with Nvidia for OpenAI.

In any case, working with OpenAI is bullish for Broadcom. In July, after news of the collaboration came out, Citigroup analysts released a report stating that following Google, Meta, and ByteDance, OpenAI will become Broadcom's fourth largest ASIC (custom chip) customer, and Broadcom is expected to deliver to OpenAI after the second half of next year.

After news of OpenAI's collaboration with Broadcom surfaced again on Tuesday, Broadcom's stock price rapidly expanded at midday on Tuesday. In less than 20 minutes, the intraday gain increased from less than 1.7% to nearly 3.7%, hitting a new high before closing, with an intraday increase of nearly 4.8% and ultimately closing with a 4.2% gain.

It would not be surprising if OpenAI indeed abandons the dream of a chip empire as mentioned in the news on Tuesday. Some previous reports considered this dream to be crazy.

7 trillion US dollars is equivalent to 10% of global GDP, almost 14 times the total revenue of the entire semiconductor market last year. Based on the market cap calculation at the stock price in February, this amount of money can basically buy the entire semiconductor ecosystem including Nvidia, Taiwan Semiconductor, and Broadcom, with the remaining money enough to acquire Meta for another 300 billion US dollars.

The media points out that many executives in the chip industry believe that Altman's factory expansion ambitions are unlikely to become a reality because it requires a large amount of funds and skilled labor. For example, when asked about the prospects of a new factory, the CEO of Taiwan Semiconductor, Wei Zhejia, said in June that Altman is "too aggressive, I can't believe it."

Editor/Lambor