Roughly twice the GDP of Japan! Explore the background and details of Sam Altman's "AI semiconductor development plan" to invest "up to 1,000 trillion yen".
1)
Sam Altman's outrageous plan
Wall Street reported last week that OpenAI CEO Sam Altman plans to raise a total of $5 trillion to $7 trillion (more than 750 trillion to 1,000 trillion yen) to develop new semiconductors for AI. The Journal (WSJ) reported.
To realise this plan, he is said to be in talks with the United Arab Emirates (UAE) sovereign wealth fund, which will be the primary source of funding, and Taiwan's TSMC, which will be responsible for the actual semiconductor manufacturing.
2)
Again, the total amount of funding he is trying to raise is up to $7 trillion. I'm not sure what to compare that to, but for example, it's about five times the total amount of corporate bonds issued in the US in 2023, which was $1.44 trillion.
It is also much larger than the combined value of Microsoft and Apple, currently the world's first and second most prominent companies in terms of enterprise value (stock market capitalisation), which is around $6 trillion. Or, since Japan's GDP (Gross Domestic Product) is about 560 trillion yen ($3.7 trillion), it could be seen as about two years of that amount.
3)
It's a monumental undertaking. Still, his and OpenAI's ultimate goal is "AGI (artificial general intelligence)", that is, to surpass the intelligence of humanity as a whole. Achieving "super-intelligence" may require that much money.
4)
A scale that will fundamentally reshape the semiconductor industry
Leaving aside future issues such as AGI, the immediate motive seems to be a shortage of GPUs (Graphics Processing Units). This semiconductor product is essential for developing generative AI, such as ChatGPT.
The supply and demand for GPUs has recently become tight worldwide, with around 80% of the market controlled by the US company Nvidia. It can be said that it is close to a semi-monopoly situation.
Not satisfied with this, Altman seems to be taking the initiative to develop a new processor for AI (to replace GPUs) and wants to assume control of the semiconductor industry, which is the basis of generative AI.
5)
But even here, the total investment of up to $7 trillion is extraordinary.
Global semiconductor sales in 2023 will be about $527 billion (about ¥79 trillion) and reach about $1 trillion (about ¥150 trillion) in 2030. In other words, the total amount of money Altman wants to raise and invest will be about seven times the total amount of products sold by the world's semiconductor manufacturers in about seven years (at the current rate).
6)
It's unclear when Altman intends to implement this plan, but it's probably not too far off. This means that current investment is far more significant than future sales. On the other hand, if this is not the case, it would suggest that the market forecasts for around 2030 would be revised upwards automatically and significantly.
Rather than simply taking control of the industry, this is an attempt to fundamentally reshape the global semiconductor industry through one person's will.
7)
Big Tech developing AI semiconductors independently
However, Altman (OpenAI) is not alone in his efforts to create new processors (semiconductor products for logical operations and data processing) for generative AI to compete with Nvidia's GPUs.
8)
The market for AI processors such as GPUs is expected to be worth around $140 billion (£21 trillion) by 2027, more than double its current size. Traditional semiconductor manufacturers such as Intel and AMD, which have dominated the conventional CPU (central processing unit) market, have recently focused on developing AI processors such as GPUs to tap into this vast market.
9)
But perhaps even more remarkable is that the major tech companies developing and offering generative AI, such as Google, Meta (formerly Facebook), Microsoft and Amazon, independently develop AI-specific processors. The primary purpose is to reduce the enormous costs of creating and operating generative AI.
Large-scale generative AIs, called "foundation models", such as OpenAI's ChatGPT or Google's recently launched Gemini, require tens of thousands of GPUs for machine learning and subsequent operation. Or they need an AI processor.
10)
Nvidia's GPUs are expensive, averaging about $15,000 (£2.3 million). Still, the development cost of the semiconductor chips (processors) called TPUs that Google makes in-house to generate AI is about $2,000 to $3,000 (£300,000 to £450,000).450,000), which is relatively cheap.
In other words, by producing processors for generative AI in-house, Google can significantly reduce the cost of procuring such processors and, thus, the cost of developing and operating generative AI.
11)
In a bid to compete with Google, Meta announced last May that it had begun developing its first AI processor. Microsoft also announced an AI processor called "Maia" in November of the same year. Amazon is also said to be creating its own AI processor. In both cases, the main goal is to reduce generative AI's development and operating costs.
12)
In the future, however, these big tech companies are expected to offer their AI processors to companies other than their own. Indeed, in September last year, Amazon invested up to $4 billion in Anthropic, a startup developing generative AI. Still, one of the conditions was that the company adopted Amazon's own AI processor. It was something to do, they say.
13)
In addition to Amazon, other big tech companies, such as Microsoft and Google, offer generative AI services to customers through their cloud services. We believe that if we replace each of the many GPUs installed in these generative AI cloud platforms with our own AI processors, we can break Nvidia's stronghold that currently dominates the GPU market.
14)
This is just speculation, but Mr. Altman mentions a whopping $5 trillion to $7 trillion in funding for developing AI processors due to the company's ability to create a new semiconductor market to compete with these formidable Big Tech companies. It may have been to show determination or enthusiasm for pioneering.
15)
The US government is concerned about information leaks to China via the Middle East.
Mr Altman recently met with UAE strongman Sheikh Tahnoon bin Zayed Al Nahyan to raise this enormous amount.
16)
Incidentally, "Sheikh" is an Arabic word meaning ``leader'' but is generally used as an honorific or title in Islamic society.
Sheikh Tahnun is the younger brother of Sheikh Mohammed bin Zayed Al Nahyan, the current president of the United Arab Emirates. He has been a leader in moving the country's industry away from its traditional dependence on oil.
It plays a significant role in weaning the country off the sector and transitioning to next-generation industries, such as AI 2017, the country's government created the world's first 'Minister of AI'.
17)
At the same time, the United Arab Emirates has recently deepened its ties with China, accelerating its expansion into the Middle East and causing concern for the US government.
18)
Under these circumstances, in October last year, OpenAI partnered with G42, an IT and AI development company in the United Arab Emirates, whose CEO is from China.
Although not necessarily related, the US CIA (Central Intelligence Agency) is concerned that information on cutting-edge technologies related to national security, such as AI semiconductors, will be leaked from the US to China via G42.
19)
Altman's meeting with Sheikh Tahnun appears to be to negotiate funding for the development of AI semiconductors, estimated to be worth up to $7 trillion.
Sheikh Tahnun holds a key position in the United Arab Emirates sovereign wealth fund and serves as chairman of the G42, which appears problematic for US government officials.
20)
To alleviate these concerns, Altman reportedly recently met with US Secretary of Commerce Gina Lemond and other government officials to explain the situation.
There are three central sovereign wealth funds in the UAE, including the Abu Dhabi Investment Authority (ADIA), which is said to be one of the largest in the world.
Still, the total amount of money they can manage is not officially disclosed. Although the figure is not fixed, some estimates put it at around $1 trillion.
21)
In other words, even with one of the world's most significant Arab sovereign wealth funds, he will still fall far short of securing the "$5 trillion to $7 trillion" investment that Altman said he would. So where is he going to find the rest?
Roughly twice the GDP of Japan! Explore the background and details of Sam Altman's "AI semiconductor development plan" to invest "up to 1,000 trillion yen".
https://gendai.media/articles/-/124359?page=4
Masayoshi Son's "Izanagi" plan to establish a new company for AI semiconductors for 15 trillion yen...US report
https://www.yomiuri.co.jp/economy/20240217-OYT1T50100/?from=smtnews
