Sam Altman's plans to overhaul the semiconductor industry with an injection of up to $7 trillion in private investments could be as unrealistic as they are ambitious.
The CEO of OpenAI, whose company recently shattered growth records with revenues of over $2 billion last year, is chasing the quest of building an artificial general intelligence: an AI model able to universally perform better than humans on any given task.
For this, the executive is embarking on a quest to raise a sum that seems impossible for a company its size.
Earlier this month, a report by Cathie Wood's Ark Invest predicted that the cost of training AI programs would drop by 75% on a yearly basis until 2030.
In a Monday speech, Nvidia Corp's
Altman himself confirmed that the cost of training GPT-4, the latest AI language model behind the company's successful ChatGPT, was above $100 billion.
However, the processing capacity needed to build exponentially more powerful models is beyond today's reach. The complex process of manufacturing high-performance chips needed in AI system training requires the coordination of many areas including supply chains, facility development and expert training, many of which fall short of today's resources.
There are currently only a handful of companies in the world with the infrastructure and the know-how to manufacture the most advanced chips that will train the next generation of AI systems. Taiwan Semiconductor Mfg. Co. Ltd.
Both Big Tech companies and governments have become aware of the need to support the growth of AI as a strategic asset in securing market dominance and geopolitical power, respectively.
Securing the necessary infrastructure to make those developments is a key step along the way. In many cases that means investing heavily in chip factories.
President Joe Biden's CHIPS Act, which intends to invest billions in the development of a domestic semiconductor industry, has so far experienced delays in its implementation, but the effort continues to be a major angle in his Administration's push to confront China's growth in the AI race.
Some of the same hurdles that are causing difficulties implementing CHIPS Act policies can become a challenge to Altman's ambitions to obtain unprecedented levels of processing power for AI training.
A shortage of experts is already a problem for expanding manufacturing capacities in the semiconductor industry. It has been estimated that 67,000 professionals will be lacking from the industry by 2030.
Brain capital shortages have been behind China's failure to build up a local semiconductor industry that can overcome Taiwan's technological achievements, according to a report by The Wall Street Journal. Although the country invested almost $300 billion in the industry in the period 2021-2022, according to JW Consulting, China continues to rely on foreign-made chips to train AI platforms.
Moreover, the sum of between $5 trillion and $7 trillion that Altman reportedly mentioned as his company's new investment goal can be seen as out of reach when put in context. It exceeds the entire U.S. government spending for 2023, as well as the combined market capitalization of Alphabet Inc
According to the WSJ, Altman has been in talks with key investors and companies in the chip space in order to raise the funds to build dozens of new manufacturing plants. These include U.A.E.'s Sheikh Tahnoun bin Zayed al Nahyan, a top security official and brother of U.A.E. President Mohamed bin Zayed al Nahyan. The country has already invested billions in the industry, as per The National Interest.
Others include Masayoshi Son, the CEO of SoftBank Group Corp
