Sam Altman’s $7 trillion AI chip project might not be very realistic

Sam Altman’s $7 trillion AI chip project might not be very realistic

OpenAI CEO Sam Altman wishes to raise trillions of dollars to improve the worldwide semiconductor market, The Wall Street Journal reported previously this month, an effort to enhance chip-making capability and power more expert system. It’s an eye-boggling quantity, one that was put to Nvidia CEO Jensen Huang– the male behind the AI business of the minute — for his ideas.

Tarek El Moussa’s roadway out of financial obligation to being a millionaire|Your Wallet

When asked throughout the World Government Summit in Dubai today the number of GPUs can be purchased for $7 trillion, Huang jokingly reacted: “Apparently all the GPUs.” (GPUs, or graphics processing systems, power generative AI applications like ChatGPT and OpenAI’s brand-new video-generating AI Sora)

Huang then revealed apprehension about the figure. He stated computer systems powering AI will continue to advance, which drives down expenses.

“You can’t presume simply that you would simply purchase more computer systems, you likewise need to presume that the computer systems are going to end up being quicker, and for that reason, the overall quantity you require will not be as much,” Huang stated.

“Otherwise,” he included, “the mathematics, if you simply presume that computer systems never ever get any faster you may pertain to the conclusion we require 14 various worlds and 3 various galaxies and 4 more suns to sustain all this. Certainly, computer system architecture continues to advance.”

Will Sam Altman invest trillions on information centers to sustain larger OpenAI designs?

Let’s take an action back and take a look at exactly what $7 trillion may money.

The AI designs behind ChatGPT do need a great deal of calculating power, more than many individuals understand, stated Willy Shih, a teacher at Harvard Business School who formerly operated at IBM.

If Altman’s aspiration is to make larger designs for OpenAI, he might invest trillions on information centers, which house the GPUs required to train AI designs that power items like ChatGPT and Sora. The U.S. information center building market was valued at $24.63 billion in 2024, research study company IDC quotes. If he invested $1 trillion on chips, he might purchase 40 times as numerous information centers than presently exist.

Information centers presently utilize less than 1% of the electrical energy supply of the U.S., Shih stated. Altman would require to construct a lot of electrical energy generation centers– which produce electrical power from different energy sources– to support his brand-new information. He would require to update the extremely electrical grid that really disperses the energy to the information. When you think about the cash being invested through the federal Inflation Reduction Act and the Facilities and Investment Act to incentivize tidy energy production and grid modernization in the U.S., a trillion there would most likely be a great financial investment, Shih stated.

Or will Altman invest trillions on structure numerous chip factories?

Maybe Altman wishes to broaden international chip-making capability. There are simply a handful of leading-edge fabs, which are producing plants that produce chip parts, being integrated in the world today: TSMC in Taiwan, Arizona, and Japan, Samsung in Korea and Texas, and Intel in Arizona, Ohio, and Israel, to name a few.

$7 trillion might purchase more than 200 leading-edge semiconductor fabs for $30 billion each, Berstein semiconductor expert Stacy Rasgon approximates.

With 200 or perhaps 100 fabs, you would require to begin developing out steel mills and concrete plants, Shih stated. Altman would likewise require to purchase a great deal of building devices. Getting a provider to produce the leading-edge UV devices required for the scale of Altman’s job might take years, Shih included.

There’s the cash required to train the employees to fill the factories. Chip business like TSMC have currently grumbled that employees aren’t proficient enough for their CHIPs Act tasks in Arizona, postponing the opening of brand-new factories

Anybody can toss out a huge number

If cash might purchase what we desire, China would have gotten much even more with the $150 billion Made in China 2025 financial investment into its domestic chips, Shih stated. China hasn’t rather attained self-reliance. The nation, for example, invests two times as much importing semiconductors as it invests in oilaccording to a report from the Canadian bank RBC Wealth Management.

The concern then is not whether one can invest all that cash, however how far will all that cash go?

A minimum of in the meantime, the mathematics does not appear to accumulate.

Find out more

Leave a Reply

Your email address will not be published. Required fields are marked *