Morgan Stanley: Computing power investment enters the second phase, "AI Data Center" takes over from "Chips".

Author: Zhu Xueying

Source: Hard AI

As Huang Renxun boldly declares that the next four years will see a $1 trillion investment in data center upgrades, the development of computational power has entered its second phase, with dividends gradually shifting from chips to infrastructure. Morgan Stanley specifically points to data centers, believing they will usher in a new wave of AI investment.

In their latest research report on March 1st, Morgan Stanley stated that the next wave of investments will focus more on other infrastructure of AI data centers, such as servers, network equipment, cooling systems, data storage, and even physical real estate.

Semiconductors Leading the Way, AI Data Centers Up Next

According to Morgan Stanley's analysis, each new computing cycle that emerges every decade increases the ubiquity of computing power by 10 times, following a similar path of monetization.

Therefore, with the development of AI, investments in data centers will go through the traditional monetization cycle from semiconductors to infrastructure and then to software/services.

In the past 12 months, NVIDIA's stock price has risen by approximately 250%, and the semiconductor industry (such as the SOX index rising by about 60%) has also led the way in growth, indicating that they have already benefited from the early growth of the AI industry. In contrast, other companies in the data center technology field have seen slower growth (with a median of only 20%), indicating that there is still plenty of room for investment and value growth in the data center infrastructure sector in the future.

Currently, investments in AI data centers are accelerating in the market, with indications seen from land sales to power demand.

Looking ahead, with the increasing demand for AI data center infrastructure, including servers, network equipment, cooling systems, data storage, and even physical real estate, a new wave of investment boom is expected in these areas.

What Could Hinder AI Data Centers?

It is worth noting that Morgan Stanley also emphasized in the research report the potential constraints that AI data centers may face in the future, especially power limitations.

Compared to the present, the technological and energy infrastructure required for AI data centers in the future is incomparable. In addition to traditional technical infrastructure, the importance of power and building infrastructure is also increasing. This is particularly crucial as data centers face more and more challenges due to power limitations.

According to relevant data, the power consumption of global data centers was 10 billion watts ten years ago, and now the level of 100 billion watts is very common.

Although artificial intelligence currently accounts for only a small portion of the global data center market, according to the forecast by the U.S. Uptime Institute, by 2025, the proportion of electricity consumption by artificial intelligence businesses in global data centers will surge from 2% to 10%.

Elon Musk has also expressed concerns about the future of energy supply. At the end of last year, he mentioned in a podcast that the United States is currently facing a shortage of chips, which will be followed by a shortage of transformers in a year, and there will be a shortage of electricity in about two years.