Artificial intelligence may seem a new concept, but they are fascinated technologists, inventors and science fiction fans for decades.
The mathematician and computer scientist Alan Turing contemplated AI computers in the 1950s, and Rand Corp. built the first IA program in 1956. Isaac Asimov wrote “I, robot” in 1950, and “The Terminator” of film director James Cameron “launched in 1984.
However, it was not until the Opengai chatpt launched in November 2022 that AI became the mainstream. Chatgpt was the fastest application to reach 1 million users, and its generalized lawyer began a torrent of research and development of AI, even some of the largest companies on the planet.
Amazon: $ 31.4 billion. Source: SEC presentations of the company.
The money spent is undeniably striking, and the CEO believes that it is necessary, including Sam Altman de Operai, who said recently:
You must expect Operai to spend billions of dollars.
The bets are high and have caught the attention of investors, including Wall Street billionaire, David Einhorn. Einhorn is the portfolio manager behind Greenlight Capital, a coverage fund with $ 2.3 billion in assets under administration.
Einhorn recently commented on the frenzy of AI spending, and given his more than 30 years on Wall Street, investors may want to consider their opinion.
David Einhorn delivered a clear warning about the recent increase in the expenditure of AI. Bloomberg & Sol; Getty images
Companies are scheduled to spend between $ 500 billion and $ 1 billion a year on nuts and bolts necessary to support artificial intelligence.
Much of that money is being spent on data centers equipped with chips of the next -generation graphic processing unit (GPU) of NVIDIA, which are much more suitable for handling computing workloads associated with training and training chatbots and the agent programs.
Related: Bank of America renovates the price objective of Palantir shares
These programs are intensive in data, which require significant updates for pre -existing network servers and switches that connect them. As a result, the expense has been a blessing not only for Nvidia, which controls 90% of the AI ​​GPU market, but also server manufacturers, such as Dell and Supermicro, memory chips, such as Micron and Storage, including Seagate.
Today’s action reminds the flooding of IT expenditure at the end of the 1990s, at the Internet dawn.
That is worrying for Einhorn, who described the size of the dollars destined for AI with a word: “extreme.”
“The numbers that are being launched are so extreme that it is really very difficult to understand them,” Einhorn told Simplify Asset Management Panel.
The budget change of companies could be brilliant, but the justification offered by companies for disbursements has lately been “because we have to” that a carefully arranged return plan.
Einhorn believes that there is a “reasonable possibility that a large amount of capital destruction will pass through this cycle.”
That is shorthand for many of the things in which people are investing, it will not work.
The risk that companies burn money in AI projects are not covered by appetite expenditure yet. Recently, there has been a wave of mega high profile agreements that involve a lot of money from the best players.
Nor does it seem that hyperscalers are ready to delay their expenses in data centers. The Meta Platforms CEO, Mark Zuckerberg, aims to invest from $ 64 to $ 72 billion this year, and worries that “it is not aggressive enough.”
The market analyst for a long time, Beth Kindig, recently wrote in X:
It is projected that the server market of the AI ​​data center grows at a 38% TCAG from 2024 to 2029, reaching more than $ 580 billion, or 5 times from the $ 115 billion of 2024.
If you combine all the expenses made in AI, including the development of agent’s applications that one day could replace or increase many jobs, the amount spent this year totals $ 1.5 billion of $ 1.5 in 2025, according to Gartner.
Next year, Gartner says that the figure will be even higher:
Looking around 2026, a general global expense is forecast to exceed $ 2 billion, largely led by AI integrated into products such as smartphones and PC, as well as infrastructure.
This is not the first time that companies obtain mass spending to take advantage of technological innovation (and will not be the last).
More technology:
During the Internet boom, companies fired to invest in the necessary equipment to build the internet spine, causing actions in Intel, Cisco and many others.
“The real investment of IT was especially solid between 1995 and 2000, averaging 24 percent per year and adding an annual average of more than 3/4 percentage points for GDP growth,” according to the Bank of the Federal Reserve of San Francisco. “In 1990, the nominal investment in IT goods totaled only $ 131.5 billion, a little less than a third of the private investment of non -residential equipment and software (E & S). By 2000, IT investment had increased to $ 401.6 billion.”
This expense fed a lot of growth, but also created a bubble that, when it exploded, bankrupt many companies and, unfortunately, individual investors who threw caution in the wind, buying companies prior to income with profits instead of earnings.
“In 2001, IT investment contracted sharply, with a real investment of you falling almost 11 percent and nominal investment falling almost 17 percent,” according to San Francisco’s Fed.
The stock market fell as the expense collapsed.
2002: -23.4%
2001: -13%
2000: -10.1%
Even the actions that survived the retirement took years and, sometimes, for more than a decade to recover at their peaks on the Internet. For example, Cisco has not yet reached its March 2000 peak of almost $ 80, and Intel was needed until 2014 to eclipse its Internet boom peak.
If we see something similar this time it is not uncertain, but there is much speculation, and Einhorn is not too excited about economic perspectives.
“I am a little more about the opinion we are going to or that we have been in a recession,” Einhorn warned.
Related: The Blockbuster Nvidia OpenI agreement raises important questions
This story was originally informed by Thatleet on September 27, 2025, where it first appeared in the technology business news section. Add Thestreet as a preferred source by clicking here.
