Skeptics of the artificial intelligence revolution have recently had the upper hand.
In July, the semiconductor sector of the U.S. stock market began to underperform; chip stocks fell precipitously, down 20% and more. Investors started to demand clear evidence that spending on AI hardware would generate sufficient financial returns to justify investing trillions of dollars. Profit-taking and extreme overweight positions amplified the selling. Talk of an AI bubble began to emerge. Selling intensified in the first half of this month. By Oct. 15, the major semiconductor exchange-traded funds had declined for eight consecutive sessions with drawdowns of around 12%.
This week, however, with Anthropic’s filing of its so-called S-1, a necessary document for a company to go public, a degree of optimism is returning to investments in the AI complex. Stocks in the AI ecosystem rallied, especially semiconductor names such as Micron and Nvidia. In its S-1, Anthropic details an extraordinary growth rate coupled with strong operating margins. At the end of its June quarter, Anthropic reported annual revenue growth of over 1,400%, with a positive operating margin of 19%. Anthropic plans to go public around Nov. 16, with a valuation of somewhere between $1.5 trillion and $2 trillion. Anthropic acknowledges that it plans to invest another $800 billion in its operations.
So the million-dollar question is when Anthropic and the other American hyperscalers — Amazon, Google, Meta, and Microsoft— will start generating major capital returns on the trillions of dollars invested in AI?
From a basic capital-investment standpoint, the hyperscalers are currently in the most intensive phase of cash burn, where investment exceeds returns. For example, Meta just announced it is raising about $95 billion in new capital to help fund its planned investment in AI, and Microsoft is offering buyouts for 60,000 of its employees as it struggles to invest in AI and protect its balance sheet. The general consensus on Wall Street is that by the end of 2027, incremental cash flow from basic AI services will begin to exceed the marginal growth in new investment outlays.
But earning necessary returns on capital (mid-teens and higher) from the almost $8 trillion that will be invested in AI over the next five years will probably not occur until early in the next decade. The major U.S. investment banks project that by 2030-2031, the hyperscalers will begin generating cumulative operating profits exceeding $1 trillion, which would translate into about a 12%-14% return on capital investment. So, within five years, the current aggressive AI investment theme will transition into a highly cash-flow-positive business operation.
Wall Street traders are not noted for their patience, but retail investors should be reminded of the history of Amazon. Today, Amazon has a market capitalization of about $2.7 trillion with a stock trading around $250. Adjusted for a 20-for-1 stock split in 2022, Amazon sold at 30 cents a share in September 2001. The stock had traded as high as $106 in December 1999, but then Wall Street gave up on Amazon’s aggressive investment strategy that generated very low returns on capital.
HERE ARE THE STATES WITH VOTER ID ON THE BALLOT THIS NOVEMBER
However, investors who believed in Amazon, bought the stock 20 years ago, and did not sell, would have realized stock gains of over 9,000%. In other words, a $1,000 investment in Amazon 20 years ago would be worth more than $90,000 today.
On Wall Street, patience is often an underrated investment skill.
The writer owns shares in Google, Micron, and Nvidia.
James Rogan ([email protected]) is a former U.S. diplomat who later worked in law and finance for over 30 years. He writes a subscription-based daily note on markets, economics, politics, and social issues.