Download App

[US AI Development Slows Down, Market Reactions Diverge] ⑴ Voters clearly put the brakes on AI development this week, but large tech company stocks are still rising. Gains in companies like Microsoft and Amazon reflect Wall Street's growing recognition that a slowdown in the AI race may not be a bad thing. ⑵ If companies can postpone massive spending commitments, they can demonstrate healthier cash flow, avoid financing costs, and potentially circumvent high-risk investments. ⑶ Large tech hyperscale cloud service providers were among the big winners in Wednesday's post-election relief rally, with Microsoft rising about 7%, Amazon about 6%, and Alphabet about 3%. ⑷ These companies generate substantial cash flow through businesses like advertising, cloud computing, or software and don't necessarily need to rapidly transform into AI companies. ⑸ Some commentators point out that while the delays are frustrating, they also reduce the risk of trillions of dollars invested in equipment becoming obsolete or depreciating faster than financiers anticipated. ⑹ Not all hyperscale cloud service providers rose; Oracle fell nearly 2% on Wednesday, suggesting that the challenges faced by smaller AI companies may eventually spread to larger companies. (7) Related challenges may include slower sales growth for companies providing the “shovels” of AI transactions, such as fiber optic connectors, servers, and storage devices; the market appeared to be pricing in this outlook on Wednesday. (8) Engineering and construction companies closely linked to AI construction saw particularly sharp declines, including Quanta Services, Emcor, Vertiv, and HVAC contractor Comfort Systems. GE Vernova, which provides power plant turbines for AI computing, was also under pressure. (9) Smaller data center operators also suffered on Wednesday. These companies have borrowed heavily to build data centers and purchase semiconductors; a slowdown in construction means delayed revenue during debt repayment. (10) Democratic victories at the state level and dozens of referendums appear poised to slow data center construction. JPMorgan analysts noted that most construction in 2027 may be unaffected, but 37.5 gigawatts of planned capacity in 2028 could face suspension, and another 15 gigawatts could face “conditions.” (11) These conditions could include requiring data centers to pay grid costs to protect electricity bill customers, eliminating tax breaks, restricting water usage, setting on-site backup power requirements, imposing labor requirements, and prohibiting the use of confidentiality agreements during the approval process. 12. Some economists and investors believe that slowing down the AI capital expenditure boom may be generally beneficial to the economy from a macro perspective, helping to alleviate the risk of overheating in the next 1 to 2 years and potentially extending the AI investment cycle.