Fynsa
U.S. equities have traded more volatile, particularly investments linked to Artificial Intelligence (AI), amid a well-valued market and recent questions about whether large AI investments can pay off in a timely manner and how CAPEX can continue to be financed when cash flows dry up and require large debt issuance, as is already occurring. when cash flows dry up and require large debt issuance, as is already happening.
As of last week, the S&P 500 went on to correct nearly 5% from its October highs (to recoup much of those losses this week). "Bubble" headlines abound in IA, but we have no way to conclude that, with corporate results still healthy, earnings continuing to be revised upward, an economy that continues to grow, expansionary financial conditions, and a Federal Reserve with some room to continue lowering interest rates.....
However, below the surface, the market has been "cleaning up" or reversing excesses.. The S&P 500 adjusted by 5%, but we have more speculative assets that have done much more. For example, Bitcoin has already accumulated a 30% drawdown, an unprofitable technology index has adjusted 24%, and a basket of retail investor favorites, many of them linked to AI (highly speculative flows and already representing more than 30% of the market), has fallen almost 20% from its October highs.
Some will say that the overall market then has room to adjust further. Maybe so, and ideally we would like to see a correction that allows us to buy at more reasonable valuations. But beyond short-term volatility, we still have a constructive medium-term view, as the market is differentiating quality and reversing excesses in assets with poorer fundamentals, and that will limit seeing much larger declines than we have seen so far.
That's on the equity side. The debt issue is important, because we see everyone worried about equities tied to the AI investment cycle, but the key may be in credit..... Capital spending on AI data centers is being funded by a flood of investment-grade loans and issuance that have skyrocketed.
Known IA big Tech issuance this year is already approaching US$150 bn. Net GI debt issuance in the US is forecast to increase 54% next year to US$802 bn. Also, 2026 is expected to be a record year for gross supply, with an expected US$1.81 trn (US$400 bn just to fund the AI investment boom).
According to JPM, the development of the global AI data center infrastructure, as well as the related power supply, could cost more than US$5 trn.. Annual data center financing needs in 2026 are around US$700 bn, which could be fully funded from hyperscaler cash flow and the bond market.
However, the financing needs for 2030 exceed US$1.4 trillion, which will likely require contributions from all capital markets.1.4 trn, which will probably require contributions from all capital markets.
Since 2021, the hyperscalers have collectively increased net debt on their balance sheets by US$295 bn, but their collective net debt/EBITDA leverage remains at just +0.2x. These companies could add US$700 bn of net debt to their balance sheets without their net leverage exceeding 1x. While the largest AI infrastructure companies enjoy strong balance sheets, leverage and cash flows are more pressing challenges for many of the other public and private companies involved in AI infrastructure development.
Thus, the recent pace of growth in both debt and alternative financing has raised concerns among investors.. The feedback loops created by the revenue and equity relationships between some of the largest U.S. public companies and smaller AI companies raise the risk that stress in one part of the AI ecosystem will affect investors across the complex. The important thing, then, is to do proper risk management and favor quality assets.
Finally, with respect to "circular risk" Capex and Monetization. The AI ecosystem has become a closed loop where capital, compute and infrastructure feed back into each other: OpenAI concentrates the demand, Nvidia provides the compute power, and Microsoft and Oracle fund and operate the cloud that makes it possible.
The major concern is the monetization of these investments. To achieve a 10% rate of return on modeled investments through 2030, approximately US$650 billion in annual revenues would be required in perpetuity, according to JPM estimates. Sources warn of the risk that the revenue curve may not materialize at the rate that justifies the investment, citing the telecom and fiber boom as a possible historical parallel of what could go wrong.
A final message to our clients.... In most cases, a well-diversified, multi-asset portfolio is the best way to deal with market risks, and investment strategy should continue to be tailored to personal timelines, liquidity needs and risk tolerance levels.
