Ketu Desai
By Ketu Desai
September could be remembered as the month when two important mistakes were made. The first by the Fed, which raised rates with market odds that this will be a hiking cycle. While the hike helps the new Fed Chair with credibility, the hike is unlikely to have a meaningful impact on inflation.
This inflation problem has created a supply problem of oil, food, and other commodities due to two wars. Supply of semiconductors and other AI infrastructure. Supply of housing. Rate hikes won’t solve these issues but instead raise financing costs that prevent new supply from coming. These hikes also flatten the yield curve, which hurts banking activity, especially for small and mid-sized regional banks.
Breakeven rates across the curve have moderated, after peaking in early May. The 5YR breakeven rate is 2.3%, not far from the 2% target. Core PCE (Personal Consumption Expenditures) over the last three months is running at a 2% target and expected to decline meaningfully next year. One-time factors from portfolio management, software, war and tariff impacts peaked this summer and are expected to roll off. As oil and diesel cool off, it will have a meaningful impact on inflation.
The stock market is often the best economist, sending a clear message that hikes are crushing housing, retail, banks, transports, and consumer discretionary. Even large businesses such as Walmart, Costco, and McDonalds are down nearly 20% from their usual profits this time of year. Other discretionary services such as cruises, restaurants, homebuilders, and retailers are down even more.
The second mistake seems to be frontier AI lab executives making doomsday predictions. These warnings could be for commercial reasons such as regulatory capture, liability protection, and burden of care. It is clear that they are losing market share to open-source models. Anthropic is losing public trust very quickly, with recent polling showing that just 5% of people trust Anthropic with their passwords. Many corporations will be hesitant to hand over their data to Anthropic. Latham & Watkins, the nation’s second-largest law firm, bought its own Nvidia GPU servers and started customizing its own AI models not using OpenAI or Anthropic. The firm’s chief information officer said they don’t want to put their client information on any cloud vendor’s platform.
Tighter monetary policy, distrust of frontier labs, and the release of Meta’s Muse is driving investor capital to large cap tech names. Narrow breadth has benefited market cap weighted indices and to the detriment of equal weight and smaller caps. The market cap weighted S&P has outperformed both equal weight and small caps by nearly 5% over the past month. Investors prefer the secular growth of large cap tech and AI infrastructure when the Fed is in a hiking cycle. An additional headwind for many of the names outside of tech is that they could be a candidate for disruption from consumer AI agents.
Names across insurance, travel, banks, brokerage firms, and retail sold off upon the release of Muse. Valuation for the market cap weighted S&P is not demanding as gains for the year so far are entirely through earnings growth. The multiple is down 13% for the year, while margins are near 17%, an all-time high. The ratio of P/E to margins is at the same levels of major bottoms such as in 2008, 2011, 2018 and 2022. The PEG ratio is at its lowest in 30 years, meaning you are not paying as much for earnings growth.
Frontier labs and companies such as Google and Meta will battle to be the preferred consumer AI agent. The ultimate winner will be the semiconductor companies. There simply isn’t enough hardware to fully utilize consumer agents for large user bases such as Meta. Bank of America expects the semiconductor industry to double by 2030. Citi expects the memory undersupply to last until 2031. Goldman expects the hyperscalers to increase CapEx by 54% in 2027, a significant amount of which will go to semiconductors. Analysts expect semiconductor earnings to grow 73.5% in 2027 with 51.2% margins.
Korean customs data offers the earliest and cleanest reading on global technology demand. The latest semiconductor exports surged 259.4% from a year earlier. Semiconductors continue to trade below a market multiple as most trade is below a 1.0 PEG ratio. Alternatively, shareholder return of capital is high. Nvidia announced its largest buyback in history. Expect a large one from Micron in December. Positioning has been cleaned up from the Situational Awareness led sell-off over the summer. Semis have built a beautiful base they are ready to break out from. The bottom line is that AI buildout has reached its largest capital expenditure in history and semiconductors are one of the biggest beneficiaries.
Looking forward, the market will focus on earnings, geopolitics, and a Fed meeting.
Ketu Desai is the Principal of i-squared Wealth Management Inc. (www.isquaredwealth.com), an investment management firm based in New Jersey. ketu@isquaredwealth.com
