Market backdrop
Oil jumped over $104 a barrel. The 10-year bond yield sat just below 5%. All eyes are on the Fed, which announces its rate decision Wednesday afternoon, when chair Kevin Warsh speaks to investors.
A large debate is running in the AI trade. Over the past week an Anthropic researcher left, and Dario Amodei's letter came out over the weekend. Nvidia (NVDA) CEO Jensen Huang spoke at a Goldman Sachs (GS) conference Thursday and said he did not find that researcher's claims credible. The Trump administration appears uninterested in stepping into the regulatory role Amodei asked it to take.
An AI growth scare has been building for some time - rogue agents escaping, a Hugging Face hack. Sam Altman agreed with Amodei, saying the industry needs to pace the frontier, with more to say soon. That talk pulls a lot of the tech trade down, though not all of it, and could slow economic growth.
AI capex versus model pacing
Much of the GDP growth the country has enjoyed comes from the AI data center capex investment cycle. AI is also a big driver of inflation today, through memory shortages that push up iPhone prices and electronics prices across the board.
There is a knee-jerk reaction for AI data center capex beneficiaries to fall after the weekend news, visible in pre-market trading. Investors should separate two things in their minds: pacing AI model improvements, which is what is being proposed, and actual data center capex spending. The gating factor on capex is not model quality. There are years of improvements, product development, and research still to come before the models already on the market can be used at full capacity. So pacing model quality has no direct effect on AI data center capex over the next couple of years.
This morning's market reaction may be a smaller version of the DeepSeek response from early 2025, when Nvidia (NVDA) and many other companies fell on the launch of a new Chinese model that scared people at first. A month or two later, it was clear the worry was over close to nothing.
Some Mag 7 names held up with gains this morning while many other names fell 5% to 8%. A Microsoft-backed OpenAI agent was linked to a cyber attack.
Microsoft (MSFT)
I am a fan of Microsoft (MSFT). Its legacy software business should be very durable and keep growing. The company will build AI features into its products. Its deep ties with customers span dozens of products - productivity apps, security tools, cloud storage - and its huge capital base and ability to invest should make it a successful company over time.
If the industry paces the frontier, that is incremental good news for legacy software firms like Microsoft (MSFT), Oracle (ORCL), ServiceNow (NOW), and IBM (IBM). Even setting that aside, Microsoft is a very good, well-managed business with good growth prospects, and the valuation is attractive. I like it today, and it is a top 10 position in the fund.
Cintas (CTAS)
A well-balanced portfolio needs beneficiaries of the AI trade - capex plays or productivity tools that use AI to improve their products - and some defensive balance. Cintas (CTAS) provides that balance. It is a uniform rental company known for blue-collar industries, and it also supplies facility maintenance and fire and safety protection equipment to many of the same customers. Cintas is consistently gaining market share and is the biggest player in its industry. It is buying UniFirst (UNF), one of its larger competitors; that deal should be nicely accretive if it closes in the next couple of months. It is a well-managed business and an attractive stock for a balanced portfolio.
How to position
Steer your portfolio toward higher-quality, more profitable, more durable businesses. To get AI data center capex exposure today amid the bad headlines, Nvidia (NVDA) is probably the place to go. It has the most compelling valuation in that group and should gain market share of the large capex the big hyperscalers are spending. Quality is attractively priced across every industry right now.
On Amodei's letter
Amodei's letter was very reasonably and very well written. He has information the average investor does not, so his view matters, along with others like Jensen Huang's. But consider his track record on long-term calls: not long ago he predicted a 10% U.S. unemployment rate and roughly 25% of white-collar workers losing their jobs. We are nowhere near those outcomes. Predictions five, six, seven years out are extremely tough to get right after 20-plus years of doing this. I would not base investment decisions solely on that letter today. Today's models are already very capable, and people will keep finding new uses for them for years. So I am not terribly concerned about the AI buildout because of the weekend letter.
Fed call
The market now thinks the Fed will raise rates this week, a shift from a week ago. Inflation data has come in hot, and labor data has been fairly good. It seems like a bad time to raise rates ahead of an election. The person who nominated the chair wants rates to go down and says so loudly. It will be hard for the chair to defy the market and not raise. I expect a quarter-point (25 basis point) rate increase Wednesday afternoon.
