Wall Street ended last week looking calm on the surface. Oil prices were mostly steady and stocks barely moved. Underneath that calm, however, bond yields climbed to multi-decade highs in several countries as central banks continued to raise interest rates. Investor and portfolio manager Dan Niles believes this could make for a difficult week ahead. In a note on X, Niles said investors should approach the markets with caution.
Three rules are flashing red
Niles follows three simple rules in markets: “don’t fight the Fed, don’t fight the bond market, and don’t fight seasonality.” Last week, he said, all three pointed in the same direction. The global rate hiking cycle appears to have restarted, according to Niles. Bond yields have reached multi-decade highs across several countries. According to Niles, September is also historically the weakest month for stocks. Years in the middle of a US presidential term have also tended to be difficult for markets.
Niles uses these three factors when deciding how much leverage to use, how much of his portfolio should be in long or short positions, and how large individual trades should be. “Investing is hard enough without fighting headwinds,” he wrote on X. He also pointed to Warren Buffett, whose 61-year run at the top of Berkshire Hathaway ended last week.
Niles recalled one of Buffett’s own lines, “the market has to keep pitching but you do not need to swing.” Investors do not have to act on every market move, especially when several risks are pointing in the same direction.
AI industry splits over safety
The market caution comes as the artificial intelligence industry is also facing a debate over how quickly new AI models should be released.Niles described a growing divide within the industry over AI safety and regulation.
He said OpenAI and Anthropic are pushing for regulatory capture, meaning rules that could favour companies that are already ahead of their competitors. Nvidia and Microsoft, on the other hand, are pushing for more testing before new AI models are released. Niles supports the second approach. “No product from any company should be released before it is safe,” he said. He also argued that a slowdown in AI model training does not necessarily have to be bad news for the industry.
More focus on security, greater spending on testing and a shift in hardware demand towards infrastructure rather than pure model training could continue to support AI-related stocks, even if the pace of model development slows.
He says, last week’s market action appeared to support this view. The SOX semiconductor index, which is often used as a proxy for AI infrastructure, rose 0.8%. The software ETF IGV gained 2.8%. Cybersecurity stocks performed even better. The HACK cybersecurity ETF jumped 8.0%, compared with a 0.1% decline in the S&P 500.
Cheaper AI could mean more demand
Niles believes most companies will not need the most advanced AI models for most of their everyday work. Instead, he expects more users to turn to open weight models. These are AI models whose underlying model weights are made available for others to use, modify or build on. If that trend continues, the language model layer could eventually become more like a commodity, with many models offering similar capabilities at lower prices.
There are already signs of this happening. Niles pointed to pricing data showing that the average price per token peaked in late May and has since fallen by roughly 50%. At the same time, token production has increased roughly fourfold. He described this as Jevons paradox in action. The idea behind the paradox is that when something becomes cheaper to use, people end up using much more of it.
In case of AI, this could mean lower prices do not necessarily reduce demand for computing power. Instead, cheaper AI could encourage companies and consumers to use AI more frequently, increasing the need for data centres, chips and other infrastructure.
Who could benefit as AI models become commodities?
As AI models become more interchangeable, Niles expects four factors to become increasingly important. Companies will need strong open weight models, wide distribution, large amounts of training data and a core business that generates significant cash flow on its own.
Niles pointed out three companies in particular. Meta could benefit from the strength of its open weight models and its large distribution network. Niles also said Google is likely to release a new frontier-level model soon, which he believes could help its stock.
Microsoft could benefit from the growing use of Copilot among its enterprise customers. Niles described Copilot as the “safe” choice for roughly 450 million Microsoft 365 enterprise users. He also flagged a concern for Microsoft. The company owns a 27% stake in OpenAI, and it has a smaller pool of proprietary training data compared with Meta and Google, according to Niles.
AI agents could keep chips in demand
Another major driver of AI infrastructure demand is agentic AI.These are AI systems that can carry out several steps of a task on their own instead of simply responding to a single question.
According to Niles, agentic AI systems drove token production up by 10 to 100 times earlier this year. He expects this increase in usage to continue supporting demand for semiconductors and other companies that provide the infrastructure needed to run AI systems. This could create an interesting split within the technology sector. The AI model layer may face falling prices and increasing competition, while the infrastructure and security layers could continue to see strong demand.
Geopolitics adds another risk
Markets also face a fresh geopolitical risk after a series of attacks over the weekend. Ukraine launched more than 1,000 drones at Russia, including what was described as the largest-ever attack on Moscow. A missile also struck Riyadh, marking the first such attack on the Saudi capital since July.
Niles warned that both events could affect oil prices. Higher oil prices could then push bond yields higher, adding another source of pressure at a time when rates are already rising and bond yields are sitting at multi-decade highs. This would add to the risks Niles has already identified through his three rules: the Federal Reserve, the bond market and seasonality.
Trump-Xi meeting gives one bright spot
There is, however, one event on the calendar that Niles is watching with some hope. US President Donald Trump and Chinese President Xi Jinping are scheduled to meet on Wednesday alongside business leaders from both countries.
Niles said he is hopeful that something constructive comes out of the meeting. For investors, the week begins with several forces pulling in different directions. Interest rates are rising, bond yields are elevated and September has historically been a weak month for stocks.
Geopolitical tensions could add pressure to oil prices and yields. At the same time, the AI market is undergoing its own transition, with cheaper models potentially driving greater usage, while infrastructure and cybersecurity companies continue to benefit from rising demand. Niles is not calling for panic. He says to deal with things cautiously and with discipline.
Disclaimer: This article provides factual analysis only and is not, and should not be construed as, an offer, solicitation, or recommendation to buy or sell securities. Investors must conduct their own independent due diligence and seek advice from a registered financial advisor in the respective jurisdiction.