Skip to content

Dan Niles

Dan Niles: The AI Trade Has a Cisco Problem (September 18, 2026)

Link.

Dan was in the tech industry in the 90s, then went to the sell side, and now is an investor.

He doesn't want to fight seasonality, the Fed, or the bond market.

Yields are at 5% and going higher, but the stock market remains at all-time highs.

AI has powered the market this year. The semiconductor index is up 60% this year. This is a relatively small sector, but has been driving the S&P 500 returns. This semiconductor growth is related to data centers.

Three concerning developments:

  1. Texas wants to slow down data center development

  2. Last week we saw Sam Altman (OpenAI), Dario (Anthropic), and Elon Musk say we need to slow the pace of innovation.

  3. Customers are saying "we've been using a Ferrari to go to the corner store to get milk." They are switching from the expensive models to the cheaper open source models.

The cost per token is down 50% since May. The number of tokens produced has quadrupled. The open source models are where the new volume is accruing.

A quarter point raise doesn't do anything. Inflation has been running above the 2% mark for 66 months now. The market has historically underperformed when the Fed is hiking, that's where the phrase "don't fight the Fed" comes from.

All of these data center companies are now competing with the government for capital.

Unless you invested thru 2007, then you can believe that "buy the dip" always works.

Central banks and governments are running out of ammunition.

A risk-free 5% return at the current valuation levels may be a good investment.

The last Fed decision to raise rates 25bp was unanimous. There are rate hike cycles happening at the ECB, Japan, and the UK. This is happening across the globe.

The socialist democrats are picking up a lot of seats in recent elections. By definition, they are anti-big business and anti-data centers. If they get a lot more seats, will this be bad for data centers?

He doesn't think we're at the top yet because the agentic AI trade just started in January.

If CAPEX slows then the AI trade can unwind. In 2001 and 2002, the internet was growing massively, but the dot coms still crashed.

He believes the AI is the most transformational technology we've seen since the internet. These companies are all overinvesting because all these companies know they will make a fortune if they're the dominant AI company. But there will also be train wrecks along the way.

He thinks we have another year or two before we hit the top.

He thinks Google will eventually be the winner for the consumer sector. The dark horse is Meta. They have 3.6 billion daily active users. They could create a large user base from their current platform.

He thinks one company will be dominant. Look at other areas of the internet:

  • Social: Meta
  • Streaming: Netflix
  • ecommerce: Amazon
  • Google: search

The #2 companies in these areas are far, far behind.

He's very bullish on Google. They have a lot of data to train models on. It wouldn't surprise him to see Google at the frontier of models in the near future.

The circular financing will only start to matter when things start to slow.

He doesn't think any of these companies are going to slow, except for a few quarters before an IPO to make cash flow look better.

He thinks the iPhone Duo will be a great selling phone by the end of 2027.