DAILY STRATEGY: Earlier this week, we shared investor sentiment from a dinner we hosted (HERE). Yesterday, we hosted a larger investor breakfast, and the message from both events was similar: investors are constructive on the AI buildout, but their confidence rests on AI lab revenues (Anthropic, OpenAI and the others) reaching the roughly $400bn-600bn ARR range exiting 2027. And getting SOLID information on AI Lab ARR trends is critical. Anthropic IPO should be helpful in this regard.
Some Background on Why This is So Important – At $400-600bn ARR, 2029/2030 AI related capex plans are likely to come to fruition. Currently, the stocks most levered to AI related capex are getting very little credit for 2029 and beyond AI capex according to 22V Data & Infrastructure analyst Dauvin Peterson. Hence the significant compression in PEs for many AI related buildout baskets. We have recommended being long a basket of Dauvin’s “coiled springs” AI buildout names based on the idea that the $400-600bn ARR will be achieved (see my conversation with Dauvin HERE). These names have witnessed PE compression at the same time EPS expectations have increased meaningfully. See more HERE.
All the above helps explain the severe sell-off in AI related stocks AND 10yr yields yesterday. It was reported that OpenAI 2026 ARR was tracking $50B ARR vs $70B previously signaled according to the FT. Two reactions fell from that headline, AI buildout baskets and Semis sold off as any downside in revenue from one of the labs puts the $400bn-600bn in ARR required to support AI capex at risk. Additionally, downside risk to capex expectations longer term = downside risk to longer term GDP growth and inflation expectations. All things equal.
Capex expectations need to hold up (chart below courtesy of Dauvin Peterson) for AI buildout baskets to move higher. Our call is capex expectations will hold through 2027. Also, it turns out, at least according to Bloomberg, the $50bn ARR from OpenAI may not be correct. From the Bloomberg article. “OpenAI is expecting to reach or exceed $70 billion in annualized revenue by the end of the year, driven largely by growth in its enterprise business… The company’s annualized revenue was roughly $50 billion at the end of September, which represents a projection of OpenAI’s yearly sales based on a shorter period.”
More Survey Results – Dinner attendees saw economic growth as sustainably strong despite 10yr yields being near 5.3%, and the breakfast survey is consistent with that: 62% see the 10yr at 5.25-5.50% over the next six months and another 15% at 5.50-5.75%. Only 23% expect yields below 5.25%. The S&P 500 view was more divided than usual: 46% expected the S&P to be up 5-10% and another 8% expected more than 10%, but 38% expected it down.
On AI, 62% see mega-cap tech/AI leading and 23% pick broader AI infrastructure, while industrials, consumer, healthcare and energy/materials drew zero. As discussed in the conversation with Dauvin, (HERE) no one expects compute oversupply in 2027, and 54% say later than 2030. That is more bullish than equities imply, since multiples across the group are compressing as if compute turns long by 2029. It also fits the dinner’s central point that the trade needs $400-600bn of ARR by end-2027 and a path to roughly $1tn in 2028-29 to earn a 15% ROIC on the $1.4tn 2028 capex year. Missing that ARR marker is the 2029 capex “cliff” risk that investors worry is may be holding down multiples.
Risks – 54% said price competition from Chinese AI models colors their view of the US capex outlook, which feeds the worry that token prices fall faster than volume grows. If revenue falls short, financing $1-2tn of capex at sub-15% returns will get much harder with yields rising. Dauvin has moderate confidence in the revenue targets today, mostly because disclosure is limited, and expects that to improve once Anthropic’s S-1 shows realized pricing. Until then, the trade works as long as ARR tracks toward the marker and yields stay contained.
Charts…
~54% of investors expect the S&P 500 to move higher over the next 6 months. ~38% expect the S&P 500 to move lower. This is unusual dispersion from our surveys.
Investors expect the 10y to be in the 5.25%-5.5% range in the next 6 months.
Market leadership is skewed towards AI.
Investors do not see compute oversupplied until 2031 and later.
Investors believe Anthropic will have a higher market cap than SpaceX at the end of 2028.
Investors see price competition from Chinese AI models affected their view of the US AI capex outlook.
