DAILY STRATEGY: 10yr yields were flat yesterday despite lower oil prices and 1yr inflation swaps. The 10yr-2yr curve steepened slightly while longer-term inflation expectations held steady — consistent with an improving real growth outlook as the energy shock headwind fades.
Why yields didn’t rally: consumer spending has remained strong during the energy shock. Banks flagged at last week’s financial conference that spending is holding up, with some improvement at the lower end (HERE). With the Energy shock fading, real incomes will move higher and downside risks to consumer spending are fading. We don’t expect the real income lift to fully offset fading fiscal impulse, but that will take time to play out. So, our base case is yields stay rangebound (4.4-4.5% range) until there’s more clarity on how much growth slows. Or not.
Assuming the economic outlook keeps 10yr yields below ~4.5%, which is our call (with lower conviction now unfortunately HERE), fundamental factors (EPS momentum, Earnings Growth, GARP) and Cyclical sector will be supported.
TOKENMAXXING UPDATE: Two weeks ago, we wrote about investor concerns that AI demand is “inflated” at the same time value creation is unclear. The concern is large AI bills without monetizable productivity gains. The AI buildout baskets were perceived as at risk because of their sensitivity to usage statistics and/or token pricing (especially after 50%+ YTD price gains).
Since then, token prices dropped and buildout beneficiaries (memory, compute, power generation, liquid cooling, neo clouds) fell initially but have now pared losses.
Dauvin Peterson, head of 22V Data Infrastructure/Commodities, wrote last week how lower token prices encourage firms to adopt AI, so lower prices are not a long-term bearish signal for the buildout, even if tokens become “commoditized” (HERE). Investors may be pulling forward that view. It was one of the most asked questions we got last week. Employees can go back to testing AI tools more economically. The effectiveness of these tools is still a risk (more on that below), but at least the tools are cheaper again.
During the great AI buildout drawdown of June 2026, the baskets of AI Disrupted stocks stabilized but did not outperform to near the same extent. Investors were NOT reconsidering whether AI as a technology is effective. The terminal values of companies disrupted by AI were still perceived as at risk. That implies the fundamental thesis for the AI and AI buildout trade are intact.
Price Mo is sensitive to inflation and supply shocks, given inflation is the limiting factor on the expansion and Price Mo is the first derivative trade in a VAR shock/broad derisking. Momentum began its drawdown when labor data was stronger-than-expected, deepened when the US and Iran resumed missile strikes, and then began pared losses when peace talks perceivably gained traction late last week.
The Most Important Longer Term Demand Driver Is Enterprises Making More Money with AI – We continue to track what companies are saying about their AI deployment and its impact on margins/productivity. There is no update here. 2Q earnings reporting season will be important. What we know is companies that use AI are more optimistic about their forward outlooks for margins (across sectors, not just Tech) than companies that do not. The early quanitifiers in Q1 guided to ~85bps of margin improvement from AI tools (HERE). It is difficult to be short Price Momentum unless margin sentiment declines, the quantifiers guide to worse margin improvement, and/or no additional companies quantify the impact from AI next quarter.
Recall, the negative argument for AI Usage (and by extension the AI demand outlook) is as follows – It is possible that recent margin expansion/productivity gains have VERY little to do with AI. As Peter Williams pointed out in a report last week, when accounting for the fact that businesses are simply working their existing workers and machines harder — which naturally happens after an economic slowdown, the recent productivity gains appear to be driven by a surge in capital and labor utilization rather than AI. Theoretically, this is a temporary lift to margins/productivity.
Also, the immigration slowdown could be temporarily boosting productivity now. Fewer low-wage immigrant workers means total work hours are dropping faster than economic output, which makes productivity look better for a time, but should fade (HERE).
All of the above suggests AI is not really the driver of recent margin gains. If companies are struggling to find value using AI tools, the demand outlook will be much less certain. The ROI assumptions are more at risk.
Charts…
Momentum baskets have pared losses. AI buildout baskets have pared losses.
During the drawdown, names that face the greatest risk from AI stabilized but did not outperform to the extent that the buildout baskets underperformed.
The LLM Token Expenditure Index (BBG: SDLLMTK) is a daily financial benchmark that tracks the effective aggregate spend on large language model (LLM) APIs rather than just nominal price per token. It measures how much enterprises are spending on AI inference.
The current level of margin sentiment suggests an INCREASE in margins in 2Q26. Unless margin sentiment declines or actual margins decline, it will be difficult to be short Price Momentum.
Companies that quantified the impact from their AI tools have much higher earnings sentiment than the rest of the index. Companies that mentioned use cases of AI, but did not quantify the impact, also have higher margin sentiment than the index.
Consensus margin estimates are higher for AI users as well. That is NOT just concentrated in Tech.
