Strong Earnings and Broadening Capex Keep Us Constructive

We remain constructive on risk assets. We stay overweight U.S. large cap equities. Thematically, we favor the AI supply chain, cybersecurity, energy infrastructure, and natural resources. We remain underweight duration, and we hold gold as a partial duration replacement.

Prices moved over the past few months. Our theses did not. Fundamentals continue to back our thematic work; earnings are exceeding lofty expectations, AI demand remains robust, and infrastructure investment remains elevated across sectors.

Gold has outperformed long duration bonds since we first added it. Since June 2025, gold has gained 30.9%, while long duration bonds fell 5.0% in price and returned roughly 0.1% including coupons. Since we added more gold in December, gold has gained 4.8% while long duration bonds fell 6.6% in price and 3.7% including coupons1.

We apply the view across three horizons. In the short term, we would deploy excess cash during equity market pullbacks and add to gold where clients have appetite. In the medium term, we would build thematic positions into volatility. In the long term, we would maintain the U.S. large cap overweight and hold allocation discipline.

Bottom line: Our core views have not changed, and data continues to support our investment theses. We would maintain the U.S. large cap overweight, hold gold as a partial duration replacement in a world of higher fiscal spending and sticky inflation, and put excess cash to work on weakness. 

Q2 results reinforce our constructive U.S. equity outlook. Roughly 94% of reporting S&P 500 companies met or exceeded EPS estimates, the highest post-pandemic beat rate in our data2.​

The growth number matters just as much. Companies delivered 46% EPS growth against 21% expected, roughly twice the forecast3. This is not a multiple expansion story.​

Revenue growth, earnings growth, cloud growth, and AI demand remain robust. We watch that distinction closely, because fundamental delivery, rather than sentiment or valuation, is doing the work.​

Cloud results illustrate the demand backdrop. One large hyperscaler grew cloud revenue 37% year-over-year toward a $200 billion run rate, and its custom chip business now exceeds a $25 billion annual run rate. Management framed the spending as demand-led rather than speculative and noted that AI demand continues to exceed available capacity.​

Taken together, these results support our confidence in the U.S. large cap overweight and our preference for AI-related themes that range from the supply chain to cybersecurity.​

Bottom line: Companies delivered real earnings growth rather than a valuation rerating, and broadening AI demand sat behind much of it. That combination argues for staying invested in U.S. large caps and in our preferred AI related themes. ​

Capital spending has moved beyond semiconductors and memory. In Q2 2026, 51% of S&P 500 companies grew capex by more than 10% year over year, above the long-run average near 41%. That level of breadth has been rare over the past 25 years.​

Breadth matters because one company’s capex becomes another company’s revenue. As the spenders broaden, the beneficiaries broaden with them. The story increasingly centers on the infrastructure that supports AI deployment rather than on the chips alone.

We see four areas of potential opportunity. The AI supply chain covers data centers, networking, and compute infrastructure, where demand still runs ahead of available capacity. Energy infrastructure covers power generation, transmission, and grid expansion, as rising AI workloads lift power demand. Natural resources cover metals, mining, and materials that historically benefit from higher capital spending. Cybersecurity rounds out the list, because AI adoption expands digital attack surfaces, and, in our view, the market underappreciates long-term enterprise security budgets.​

Early cybersecurity earnings back that last point. Vendors reporting so far have posted meaningful acceleration in both revenue growth and backlog, including some of the largest top-line beats in years. Customers appear to be funding this with incremental budget rather than cannibalizing existing IT spending, and the dollars are moving toward identity, data security, zero trust, and platform consolidation. Hyperscaler results and commentary tie security spending directly to enterprise AI adoption, which supports our conviction in how long this growth can run. Results are not uniform, though. Legacy security assets have delivered mixed numbers, so positioning within the theme matters. Most cyber vendors report off cycle, so we will keep testing this thesis as those results land.​

We also watch four risks. Inflation could reaccelerate and push global yields higher. The Middle East conflict could escalate materially and pressure energy and food prices. Growth could slow meaningfully and entrench stagflation, leaving the Fed behind the curve. AI investment could moderate if infrastructure spending falls short of expectations.

Volatility comes with the territory. Since 1950, markets have delivered 3% drawdowns roughly every seven months, 5% drawdowns roughly every 13 months, and 10% corrections roughly every three years4. If earnings, AI demand, and capital investment stay supportive, we would stay invested, deploy excess cash, and build thematic exposure during pullbacks.​

Bottom line: The AI build out now reaches infrastructure, power, resources, and security, which widens where clients can participate. Risks remain in this fundamentally robust environment, so we would treat normal pullbacks as entry points rather than exit signals. ​