Capex Broadens as Labor Holds and Themes Converge
Weekly Market Update
Takeaways
The U.S. labor market posted another month of strong job gains, doubling expectations. Global PMIs strengthened in April, and South Korean export data accelerated, signaling continued upstream demand across the technology supply chain.
Capital expenditure (capex) growth is no longer just a hyperscaler story. Nearly half of S&P 500 companies are now growing capex above 10% year-over-year, the highest share since 2022 and well above the long-run median.
Two structural themes are gaining earnings confirmation: physical AI is contributing to exponentially higher memory and compute requirements, while energy infrastructure investment is accelerating globally as grid hardening becomes a national security priority.
This Week in Charts
Looking Closer
Capex spending is broadening beyond just the hyperscalers, with 48% of S&P 500 companies growing capex by more than 10%. The median of this dataset going back 25 years is 40%. When capex broadens across sectors and company sizes, the beneficiaries typically extend across industries and borders.
Market and Data Recap
Labor markets and the global macro picture
The latest U.S. monthly payroll gain came in at double expectations. Non-farm payrolls have now exceeded 100,000 in three of the last four months. One softer reading in February largely reflected idiosyncratic factors, including a healthcare sector strike and workforce adjustments at some major couriers. However, we believe the underlying trend remains intact.
Within the report, the Information sector, which includes computer and data processing employment, was the one area of relative softness. That category is down roughly 5% year-over-year. Even so, computer and data processing payrolls remain approximately 30% above pre-pandemic levels, making it difficult to frame the data as anything other than a normalizing cycle within a structurally healthy labor market.
Weekly jobless claims reinforce the picture that there are no clear signs of broad deterioration at this moment. In our view, the current data argues against any near-term labor market breakdown.
The global picture adds to the constructive read. Japan reported its third consecutive month of real wage growth year-over-year, a figure the Bank of Japan (BoJ) monitors closely as it weighs further rate action. The June meeting remains a live possibility for an additional increase. Global Purchasing Managers’ Indices (PMIs) for April came in stronger than March, with broad participation from the U.S., parts of Asia, and select emerging markets (EM) like India and Brazil. Europe showed softness, but the aggregate trend held.
South Korean export data, which we track as an upstream indicator of technology and AI-related demand given its heavy semiconductor weighting, rose 44% year-over-year in the first ten days of May. That compares to a 37% pace in the same window in April.1 We believe this acceleration matters.
Investor sentiment has shifted sharply. Markets have moved quickly from fear to optimism as markets look through geopolitical uncertainty and focus on ongoing peace talks and earnings durability (Figure 2). That said, complacency signals do not change our constructive view on equities. The macro backdrop remains supportive, and capex continues to broaden. We believe near-term positioning risk does not alter the durability of those fundamentals.
Bottom line: Labor markets remain resilient at home and abroad. Global growth data is broadening rather than narrowing. The macro backdrop continues to support the constructive positioning we have maintained.
Capex: the surge is real and it is broadening
With 90% of the S&P now reported, 84% of companies have beaten consensus estimates. That is the strongest beat ratio since the first quarter of 2021, and much of the earnings strength ties directly back to the capex cycle we have been tracking.
The hyperscaler commitment to capital investment is the headline figure. 2026 hyperscaler capex estimates are now tracking at roughly double what they were one year ago. Estimates for 2027 are already running approximately 40% above where they stood just six months ago. The scale of that commitment suggests the demand running through the technology supply chain is durable.
As we noted in our Q1 Macro Investment View, “As AI technology advances, the contours of capex will evolve. This evolution will create new engineering requirements, new supply bottlenecks, and introduce niche companies as key drivers of innovation.” We are seeing this transition occur real-time as agentic AI changes the overarching workflow, compute demands, and memory needs.
But the capex story is no longer confined to a handful of large technology platforms. Currently, 48% of S&P 500 companies are growing capex at a rate greater than 10% year over-year (Figure 1). The long-run median for that figure going back to 2002 is approximately 40%. We are meaningfully above that baseline and rising for the fifth consecutive quarter.
Breadth matters as it alters the portfolio opportunity set. When capex broadens across sectors and company sizes, the beneficiaries typically extend across industries and borders. Memory producers, semiconductor manufacturers, industrial automation companies, and energy infrastructure providers all sit downstream of that spending.
Earnings this season have confirmed the supply constraint dynamic. Hyperscaler free cash flow has declined as capex represents an increasing share of operating cash generation. At the same time, high bandwidth memory provider earnings per share have risen sharply, and the gap between the two continues to widen.
Bottom line: Capex is rising and broadening. We expect the cycle to continue, with supply constraints in memory, compute, connectivity, and physical infrastructure positioning those sectors as the primary beneficiaries.
Two themes with structural tailwinds
AI is moving from the cloud into the physical world. Autonomous mobile robot penetration in warehouse automation is tracking a 20% compound annual growth rate, expected to reach an estimated 20% penetration by 2029.2 Demographic pressures, lower labor supply, and the global push for manufacturing onshoring are structural drivers. This buildout requires a different set of components than the model training cycle that preceded it.
Agentic AI3 and robotics demand more memory (a robot must be able to operate with near-zero latency) and enhanced computational infrastructure. Specifically, model training and inference remain in a graphics processing unit (GPU)-anchored architecture. Conversely, agentic AI is a fundamentally different workload profile: a long-running, branching sequence of decisions and tool calls strung together over minutes or hours. Above all, the real-time processing of complex tasks drives higher memory requirements, higher bandwidth needs, and lower-latency interconnects between central processing units (CPUs), GPUs, and memory. As a result, the demand for AI compute resources has been unprecedented and that revenue could have been higher if supply kept pace.
Grid hardening has moved to the center of the investment conversation alongside the physical AI buildout. Transmission, electrification, storage, and system-level resilience are the common denominators across geographies and energy mixes, regardless of how the ultimate fuel mix evolves. Countries are pursuing multi-pronged Energy architectures designed to seek reliability and domestic control. Energy-importing regions have structural incentives to invest, and the data reflects that.
Q1 earnings provided strong confirmations across both themes. Companies reported growth and margin expansion as they work to keep pace with rising electricity demand, grid stability needs, and national security interests. Others highlighted demand for grid independence and on-premise power solutions at data centers.
What connects physical AI and energy infrastructure is the same underlying dynamic driving capex across both: the world is building physical systems that require computational power and connectivity at a scale that did not exist even five years ago. As we said in our Q2 Macro Investment View on this theme: “This [physical AI]transition raises technical requirements for operators across transportation, logistics, manufacturing, and resource extraction, while tightening constraints in power delivery, networking capacity, memory, and precision components.”
Bottom line: The transition from software-driven to physical AI is tightening constraints across memory, compute, power delivery, and connectivity. We view both themes as secular in nature rather than cyclical and expect the earnings confirmation to continue as capital deployment accelerates.
1 Korea Customs Service Export Data as of May 1, 2026.
2 Citi Research, "Embodied Intelligence: The Rise of Physical AI." December 5, 2025.
3 Refers to autonomous AI systems designed to achieve specific goals with minimal human supervision.
See our weekly CIO Strategy Bulletin for more details