Monologue
Not every week brings regime-shifting news. Sure, hyperscale names are pulling US stocks down, but at current valuations, it's hard to make the case for an infinite rally, and a bit of caution (not panic) at these heights is healthy.
Meanwhile, in a slow week for new macro data, everything we did get confirmed our view of near-term economic dynamics. Inflation remains high (especially the upward drift in "core" or underlying measures), but consumer spending growth is more than keeping pace. This justifies the Fed's hawkishness on rates, and we continue to expect a two-hike (50 basis point) tightening cycle to keep inflation in check without troubling low-income consumers or causing excessive problems in the subdued housing market.
In this week's Memo, we turn our attention to what seems to be the big debate of the decade: will AI eat your employment prospects? By framing the new technology as a type of capital deepening for service workers, we can compare the current investment spree with prior capital deepening periods centered more on manufacturing. We find reasons for optimism.
With that, enjoy this week's note.
Dylan Smith
Founder and Chief Economist
Marginal Movers
Rising 👆
- Korean AI adoption: Does AI Adoption Improve Productivity? Effects Over the First Three Years — "AI has currently entered the 'efficiency' stage but has not yet fully transitioned to the 'productivity' stage"
- Inflation damage: The Search Costs of Inflation in the Labor Market — "When wages are contracted in nominal terms, inflation reduces real wages, leading workers to intensify on-the-job search to obtain wage-adjusting outside offers. Both the search effort and the resulting job mobility are costly"
Falling 👇
- Interest rate stability (?): Will the Shadow Fed Chair Please Stand Up? — In the interest of balance, a counterpoint to our views on Warsh: "A muddier signal raises the odds that the Fed will surprise markets and that adds to borrowing costs. And it leaves the public with a looser claim on an institution that owes it both accountability and transparency."
- World Cup Finance Content: PE World Cup: Who wins when capital calls the shots? — World Cup-related financial research distraction papers used to be great (see if you can spot a certain Chief Economist's contributions). Now they're this.
Macro Monitor
PCE (Inflation edition)
Core Personal Consumer Expenditures (PCE) inflation – the Federal Reserve's preferred benchmark to gauge its performance against its 2% target – rose to 3.4% YoY in May, confirming the signal we already had from the Consumer Price Index (CPI) that even as energy price growth starts to slow, core prices are still accelerating.
Meanwhile, consumer spending is rising faster than inflation (+0.7% MoM in May, vs. +0.4% for headline PCE inflation).
No thematic change in these data. Inflation is high, and consumption is rising regardless.
Good Canadian Behavior (Inflation edition)
Headline Canadian Consumer Price Index (CPI) inflation was 3.2% YoY in May, and stands above the 1%-3% target range for the first time since December 2023. Looking beyond the rise in energy prices driving this move, underlying inflation remains completely benign. The Bank of Canada's suite of core inflation measures ranged between 2.0% and 2.7% YoY.
Looking at the change in these indices on a 3-month annualized horizon provides the cleanest signal of near-term momentum shifts, and, conveniently, covers the period since the start of the Iran conflict. The BoC's underlying inflation metrics are well-contained at the midpoint of the target range. Meanwhile, the traditional measure of "core" inflation – the headline CPI excluding energy and food – is at the lower end of the target range (1.0%), revealing the fundamentally disinflationary setup in Canada's consumer demand, investment, and housing sectors at present.
PMI High
The S&P Global Flash PMIs provided an early glimpse at activity trends in June. In the US, the index rose to a five-month high of 52.2, comfortably in "expansion" territory. The manufacturing component is at a blistering 49-month high of 55.7 (very robust growth). This is partially driven by precautionary inventory building, so the unrelated rise in services sector output to a 4-month high of 51.3 is welcome. The only blot on the copybook was a second sequential fall in reported employment, which was attributed to cost management.
The PMIs also showed America's relatively strong position compared with Europe. Services PMIs were in contractionary territory and falling in the UK and the euro area, with modest manufacturing growth unable to fully offset the decline.
See the appendix for arcMacro proprietary Factors and the Key Macroeconomic Indicators tracking chart.
Market Monitor
Public markets
The S&P 500 has had a volatile month, and it's clear overall that there is a lacuna in the bull market as investors vacillate between optimism over the potential earnings implications of the AI build-out and concern about the sheer scale of financing that chip makers and hyperscalers require. Bad vibes won out this week to the tune of a -2% return in the S&P 500, its second down week in four. The Nasdaq Composite (which includes AI company SpaceX) fell by 4.6%.
Within the American stock market, the pessimism was limited to technology names. A powerful rally in Treasuries supported rate-sensitive industries such as Real Estate and Utilities, while strong consumption data drove gains in Consumer Staples and Health Care.
That Treasury rally sliced off some of the extreme reaction to last week's Fed meeting, but still leaves 2- and 5-year Treasury yields more than 10 basis points above the three-month average. The dollar followed suit, with a 0.5% climb in the DXY index.
Global equity markets were mixed, with East Asia and emerging markets broadly hit hard (Hang Seng: -5.2%; TOPIX: -2.0%), Europe flat, and the UK rallying by 1.5%.
The benchmark WTI crude spot closed the week below $70 per barrel, despite a rocky start to mass-scale Hormuz transits and a shaky ceasefire. Gold was down yet again and is nearly below the $4,000 per ounce threshold.
See the appendix for the market monitor table
Memo
Services Sector Capital Deepening
Bottom line: A process of rapid capital deepening is taking place in white-collar industries and occupations. This points to higher productivity and wage growth in the services sector. But the net employment effect depends on whether AI and labor are complements or substitutes. Looking at historical deepening cycles in US manufacturing, we argue for the former.
What it means for investors: A period of high productivity and strong wage growth without aggregate job destruction implies high returns to equity in exposed industries. It also imparts upward pressure on longer-term interest rates — for the "right" reason (expected strong future returns rather than inflation compensation).
"Capital deepening" is when businesses invest to increase the amount of capital per worker. In theory, this should raise labor productivity and result in higher wages.
Usually, capital deepening is thought about in terms of investing in physical productivity enhancement for laborers in goods production or logistics – adding more and better machines to a production process to raise throughput and quality. It's therefore central to the study of industrialization and growth/development economics.
The idea of capital deepening gets far less play in the white-collar world. That needs to change. If AI is viewed as a machine that improves the throughput of a human mind, then we're entering a period of rapid capital deepening for white-collar occupations for the first time in history.
At the aggregate level, the ratio of new business investment per member of the labor force carries important information on the drivers of post-COVID-19 global economic dynamics. The United States is the only major industrialized economy that has been actively adding capital to the workforce. That growth has been more robust, and returns have been broadly higher over the same period compared with peers, follows from this observation.
This isn't your Granddaddy's capital deepening, however. It's narrowly concentrated in AI-related investment in IT systems and software. Since 2024, after roughly a year of rapid improvement in frontier Large Language Models (LLMs), businesses have invested heavily in these two areas. IT and software spending per services-sector worker (a rough gauge of white-collar capital deepening) has popped from $45k per year in 2023 to $60k now, roughly doubling the pace of growth compared to the prior two decades (measured in 2012 dollars to adjust for inflation).
Employment could rise or fall depending on whether the new capital investments are on aggregate complementary to human labor, or substitutes. Viewed through classical growth theory, this points to a period of higher services sector productivity and wages to come. But the theory is largely silent on the employment level that results from these effects.
The manufacturing industry provides precedent for both effects. As the chart below shows, there have been two major capital deepening episodes in the industrial sector since reliable records began in 1960.
The first, spanning the post-war period until the economic crisis of the early 1970s, was driven by the process of reorienting wartime technology breakthroughs and production capacity toward consumers. This generated higher productivity, higher wage growth and higher manufacturing employment. Productivity gains improved returns by more than the marginal cost of labor, leading to a broad expansion alongside improved efficiency.
The second period, roughly spanning the 1990s, also saw improved productivity and wages, but this came at the cost of a decline in manufacturing employment. "Capital deepening" was in part driven by a shrinking denominator, as automation and early robotics replaced workers, with businesses seeking to compete in a more globalized labor market.
This raises a natural question: which type of episode will AI be? If history is anything to go by, early rounds of capital deepening are employment-generating because the marginal productivity curve is steepest at this point, and the scope for improvement across the economy is wide. There are good reasons to believe that implementing AI in business processes will be employment-generating in the aggregate.
Of course, this won't hold for every occupation. But on balance, we're optimistic that AI will create more jobs than it destroys.
Appendix
Proprietary Factor and Regime Model and Key Indicators
Disclosures
AI Declaration
All written content, analysis, and opinions are original and ascribed to the author. AI tools were used for proofreading and summarization purposes only. AI tools may also have been used in the development (codebase) of the analytical models reported in this document.
Disclaimer
This publication is for informational and educational purposes only and does not constitute financial or investment advice. Nothing in this report should be construed as a recommendation to buy, sell, or hold any security or financial instrument. Always consult a qualified financial advisor before making investment decisions.