Top
Two factors make us question the quality of the market’s earnings growth: (i) the earnings boom has not been matched by cash generation, and (ii) this exceptional earnings growth comes from a starting point at which corporate margins are already close to historical highs.
The third quarter proved a challenging one for global markets. Re-escalating U.S.-Iran tensions and tightening supply drove extreme volatility in energy markets, with Brent crude oil futures surging 40% and the resulting energy shock keeping inflation fears firmly in focus. Monetary conditions tightened further as a number of central banks raised policy rates, while mounting fiscal pressures and sovereign debt concerns compounded the pressure on longer-duration bonds, pushing the 10-year U.S. Treasury yield up 50 bps over the month, closing September at 5.29% – its highest level in more than two decades.
It was also a choppy quarter for equities. July saw a reversal among some of the strongest year-to-date beneficiaries of the artificial intelligence (AI) infrastructure build-out, with semiconductors falling sharply and other areas of physical AI infrastructure coming under pressure. These losses were largely recovered as the quarter progressed, with the release of several next-generation autonomous agents and frontier models contributing to a notable improvement in AI sentiment during September. At the same time, market participation narrowed considerably over the quarter, with less than one-third of MSCI World constituents outperforming the index in September, down from around two-thirds in July.[1] The relative performance of our quality portfolios broadly followed this round trip, improving as participation broadened before weakening again as it narrowed.
The MSCI World Index’s modest 2% gain in the third quarter takes its year-to-date return to a respectable 12%. Underpinning this performance has been an extremely buoyant earnings backdrop, with the index’s 12-month forward earnings per share (EPS) rising 23% year-to-date, even as its multiple has fallen 10% to 18x next-12-month (NTM) price-to-earnings (P/E) – below its COVID-era peak of 21x and only slightly above its 10-year historic average of 17x. The earnings momentum has been even stronger for the S&P 500 Index, where forward earnings are up 31% year-to-date.
While the earnings strength has been a nearly market-wide phenomenon (with only four of the index’s 11 sectors – utilities, health care, consumer staples and real estate – delivering EPS growth of less than 10%), the bulk of earnings upgrades for the index come from two distinct pockets: energy and AI.
In energy, persistently high oil prices have driven EPS growth of 60% year-to-date, while the AI deployment boom has propelled earnings growth in semiconductors (+111% EPS), hardware (+38%) and the hyperscalers (+32%). Aggregating energy and these three “AI-enablers” areas together explains over 70% of the increase in index earnings growth. Taking into account that AI spend has also benefited pockets of industrials supporting the physical data centre build-out, the true earnings growth concentration is likely to be even greater.
Beyond the obvious heavy reliance on AI-related earnings growth and volatile energy prices, we would point to two factors which make us question the quality of the market’s earnings growth. First, the earnings boom has not been matched by cash generation: while the MSCI World 12-month forward EPS has risen by 23% year-to-date, its free cash flow growth has been 12 percentage points lower. In other words, the improvement in underlying corporate economics is less pronounced than headline earnings growth would imply, as the semiconductor companies and other infrastructure providers take 100% of the AI buildout revenue and profit today, while those spending the money capitalise most of their costs.
Second, this exceptional earnings growth comes from a starting point at which corporate margins are already close to historical highs. While today’s extraordinary AI earnings are accruing to the companies supplying the build-out, we have yet to see a commensurate benefit to margins among corporate customers buying and deploying the technology. There is evidence of productivity gains in areas such as advertising and coding, but these have yet to translate into a material earnings benefit across the wider corporate sector, which implies that the margin improvement has largely been driven by the AI-enablers, rather than adopters.
In our view, two areas in particular will test how well this earnings backdrop holds up from here.
Can corporates sustain earnings in a higher-inflation world?
Higher inflation may support nominal revenues and earnings, but periods of high inflation have not historically translated into higher returns on equity for corporates. As Warren Buffett observed in his 1977 Fortune article, How Inflation Swindles the Equity Investor, as prices rise, so too does the capital required to sustain a business, with inventories and productive assets ultimately needing to be replenished at higher costs.
Keeping quality investors on their toes, inflation can also temporarily gift even relatively low quality businesses the appearance of pricing power, as they may initially be able to pass higher costs on to customers. As inflation persists, however, and higher interest rates compound rising input and labour costs, genuine pricing power should become easier to distinguish, with those businesses better placed to protect their economics.
Can the AI earnings boom endure?
The sustainability of earnings in parts of the AI complex will largely depend on whether the AI build-out has created newfound and durable pricing power across the supply chain. Putting to one side the rare bottlenecks, the usual rule of capital cycles is that exceptionally attractive economics incentivise additional investment and bring new capacity online. Some of today’s supply constraints may persist for some time, not least because the relatively short replacement cycle for leading-edge AI chips could create a recurring source of demand. Current earnings expectations appear to assume that demand will remain strong enough to absorb much of that additional supply and sustain today’s unusually favourable economics. The question is whether AI can indeed reverse the multi-decade trend of memory, power and other components becoming cheaper per relevant unit of output, or whether we are experiencing a more traditional – albeit exceptionally strong – capital cycle in which increased demand ultimately brings forth increased supply. The distinction will be enormously consequential in areas such as memory, where operating margins have moved from a historical through-cycle range of roughly 0%-30% towards 70%-80%.
The opportunity
Today’s market is underpinned by earnings growth that is exceptional by any historical standards. More than 20% earnings accretion in just the first nine months of the year has rarely been seen outside of one-off recoveries from earlier sharp cyclical slowdowns, like COVID or the Global Financial Crisis, never mind from a very high margin base. The sustainability of this growth will have a consequential impact on the direction of broader market returns.
Against this backdrop, we find the opportunity in quality compounders particularly compelling. This group, in our view, offers resilient growth and proven pricing power, which should translate in reliable earnings compounding, with much lower sensitivity to the AI build-out, commodity prices and inflation cycles than the broader market.
Importantly, investors are currently paying relatively little for that optionality, as the premium attached to the more dependable earnings growth we seek has compressed. Our global portfolios now trade at broadly the same free cash flow yield as the market – valuation levels rarely seen in its history – despite materially higher returns on operating capital employed and gross margins, and what we believe to be a more resilient through-cycle earnings profile. Crucially, that earnings profile is more reliant on dependable top-line growth than on further margin expansion from already elevated levels.
Nor does the quality opportunity depend on either of the risks discussed above materialising. Our global portfolios are deliberately positioned across a range of outcomes: we participate in the AI build-out through select enablers and bottlenecks where we believe competitive advantages are durable; own high quality businesses with the potential to benefit from AI that continue to trade at depressed valuations; and retain defensive compounders whose investment cases are largely independent of the AI debate. If enthusiasm around AI infrastructure persists, we should participate through the first group; if market leadership broadens, the second offers meaningful rerating potential; and if earnings growth becomes more difficult to find, the resilience of the third – and indeed of the portfolios more broadly given their combination of proven pricing power, resilient revenues, and credible through-cycle compounding – should become increasingly valuable.
Featured Insights
