Where across the value chain will profits ultimately accrue?
After an exceptional AI-driven earnings cycle in 2026, the investment debate is shifting to the sustainability of returns, the durability of demand and the economics of AI. Are the returns sufficient given the investment required, and where across the value chain will profits ultimately accrue?
Global equities returned around 13% year to date to the end of September. While next-12-month consensus earnings estimates rose approximately 30%, the market multiple compressed from around 19 times to 16 times. Much of that earnings acceleration has been driven by the AI cycle.
The speed of change in market leadership illustrates just how non-linear the AI cycle has been. Micron, Samsung Electronics and SK Hynix have all moved into the top ten contributors to global index returns in 2026, yet some of these businesses were generating negative gross margins as recently as 2023. As a result, they would not have appeared in many conventional investment screens before their recent moves, leaving a number of equity strategies offside as the cycle turned. In our view, in a rapidly evolving market, the ability to identify inflection points, rather than simply extrapolate history, is increasingly valuable. This is an environment in which we believe fundamental research becomes more important, as detailed analyst insights can help identify those inflection points and challenge the assumptions embedded in market expectations.
An important component of the earnings-driven growth we’ve seen is that it has been driven in part by a transfer of cash from hyperscaler balance sheets and free cash flow into the earnings of AI hardware companies, particularly memory. The lower market multiple we now see suggests investors recognise that this balance sheet transfer is unlikely to be sustainable over the long term. The question from here is how durable those earnings prove to be and how sustainable the sources of funding behind them are.
We remain constructive on global equities into 2027, but the risks around earnings expectations are higher than they were a year ago. The market continues to expect above-cycle earnings growth, but interest rates are higher, inflation remains above target for many central banks, geopolitical uncertainty persists and US consumer sentiment is at multi-decade lows. At the same time, the largest companies are starting from a less flexible balance-sheet and free-cash-flow position following significant investment and greater use of both equity and debt markets.
We do not see those funding sources as exhausted, and importantly, the evidence on AI adoption remains positive. The next phase of the AI cycle, however, increasingly needs to be supported by monetisation. The market is watching recurring revenues at private AI companies and hyperscaler cloud growth, while companies more broadly are increasingly identifying new applications across areas including pharmaceutical and biotech research and cybersecurity. So far, the signals remain encouraging, but ultimately the scale of investment needs to be matched by growing adoption from enterprises and consumers.
As we move into 2027 and look for confirmation of the durability of earnings, the market is going to be looking for evidence about 2028. Many companies, particularly across the AI supply chain, have already sold their 2027 capacity, which means the potential for earnings beats on the scale seen in 2026 is lower. The more interesting question is therefore likely to be one ofduration: can companies demonstrate that this cycle extends into 2028 and beyond? Further evidence of that durability could support a rerating in valuations even if the rate of earnings-per-share growth begins to moderate for some of the current beneficiaries.
One of the other major investment questions is where the profits ultimately lie as the AI value chain evolves. We are looking across the whole market, including hardware, large language models, cloud providers, the industrial companies building data centres and providing their power and cooling, software, cybersecurity and the consultants helping enterprises build AI capabilities.
Technology delivery itself is also evolving because we keep crashing into the laws of physics. This is forcing a rethink of how data centres are constructed, with the move towards optical technology as one example. Solving those constraints is itself creating new investment opportunities.
At the same time, we do not think we are close to understanding all of AI’s use cases, and we have barely begun optimising the AI stack. Greater efficiency could ultimately change how much infrastructure is required.
Against this backdrop, managing investments through this dynamic environment comes back to balancing return and risk, while positioning for a range of potential outcomes. For instance, if the returns do not justify the spending, we would expect to see hyperscalers pull back on capital expenditure. In that case, they could experience a significant increase in free cash flow as they begin leveraging the investments already made. Highly leveraged neoclouds could face a much more difficult outcome. We are also asking where the credit exposure sits if the AI cycle were to unravel, and where capital would rotate in that scenario. This is one reason we continue to value attractive opportunities that are not dependent on AI.
By Rosie Malcolm, portfolio manager