Article last updated 12 November 2026.
Looking back on 2026 as the year winds down, it’s been characterised by one trade: AI enablers vs software.
In the first half of the year, that trade ran to the benefit of the companies making money out of the trillions of dollars of ‘capex’ to build AI data centres around the world – the AI enablers. Meanwhile, software got hit so hard that people coined the term ‘SaaS-pocalypse’ to prophesy the end of the software-as-a-service business models that had worked so well in the last 10 years. The popular view was that all of these companies would be displaced in the incoming reign of AI agents.
However, since June, this trade effectively went into reverse. The ‘AI disrupted’ have outperformed while the AI enablers’ momentum has stalled.
This chart demonstrates these two trading periods by showing drivers of the Bloomberg World Large & Mid Cap Index’s performance this year. You can see in the chart below that in the first half of the year the ‘AI enablers’, in dark blue, drove most of the index’s gain. They have been a headwind to performance since. In comparison, the ‘AI disrupted,’ in lilac, were a headwind in the first half of the year, yet they’ve since been a larger contributor to equity performance than all other sectors combined (right bar).
Therefore, I had planned to leave my US trip feeling reassured as to how this trade would evolve from here. Are the AI disrupted really toast? Can the AI enablers maintain today’s elevated growth rates?
What I learned on the American West Coast
Two of the companies I met during my week away, GlobalFoundries and Salesforce (we don’t own either), are at the epicentre of the trades I’ve described, as can be seen very clearly in their share price performance year-to-date. They have both moved inversely with each other and had very opposing first-half periods versus the last few months.
GlobalFoundries manufactures mid-range, workhorse chips for all sorts of devices, as well as specialised equipment used in AI data centres. It’s seeing the benefit of data centre expansion right now – its revenue from data centre business expanded 81% in its recent quarter compared with a year earlier – taking it to roughly 17% of total quarterly revenue. And there could be plenty more to come if you believe capex forecasts for data centre builders like Microsoft and Amazon.
It’s not just the volume of new data centres that provides growth – existing data centres need upgrading for AI. We learnt in the meeting that AI is creating so much data that current chip connections are becoming a bottleneck, so the industry needs much faster ways to move information, such as the new materials that GlobalFoundries is developing. GlobalFoundries noted demand for these new technologies is stronger than ever – supply remains a constraint, but customers are even willing to help fund GlobalFoundries’ capacity expansion in order to secure supply.
Over in software, Salesforce provides the world’s most-used CRM tool – a platform used by salespeople all over the world to record client data, track interactions and manage leads. This puts it at the centre of AI disruption fears: if AI agents can do a salesperson’s tasks for them, this could lead to fewer salespeople and fewer Salesforce licenses. Unlike for GlobalFoundries, this is more a long-term disruptive risk than what is actually happening today – in its most recent quarter, Salesforce’s revenues continue to grow by double digits.
In our meeting with Salesforce, they argued that they can exist in a world of AI agents. Their argument was that their recent partnership with AI lab Anthropic enables customers to run agents on top of their Salesforce data to complete tasks. Not only does this imply Anthropic is fighting with them rather than against them, but Salesforce found that pilot customers of this new agentic service show stronger engagement and increased interest in Salesforce products than before. While human users often fail to use all available features, agents automatically identify and execute the best workflow, resulting in customers discovering functionality they previously didn’t know existed. The remaining question is, of course, around monetisation: if the agent is doing the task, does Salesforce’s product still offer the same value to the customer that it has historically? We are not so sure.
In addition to the companies’ own comments, I found it interesting that the investor group I was with seemed more positively inclined towards the AI-disrupted company meetings, while I remained more sceptical. Additionally, there was more scepticism for many of the AI enablers than I had expected.
This of course reflected the share price momentum of the third quarter; however, AI enabler valuations are much more reasonable today, and the group demonstrated strong earnings momentum in the recent quarter. GlobalFoundries, for example, now sits on a PE multiple of less than 20x, well below its June highs of over 40x. Software, on the other hand, has recovered substantially from its retrenchment. Yet it’s still relatively depressed: Salesforce shares are trading below where they started the year.
What it means for our portfolios
So can both these companies co-exist in an agentic world? They certainly think so. As always, the answer probably lies somewhere in the middle.
We believe a barbell approach is important: backing our conviction in the AI rollout with a larger position in the AI enablers while retaining some of the better-positioned AI-disrupted businesses to protect against swings in sentiment. Those with the deepest moats and which are ‘stickiest’ to customers because they are heavily embedded in systems and data, such as SAP and Microsoft.
While we don’t own Global Foundries, we do own peers TSMC and Intel, which dominate their markets to such degrees that it gives them significant economies of scale.
We hold more of the AI enablers as money continues to flow into the infrastructure behind AI, and we believe we are still in the early innings of a multi-year AI data centre build-out. That said, to protect against the wildest swings in sentiment we concentrate our holdings in the companies where valuation and forecasts are reasonable, visibility is clearer and moats are significant.
For example, while GlobalFoundries is a bet on the commercialisation of its many next-generation technologies, we prefer to hold the companies that sell tools into GlobalFoundries and can share in its success – for example, ASML and Applied Materials’ tools will be used in its capacity expansion – as well as its more diversified peers – like TSMC – which has better near-term visibility from today’s data centre spending.
While we feel unsure of the levels of long-term disruption software will face, their valuation multiples remain relatively cheap versus history and recent momentum and company comments are certainly positive. In addition, the level of disruption from AI agents is unlikely to become clear for several years. In the meantime, there could be more savage swings in sentiment like we’ve seen this year.
