Key takeaways

  • AI risk is eroding once-solid sources of competitive advantage for companies, according to a study by Adrian Helfert of investment firm Westwood.
  • Companies most exposed to AI have underperformed the most AI-resilient companies by wide margin, according to Helfert.
  • Helfert says stocks including ExxonMobil and AbbVie look attractive in this environment.

Stock investors have long used economic moats as a predictor of excess returns. But the proliferation of artificial intelligence lowers barriers to entry and threatens the foundations of those moats.

Morningstar incorporates moats, which represent companies’ durable competitive advantages, into its equity research. Yet the advent of AI has raised questions about the durability of companies’ business models.

We checked in with Adrian Helfert, chief investment officer of multi-asset strategies at investment firm Westwood. Helfert recently completed a quantitative study, to be published on SSRN, that looks at AI disruption risk and how to evaluate what it means for companies. We chatted with him about the study, what to own and what to shun, and whether the market will rise in 2026.

Leslie Norton: You’ve created a way to assess AI disruption risk—an issue roiling the markets.

Adrian Helfert: I created a practical way to measure how AI is changing the value of companies. Traditionally, investors estimate how long a company can earn above-average profits, meaning its “competitive advantage period” or moat. I kept that framework but added a new layer: how exposed each company is to AI disruption.

I scored 460 of the S&P 500 companies on both traditional moat strength and AI risk, then used optimization techniques to see what the stock market is actually rewarding and penalizing. The surprising result: Four of the five classic moat pillars—switching costs, network effects, intangible assets, efficient scale—have almost no predictive power in today’s AI environment. The only moat that still clearly matters is physical cost advantage—real assets like supply chains, factories, mineral reserves, and regulated infrastructure.

On the risk side, the market is not primarily worried about “AI replacing workers.” Instead, it is pricing two things: AI-native competitors attacking incumbents, and AI eroding proprietary data advantages. Together, those drive most of the return differences. Empirically, companies most exposed to AI have underperformed the most AI-resilient companies by nearly 26 percentage points in the first seven weeks of 2026.

We now use this quantitative framework directly in valuation support. For AI-vulnerable companies, we shorten the expected life of their competitive advantage, raise their cost of capital, and apply a justified P/E haircut. The question is no longer just “how strong is the moat?” but “can AI build a bridge across it?”

AI highlights speed at which profits shrink

Norton: Let’s have a quick recap of what’s been happening in the markets over the past couple of weeks.

Helfert: What’s changed is the market’s growing concern that the key AI risk isn’t near-term earnings destruction, it’s acceleration of the fade rate—the speed at which a company’s excess profits shrink over time as competitors erode its competitive advantage in the moat model.

When I run the analysis, it’s pretty clear the market is pricing that disruption risk during specific repricing events. A repricing event is something like the launch of DeepSeek, or statements from Claude about reducing the need for lawyers because AI can do legal analysis, and you see the impacted companies’ securities move immediately. An example is Salesforce CRM. Suddenly, many of the pieces of functionality I can do myself, or two guys in a shop have rebuilt it in a way that is just as functional.

In AI-related decline, resiliency in utilities, energy, materials, defense

Norton: What is your framework telling you now?

Helfert: The data is shifting the weight of what explains market drawdowns. Historically, we leaned more on things like switching costs, network effects, intangibles, and efficient scale. But in the recent headline-driven selloffs, those factors had a much lower statistical weight than we’ve used in the past. Instead, the loss of cost advantage becomes the dominant explanatory factor in negative moves during headlines like “Claude releases a model that can do legal analysis.”

This isn’t the market pricing an apocalypse. It’s pricing steady competitive compression. The fade rate is becoming more important because cheaper, more capable AI lowers barriers to entry. The market isn’t repricing companies simply because AI replaces workers. It’s repricing because AI creates new competitors and commoditizes what used to be proprietary, including data and workflows.

When we look at sectors, those that are higher risk turned out to be financials, consumer discretionary, IT Services. We’re seeing impacts there now. We’re also seeing resiliency in utilities, energy, materials, defense. This is no surprise because it ties into the major input.

AI isn’t just software. It’s power demand, cooling systems, grid strain, water intensity, and regional policy risk layered on top. We look at AI through all of those investable paradigms. When you look at it that way, AI starts to look like the new industrial layer. Then the question becomes: what are the toll roads to that layer? In many cases, it’s utilities, energy and increasingly, water.

Norton: How should people adjust their moat-investing approach?

Helfert: Right now, cost advantage is a relatively small slice of the traditional moat framework, but we think it needs to be overweighted.

Then there’s a technology-risk classification that historically wasn’t emphasized enough. Some companies function as a system of action, meaning they close the loop between data, intelligence, and execution, like automated inventory ordering. Others are more of a system of record. Salesforce is a good example. They sit on an enormous amount of customer data, and the security and governance around that data becomes part of the moat.

We also think in terms of a judgment moat and a metadata moat. Using these classifications to model where moat erosion is most likely is becoming increasingly important in valuation. For example, automated execution workflows may be more exposed than sensitive, governed customer data. This classification system helps describe which business models are more resilient, and which are more vulnerable, as AI capabilities expand.

Norton: What have you been doing in your portfolio during this time?

Helfert: We’ve reduced exposure to certain software names where functionality is high but security and trust requirements are lower. Those are the areas most vulnerable to AI-enabled commoditization and faster fade rates.

Moat analysis is one input into valuation, but we’re ultimately focused on company-specific upside and risk/reward. We’ve added Evercore EVR as an advisor levered to the M&A and restructuring cycle, and EQT EQT as a natural gas producer we like on valuation and for the potential catalyst of a phase shift in liquefied natural gas exports. We’ve also reduced exposure to Disney DIS in part because its brand-driven, intangible moat is increasingly vulnerable to AI-enabled content commoditization and distribution disruption. That can compress the duration of its competitive advantage.

Norton: What stocks would be resilient in this new universe?

Helfert: When I talk about resilience in this new universe, ExxonMobil XOM and the large-scale energy stocks have a significant moat in large-scale physical assets. Because of scale, they have cost advantages. They will use AI to their own benefit as well. Other top companies were AbbVie ABBV, which has a large-scale set of physical assets, regulatory barriers, and other healthcare names. Utilities consistently screen well in moat work. Ameren AEE is one example of a utility that shows a strong moat profile.

More broadly, water and energy are direct inputs to AI growth. That makes parts of those ecosystems more resilient, both fundamentally and from a capital expenditure cycle standpoint.

We’re also going to see a significant resurgence in alternative energy. The nuclear resurgence is something that investors should be looking at because of how it can be deployed, researched and used. Nuclear is one of those areas within non-traditional energy companies with expanding earnings rates.

Stay away from IT services

Norton: What should investors stay away from? After all, a lot of stocks have fallen sharply already.

Helfert: Information technology services. As you know, Anthropic announced that Claude could potentially start to do COBOL (a coding language designed to process commercial data) programming and consultancy. That’s a large part of IBM’s business. So IT services is an at-risk area.

Norton: At what point do we get bullish on software or financials?

Helfert: Software and financials aren’t dead. They’re still building real utility and creating value. Some will integrate AI in a way that strengthens their proposition. Salesforce is a good example. It’s been hit hard because some of what it provided around customer data is now being built everywhere. But if you want your functionality and your customer data in one trusted place, with governance, security, and accountability, there’s still a role for a scaled platform. Trust becomes a key factor. I still believe brand and platform value can persist.

The software companies that can credibly say, “You can do it all with us” versus stitching together lots of small AI tools, have a better chance to survive and thrive. Most enterprises want one primary partner for a core workflow, not a dozen. So companies like Salesforce can find a bottom, especially as they adapt their architecture to deliver AI-driven capabilities across the product set. That said, I do think AI can erode pricing power in the interim, and that’s what the market is wrestling with.

Norton: We’ve also had spillover of the AI angst into the private credit market. Is there a broader subprime AI concern?

Helfert: I’m always looking for where the hidden leverage is, meaning the shadow system. Private credit has grown meaningfully. But there are a couple important offsets. These are generally secured loans, the portfolios are typically diversified, and there are structural buffers like risk retention/holdbacks in parts of the collateralized loan obligations (CLO) ecosystem that help incentivize underwriting discipline. And you don’t have the same concentration of exposure sitting on major bank balance sheets the way people sometimes assume. Do I have a concern that it’s systemic, as we saw during the global financial crisis? I don’t think it’s systemic.

A manager like Blue Owl [whose stock recently fell amid increasing anxiety about the health of the private credit market] can be impacted by mark-to-market concerns and pockets of risk in specific loans, but their holdings are still diversified. One other key point is that when high yield starts to break, there’s often less flexibility. In private credit, you can negotiate covenant changes. In a downturn, lenders can adjust terms to help a borrower manage through the dip, which can reduce forced defaults and disorderly repricing.

Helfert’s bullish outlook on the market

Norton: What’s your outlook for the rest of the year?

Helfert: There are times you can point to a catalyst, and today the catalyst is productivity driven by a new technological innovation that can broaden out the market.

Rates are likely headed lower. Whether that happens at the next meeting matters less than people think. What matters is the direction of travel toward a lower federal-funds rate. If you think about “neutral,” we’re probably in the neighborhood of two to three more cuts away, depending on how labor and inflation evolve under the dual mandate.

Then there’s policy stimulus in the One Big Beautiful Bill Act, which we expect to translate into roughly $3,000 to $4,000 in average tax refunds back to consumers. Historically, a meaningful portion gets spent. Consumption is still about 70% of the economy. And then there’s capex. When you build a data center, you’re hiring a lot of people, at least for now. And $700 billion by four companies at the top of the market cap stack alone is an extraordinary number for capital spending.

We have the potential to keep growing as long as earnings remain solid. For now, they are. We’re seeing roughly 14% earnings growth for the fourth quarter in aggregate, and closer to about 9% for the non–“Mag 7” cohort. If multiples don’t move at all, that’s still a 10% higher market. I want to be involved.

Get Morningstar insights in your inbox