Is AI taking over? Portfolio and productivity insights for asset allocators

Six thoughts on the AI revolution
Brian Barbetta, Global Industry Analyst and Portfolio Manager

1. The market still may not have fully grasped the implications of AI.
I don’t think the technology or what it means for investors is fully appreciated. For example, the scope of the required AI infrastructure build-out isn’t widely understood. Nor is the extent to which the adoption of AI is already driving revenue growth and efficiencies across different industries and companies. There are important questions to be answered about how the technology develops and the time horizon, but that misses the bigger picture.

This isn’t surprising. For years, I think the markets have underestimated technology's impact and growth due to an emphasis on mean reversion and market rules developed prior to the information revolution. I understand why, but today I think it’s critical to recognize that AI and its consumer and enterprise uses are developing rapidly and that this will continue.

As a simple example, consider social media apps. They’ve long been able to steer users toward content of interest, but in just the last few months, this capability has taken a huge leap forward thanks to AI. This speaks to something I've been talking about for over a decade: The ways in which companies can use technology to get a little bit better at a lot of things and then allow that to compound. Even in an area as mature as social media, the new tools that are coming offer the potential for explosive growth.

Figure 1

In many respects, this is different from some previous technological revolutions. In the industrial sector, for instance, when one company found a way to do something better — say, invent the assembly line — other companies could hire some of those people, invest their own capital, and catch up with the competition. It’s not that simple when you’re talking about these natural monopolies, and as a result, I think these businesses may still have a long runway in front of them.

To put it another way, my focus is less on how big these companies are in the index and more on why the stocks have worked and may continue to do so. From that perspective, what I continue to see is significantly better-than-market-average returns on invested capital, which comes from better growth, better margins, and the ability to redeploy capital efficiently, whether through buybacks or building new businesses. With all of that said, it's hard for me to come up with a reason these companies can’t be bigger than they are today.

3. As in prior technological revolutions, the near-term winners won’t necessarily be the long-term winners.
While scale and distribution are critical drivers of success for some of the big winners today, there are no guarantees. I try to focus on how competitive intensity and consumer behavior are changing, and on where there may be weaknesses emerging in business models. Some companies will make mistakes and miss key technology trends. The question is whether they will catch up or fall behind and start ceding share. And the goal is to be positioned accordingly.

I also think it’s helpful to spend time understanding how smaller private companies are building and advancing technology, and then to combine those insights with public company insights to create a research mosaic that can inform investment decisions.

4. AI models need to keep improving — and if they do, the massive spending on data centers may not even be enough.
As I noted, some of the companies applying this technology are seeing very high returns on their investments and their capex plans reflect an expectation that this will continue. For me, the question that is top of mind is this: If you consider what AI models are capable of today and look at the capacity builds planned for the future, is it possible that current capacity expansion plans are insufficient to meet future demand?

Right now, we’re seeing the output generated by the models growing rapidly. They are delivering much better results and are far more compute-intensive. Those advances must continue, and productization of those models must accelerate, to have a need for the capacity that will be coming online. Our research suggests they will, but time will tell.

If models continue to get better, that will improve their ability to have a meaningful impact on consumers and enterprises, including driving efficiencies, new products, and revenue growth, and that will drive the return on the capex. And then we may be talking about even larger spending plans than we are today.

5. For those who are overweight technology, there are risks to watch for.
There are technology companies that I believe are very well positioned for the future of AI that trade at a price I don't think leaves upside relative to where I expect them to be in three to five years. There’s also the possibility that either the technology doesn't deliver as promised or there's some sort of “off-ramp” in the current technology that drives things toward a different architecture and makes a lot of existing hardware less valuable.

But overall, I continue to see many green lights ahead. That includes stocks that are cheaper than they were at the beginning of the year based on their forward earnings multiples, despite having appreciated quite a bit. And again, I think the market continues to consistently underestimate the technology’s upside potential.

6. AI has a key role to play in the investment process.
Within my team, we’re primarily using the technology to help make people more efficient at what they do. For example, I might ask someone to look into what every company in health care is saying about AI. In the past, that would have required reading a great deal of research and other material, as well as talking to analysts and attending company meetings. Now, we have internal tools that can look across all the research, earnings call transcripts, and public commentary, and answer very specific queries in great detail — often within hours.

As another example, when tariffs created an opportunity to add to some holdings earlier this year, I went to one of the large language models. I uploaded a list of stocks and asked it to look at all the available filings for data on how much revenue the companies generate in certain countries and how much they import from certain other countries, and then to calculate the impact on company revenue and earnings. It was done within hours rather than days, and the team was then able to spend its time checking the work to ensure we could confidently apply it to our investment decisions.

Portfolio implications and organizational applications
Adam Berger, Multi-Asset Strategist

Let me offer a few thoughts on how asset owners might incorporate some of these AI insights in their own portfolios and approach AI implementation within their organizations.

Weighing the positives and the negatives in the tech sector
I see several reasons to consider leaning in on technology investments:

• As Brian noted, being on the wrong side of new technology (i.e., underweight) has historically tended to lead to detrimental investment outcomes.
• The market distortions created by technology’s strong performance, like the sector’s weight in the S&P 500, may feel unsustainable, but I don’t think this is a case of castles built in air. The fundamentals have been strong over an extended period (see Figure 1).
• While some companies may be expensive relative to the norms of the past several decades, they're not necessarily expensive relative to the market in absolute terms or relative to their own growth.

I also see reasons to be cautious:

• With regard to the capital investments companies are making, it may be too much too soon — and at least a reason to be alert to the possibility some of it is being misallocated.
• The race to create the “best models”— especially as it relates to large language models and the cost of building and training them — could be a zero-sum game in which there may be only a couple winners (raising the risk of misallocated capital).
• Brian’s point about the scale advantage of tech companies notwithstanding, it’s possible we might ultimately see more dispersion of AI winners. In other words, the technology may be so transformative for the overall economy, that the gains benefit other sectors as much as the technology sector.

Navigating the current investment environment
Given the opportunities and the risks, I think asset owners should consider several steps:

Make a clear call on how to weight technology exposure — Ultimately, whether asset owners are overweight or underweight technology and AI-driven stocks, they need to understand the size of their positioning, their confidence in the technology’s ability to keep driving growth, and their potential exposure if they are wrong.

In a poll of more than 60 institutional investors we conducted recently, almost 40% said they were underweight technology stocks. My worry is that some have substantial underweights that are out of line with their confidence in what is likely a “binary” outcome that is hard to predict. If that one sector bet dominates the overall active risk in their portfolios, the risk from that sector should be brought down (by adding exposure to underweights or reducing exposure to overweights), and asset owners should seek out more diversified sources of active risk.

Plan ahead for what could change your mind — Whatever weight is chosen, have a checklist of data points and a timeline that can help with decisions about whether to course-correct.

Think about two key sources of alpha — In the pursuit of alpha, look for AI winners flying under the market’s radar. And given all the attention AI is getting, keep an eye out for undervalued opportunities in other areas that are being ignored or are out of favor.

Be thoughtful/intentional about correlated bets — Watch for situations where an investment’s upside may be levered to technology. For example, infrastructure has generally not been highly correlated to the broad stock market, but that could change to a degree if infrastructure investments are tied to technology data centers. There can still be compelling opportunities in infrastructure, but it will be important to understand the balance between upside and risk. 

Use the technology yourself — Having some hands-on experience with the transformative potential of AI can give investors a concrete feel for the current and future possibilities of this technology.

AI applications at Wellington Management
In Figure 2, I offer an overview of some of the ways our own investment teams are using AI. For example, in advance of company meetings, AI can be used to efficiently gather information about the management team and their recent public comments. Or, as Brian alluded to, we can use AI to collect and quantify unstructured data, including gleaning insights from across many management reports and earnings calls, and then highlighting potentially powerful connections and comparisons in the data.

Figure 2

Figure 3

Figure 4

Risks to consider
Any discussion of AI applications also needs to take AI risks into account. Here, I would highlight three broad categories:

User input risks — When inputting information into AI tools, it has to be appropriately ring-fenced to ensure it stays secure.

Model output risks — As noted earlier, AI output needs to be carefully reviewed before it is used in any process, given the potential for hallucinations and biases in the data a model was trained on.

Legal and regulatory risks — These include risks related to the ownership of the data used to train the models, as well as emerging regulations on the permitted uses of AI.

Final thoughts on next steps
I’ll wrap up with a quick summary of implementation ideas:

For your portfolio: Ensure your top-down technology positioning (direction and size) aligns with your level of confidence in the sector and your patience. Think long term but be open to changes as the facts change. And along the way, look for alpha opportunities beyond a big top-down bet on technology, including underappreciated winners.

For your organization: Have a vision of how AI can play a meaningful role in your organization, a roadmap to get you there, and a plan for oversight and risk management.

For yourself: Invest the time needed to understand AI, follow new developments, and begin to look for easy ways to begin putting it to work in your day-to-day life.