Episode 81: AI’s Macro and Market Impact: A Framework for Investors
U.S. Economist Matt Bush and Market Strategist Maria Giraldo join the latest episode of Macro Markets to discuss insights from our new white paper, “AI’s Promise and History’s Lessons.”
Download the Transcript PDF
This transcript is computer-generated and may contain inaccuracies.
Jay Diamond: Hi everybody, and welcome to Macro Markets with Guggenheim Investments, where we invite leaders from our investment team to offer their analysis of the investment landscape and the economic outlook. I'm Jay Diamond, head of thought leadership for Guggenheim Investments, and I'll be hosting today. Now today we're discussing artificial intelligence. Our Macroeconomic Research and Market Strategy team just published a new white paper titled “AI’s Promise and History's Lessons,” which takes a broad look at the transformative impact of AI, drawing on historical examples of technological upheaval and examining the current environment of massive capital expenditures, valuation pressures, regulatory scrutiny, and uneven adoption rates.
The paper offers an approach to capital allocation across the AI value chain. The paper is in our show notes and on our website, so please check it out. To help us bring this paper to life, I'm joined today by two of its contributors, U.S. economist Matt Bush and market strategist Mario Giraldo. Welcome, Matt and Maria, and thanks for taking the time to chat with us today.
Matt Bush: Thanks for having us on.
Maria Giraldo: Happy to be here again, Jay.
Jay Diamond: Now, Matt, let's start with you. What prompted you and your team to write this paper?
Matt Bush: We see AI as a really transformative technology that is already reshaping the global economy. And with that transformation comes opportunities for investors but also risks. And so the goal of the paper is to try to cut through both the hype and the doom saying, and instead try to ground the conversation in economic data, historical precedent, and key themes across asset classes. These rapid advancements we're seeing in AI model capabilities can make it difficult to wrap your head around where things are heading and the investment pouring into these models, data centers, infrastructure is ramping up at a pretty staggering pace. And so, AI has already become the central issue, both for the macro economy and financial markets across all industries and asset classes.
And so, we want to give investors a framework to think about multiple dimensions of the AI story, the geopolitical race for AI dominance, the wave of investment we're in the middle of, a potential productivity inflection, the labor market transformation that comes with productivity, and then what this all means for investors across equity, fixed income and real asset markets.
Jay Diamond: So let's dive into the paper. Now the first section discusses the global strategic competition over AI. How is this playing out and why does it matter?
Matt Bush: Governments around the world are viewing AI as a central pillar for geopolitical strength right now. National security applications are one part of this, but it's also about long-term economic competitiveness. Major economies around the world are seeing slowing economic growth rates as population growth slows or even reverses. And so AI is seen as a critical way to offset stagnation in workforce growth by becoming a new source of this economic growth.
Right now, the geopolitical race is really dominated by the U.S. and China, who have differing advantages that we talk about. The U.S. leads in the most advanced AI hardware and frontier AI model capabilities, but Chinese AI models continue to nip at the heels of the best U.S. models and China's capacity to power its data centers runs well ahead of the US.
I think the key point is that governments are actively supporting and influencing the development and diffusion of AI with direct and indirect support for the AI sector. What that means for investors is that this is not just a private sector story. And sustained government involvement can be expected, which is likely to be an ongoing tailwind to the AI build out and to the advancement of AI model capabilities and applications.
Jay Diamond: Now, I should mention that this paper is chock full of charts and tables that help support all the arguments that you're making in the research. So it's worth checking that out on the website. Now it seems increasingly evident that AI is already having significant macroeconomic impacts. So how are you seeing this show up in the data? And is it all a positive story for economic growth?
Matt Bush: So I think last year we really saw AI move from just a narrative to actual hard numbers in the data. Looking at the U.S. GDP data since the end of 2022, we've now seen spending on computer equipment, software, data centers, and electric power rise by over $500 billion. And so it's safe to assume a large portion of that is related to AI.
So this is real, tangible economic activity that is impacting the data. Our estimates show that AI investment accounted for nearly a full percentage point of U.S. real GDP growth in 2025. Now, some economists downplay the impacts on growth by noting much of the AI hardware is imported, so it gets subtracted out of GDP. And while that's true, we think there are reasons that AI investment’s undercounted in the data, including the fact that it's hard to quality adjust for rapidly improving technology; it’s challenging to account for how much of this spending counts as intermediate versus final output; and it's difficult to account for intellectual property by U.S. companies and ships produced abroad. All of which means we think it's fair to say AI investment was a substantial driver of growth last year. And if we look ahead, we think we could see an even bigger contribution in 2026. What matters for real GDP growth is not the level of spending, but the rate of change.
Looking at the largest tech companies and how they've continuously revised up their estimates for CapEx this year to $660 billion, up $250 billion from last year—that points to another sizable contribution to growth. And it's not just through investment spending we're seeing the footprint of the AI theme in economic data. As we and others have noted, consumer spending looks to have become increasingly reliant on wealthy households in recent quarters as income growth in the labor market has cooled down.
And what's helped maintain consumer spending is strong gains in household wealth. Given how much AI-exposed companies have contributed to aggregate stock market gains in recent years, we think this wealth effect on consumption has generated around half a percentage point of consumer spending growth. So altogether, we see a substantial economic footprint from AI. But as you asked, the footprint is not all positive.
In the paper we discuss how AI is likely exacerbating the two-speed economy, creating relatively narrow drivers of economic growth. By pulling in resources and keeping interest rates higher than they otherwise would be, AI investment seems to be crowding out other areas of the economy that are more interest-rate sensitive. And we see that in investment categories less tied to AI like housing, office buildings, machinery actually contracting last year.
This is a normal part of economic evolution but is happening at a faster-than-normal pace and is important for investors to keep in mind. So overall, this phase of AI economic impact is being driven by large capital expenditures. But with investment now reaching into several hundred billion dollars, the rate of change in the spending will inevitably slow in coming years.
The more lasting economic impact and what's needed to generate sufficient return on these huge investments is a transition to AI, enabling economy-wide productivity gains.
Jay Diamond: So you mentioned productivity. Clearly, a major component of AI's promise lies in this potential to enhance worker productivity. What does that mean and how do you expect it to play out?
Matt Bush: We think this is really the key question over the long term, because enhancing productivity growth is how you get a durable economic impact from AI. It's a more sustainable growth driver than investment. Productivity allows companies to see returns on the AI spending and that justifies further deployment. But we do have to acknowledge uncertainty is really high right now. Nobody has the exact answer.
Estimates you see floated out there have massive ranges. People at the AI model companies still have economic growth rates of five or even 10% per year. On the other side, you have some prominent estimates showing virtually zero impact. And so we try to do in the paper is take two approaches to help get a sense of how to put some plausible numbers around how I can deliver productivity gains.
The first approach is really to think about how much more efficient AI tools can make specific tasks that people do, and then scale that up across the economy. And so looking at this task-based framework, we start with estimates of how many specific tasks across the economy can be automated using AI. There's no one answer here, but estimates suggest 20 to 40 percent of tasks are exposed.
The next question is how much more efficient AI will be in completing those tasks. And there's a lot of studies. Again, a big range, but generally 10 to 50 percent productivity gains on a specific task when using AI tools. And if we take the middle of both of those estimates, you get around a 10% boost to economy-wide productivity.
Importantly, though, that's not an instantaneous boost. It's going to take time. We're still in the early stages of adoption for AI, and there's roadblocks to fully utilizing AI tools. There's a lot of focus right now on coding and software development, but that's a small slice of the economy. The larger parts of the economy don't lend themselves as well to immediate integration.
And so it's going to take time to fully integrate all these AI tools. And so what that means is we estimated something like a ten-year window for the full productivity benefits to manifest. So on average, our estimate as a central range is around a one percentage point boost of productivity growth per year over the next ten years.
And that estimate happens to be pretty consistent with what we can learn from looking at the history of past general-purpose technologies, in particular railroads, electricity and computers, and the internet. Estimates of these technologies, effects on trend productivity growth tend to show a 1 to 1.5 percentage-point boost to trend productivity growth. And so if this is realized, that's a true regime shift over the long term.
When you consider estimates of trend growth without AI impacts around 2 percent or lower, faster productivity growth is how you get better economic and earnings growth without generating inflationary pressure. It's a very big deal if realized. But again, the potential is there, but we've yet to see the payoff.
Jay Diamond: So let's talk specifically about the labor market and its potential effect on workers being replaced by AI. How do you see this play out and won't it be a negative impact?
Matt Bush: It's a very real concern, one that's gotten even louder in recent weeks as the market focuses on near-term disruptions from AI. Our big picture view is ultimately AI is more of a complement than a substitute for human labor. And even for the jobs that are automated away, history has shown again and again that new technology transforms and expands jobs more than it eliminates them, despite widespread fears as new technologies emerge.
That's not to dismiss the risk entirely. The speed of AI adoption does introduce the potential for temporary but meaningful disruption. There's some evidence already emerging that we cite in the paper that job growth for younger workers in the most AI-exposed roles has seen a disproportionate decline since the launch of ChatGPT in late 2022. And that's occurred during a period of relatively solid economic conditions.
If an economic slowdown emerges, the historical pattern has been the heaviest job losses in so-called routine jobs, as businesses look to cut back on labor costs. And these routine jobs are some of the most vulnerable to replacement by AI, meaning this trend could be even more pronounced in the next downturn. So there is real potential for labor market disruptions in the near term, especially as AI adoption evolves rapidly and/or economic conditions put pressure on businesses to cut workforce costs.
But again, we think over the long term there's convincing reasons for optimism about the trajectory of employment in an AI-driven economy. It's difficult to predict in advance, but history suggests technological progress tends to create new job categories that were previously hard to even imagine. Sixty percent of jobs today did not exist in 1940, I think is a key stat that really captures that.
And one thing that's unique about AI compared to past technology shifts is that AI itself can be used as a tool for retraining workers to new kinds of jobs. And that should help with the transition to this new kind of labor market.
Jay Diamond: Now, Maria are turning to, the second half of our paper takes everything the Matt just discussed and synthesizes it into an investment approach. So how did you develop these themes, this part of the paper?
Maria Giraldo: Right. So we organized that second half around four themes specifically. And we'll walk through each of them. But at a high level, the purpose of those themes was to anchor our approach in long-term strategy rather than what we were really seeing last year, which was a lot of short-term momentum.
And even into this year, as Matt outlined, AI is moving incredibly fast. It feels like there's a constant stream of breakthroughs, new capital plans, new use cases, and sharp market reactions. Prices are moving quickly and then over the last 12 months, we've really seen a lot of opportunity across different fixed income sectors, across equities. And taking a step back, right, the promise is enormous.
Since the paper was written at the same time, disruption fears are taking hold. Investors are focused on who's going to be displaced. And we're still waiting for this evidence of economy-wide gains that distinguishes winners from losers. I guess the point here is that there's a strong combination that's happening that creates a fear of missing out dynamic alongside this really anxiety about disruption.
So our framework was constructed by working with portfolio managers, sector teams, exploring what's happening on the ground, and then looking at history and asking, what does history tell us about moments like this? So we'll get into that. I guess the final thing here is that we're really hoping investors see these themes as a kind of North Star through the investment cycle, something that's designed to have a longer shelf life and keep us focused on what is likely to be durable.
Jay Diamond: Now, the paper, as you say, contains four investment themes. Please explain the first theme, which discusses equity valuations and how to think about them.
Maria Giraldo: The first theme is about asymmetry that we were seeing in AI equity valuations specifically, and why diversification matters even if there's important differences between this cycle and the makings of the tech bubble in the 1990s, which a lot of people are drawing similarities to. Since the beginning of 2024, the hyperscalers plus Nvidia had delivered a substantial amount of returns in the equity market. Through the time of writing, the comparison was 125% cumulative returns from that group, compared to just 25% for the rest of the S&P 500. So that concentration reflected a market that was pricing in a high degree of certainty that these companies would continue to dominate and that they would turn heavy capital spending into profits almost in an uninterrupted manner.
Now, this is not the late 1990s. Unlike that period, we feel that today's AI leaders have strong profitability, substantial cash flow, and solid balance sheets. So that fundamental backdrop is very different. But history reminds us that the first movers in major technological transitions don't always retain that lasting value. You look back at railroads, electricity, personal computers, you know, in each case, the infrastructure builders often saw their returns erode, because what happens is the market starts to build excess capacity and then competition intensifies.
So what we wanted to highlight here, again, it's not that these companies are weak, it's that the valuations left little room for error. And through today, I do think that the markets have corrected a bit for that. Hyperscaler multiples have come down quite a bit, and the picture is much more balanced because leadership is rotating in the equity market.
But the overall framework for portfolios holds, and that is that diversification is central to this theme. So we still want exposure to AI, but within a broader portfolio framework that doesn't depend on a narrow set of winners.
Jay Diamond: Thank you. Now, the second theme in the paper relates to that. And it talks about how AI value won't stay confined to the companies that are building it. How do you expect that value to spread out over time?
Maria Giraldo: That builds on what Matt outlined around productivity. The key question for investors is where those gains ultimately show up. And the honest answer is that it's still unclear where value is going to migrate. And it's probably different across industries. But what we did know when we were writing this paper is that very little of the potential benefit from AI had been priced into the companies that are ultimately going to use it.
And I still think that's broadly the case, though we are starting to see where some areas where price multiples are starting to rise, like in consumer staples. But that lingering uncertainty is exactly why the opportunity exists. Now, thinking about the future, what we see ahead is that as adoption unfolds, those productivity gains are going to start showing up in businesses that integrate AI into their operations.
That's going to mean better efficiency, lower costs, enhanced customer outcomes. It could produce better profits, but investors need to find where the economic value lands. So let me put this in the context of what's happening in markets in February. Software companies have fallen over 30% since October, and that's because markets are worried about AI plugins that could essentially do the work that software companies were doing. And that's ultimately going to disrupt their revenues.
But where will value migrate? Does it go to the model developer through subscription fees? Does it go to the hyperscaler through increased compute demand? Or does it flow downstream to the software company’s client as cost savings or even further to the end consumer? Each scenario implies a different investment playbook.
So this theme is about looking ahead to that next phase and recognizing that much of that value migration is not yet priced into markets, but history suggests that eventually some value has to migrate to the end user. Otherwise the economics of adoption won't hold.
Jay Diamond: Theme three addresses current opportunities—what you're seeing in the market right now. So where are you seeing value today.
Maria Giraldo: Yeah, so this is now more about where we are confident expressing opportunity, especially as a fixed income manager. What we're looking for are parts of the AI ecosystem, where value doesn't depend on picking the winning company from a field of intense competition, but is instead rooted in stable assets, steady cash flows, and durable size of the opportunity. So let's start with data centers.
Last year, the market shifted from financing that buildout through cash flows to more debt issuance. And our own analysis found that expectations were for $3 to $7 trillion in investment that's going to be needed by 2030. So there's a lot more to come. Why are the ranges so wide? Because it's unclear how much compute is going to be needed several years from now to run future models.
And it's also unclear whether cost to build the infrastructure is going to go up or down. But what we do know is that the direction of travel is up. The challenge is that the structures of the debt that is financing this build out are not all equal. We are concerned that some investors are unknowingly getting exposure to obsolescence risk.
So that's the risk that some of these data centers are old, they run inefficiently, and they can't be retrofitted for AI use without significant investment. Those facilities could ultimately lose tenants because those tenants are going to have a lot more options with new build outs once their lease ends. So what we're favoring is secured investments tied to newer, more efficient data centers with high quality tenants and long-term lease contracts. That's the kind of differentiation that we're focused on.
The next opportunity is in energy. Running an AI data center is very energy intensive, but the country simply doesn't have the infrastructure to power it. Based on the projections of where all of this is going. And public budgets to finance more buildout for power aren't really there. They're already stretched as it is.
So private capital is going to be essential. And the other benefit in the energy space is that these are hard assets, which in this environment is attractive to investors. Energy infrastructure feels like a space that is under-owned and we're seeing a lot of opportunity to invest there.
Jay Diamond: Now Maria, the final theme in the paper looks at the opposing forces that AI could create for bond yields. So how should investors think about that big picture?
Maria Giraldo: That's exactly right. On the one hand, AI is driving a massive wave of investment. Companies are building data centers and expanding power capacity. And those that are not building the infrastructure are allocating resources to integrate AI into their workflows. All of that requires capital. And when demand for capital rises across the economy, it tends to push interest rates higher.
On the other hand, the technology itself has a potential to make the economy more efficient over time. If companies can produce more with the same amount of labor or resources, that expands economic capacity and helps keep costs in check. So in simple terms, productivity gains can reduce inflationary pressure, which would then support lower interest rates. So you have one force that's pushing rates up, which is a need to finance all this.
And then another one that's pulling in the opposite direction as efficiency improves. And then there's another layer that comes in and out as we move through the cycle, which is a potential for shocks along the way. We've seen some of that recently. Markets are reacting to concerns about how quickly AI-driven disruption could affect jobs or economic growth.
And with those fears, rising rates have moved a little bit lower because investors are more focused on downside risks. So for us, the net result is that yields stay rangebound near current levels. The ten-year Treasury yield has been in a tight range since 2023. It's now at about 4 percent. AI has a potential to keep that range intact.
And we view that range as attractive. For the past 15 years, the ten-year Treasury yield has averaged only 2.6 percent. That also has implications for the corporate bond market. Investment grade corporate bond yields have averaged just 3.7 percent and today they're at 4.8 percent. So this is why there's so much demand for U.S. fixed income these days. And we don't see those conditions changing.
So how investors should think about it is that they should be looking for opportunities in fixed income to express AI exposure as much as they're looking at equities, because the yields can provide some nice stability and steady income to portfolios.
Jay Diamond: This has been terrific. I want to thank you both for taking the time to walk us through this paper, which again is a real deep dive into a topic that's front and center for every investor. But it is a lot to digest. So I was wondering if you could kind of sum it all up in some takeaways for our listeners.
So Matt, why don’t you get started.
Matt Bush: If I were to summarize, I'd say the key questions we ask in the paper are will AI technology continue to advance? We think very likely, yes. Will AI generate durable economic benefits beyond this phase of investment spending? We think so, but the timeline there is measured more in years than in quarters and so patience will be required. The last question is will current AI investments generate durable returns?
And that's a harder question. And we think the answer depends more on where and how you're positioned. So overall, I think AI is going to be an enduring macro and market theme for years to come, and we hope this paper provides a useful framework for our listeners.
Maria Giraldo: I think Matt summarized it really well. What I would add is there's a lot of extreme narratives that are moving markets right now and, for what it's worth, from what we're seeing, we are optimistic that AI can bring a lot of opportunities. And hopefully our paper offers a balanced lens on how things will play out through this technological revolution.
Jay Diamond: Well, thank you both again, Matt and Maria, for your time and your insight. I hope you'll come back again soon and we can touch base on this topic again. And if any of you would like to read our paper, “AI's Promise and History Lessons,” again I direct you to our website or you can find it in the show notes.
And thank you again for joining us for our podcast today. If you like what you are hearing please rate as five stars. That's how people find us. And if you have any questions for Matt, Maria or any of our other guests, please send them to macromarkets@guggenheiminvestments.com and we will do our best to answer them on a future episode or offline.
I'm Jay Diamond and we look forward to gathering again for the next episode of Macro Markets with Guggenheim Investments. In the meantime, for more of our thought leadership, visit Guggenheiminvestments.com/perspectives. So long.
Important notices and disclosures.
Investing involves risk, including the possible loss of principal.
Stock markets can be volatile. Investments in securities of small and medium capitalization companies may involve greater risk of loss and more abrupt fluctuations in market price than investments in larger companies. The market value of fixed income securities will change in response to interest rate changes and market conditions among other things. Investments in fixed-income instruments are subject to the possibility that interest rates could rise, causing their value to decline. High yield securities present more liquidity and credit risk than investment grade bonds and may be subject to greater volatility.
This podcast is distributed or presented for informational or educational purposes only and should not be considered a recommendation of any particular security, strategy or investment product, or as investing advice of any kind. This material is not provided in a fiduciary capacity, may not be relied upon for or in connection with the making of investment decisions, and does not constitute a solicitation of an offer to buy or sell securities. The content contained herein is not intended to be and should not be construed as legal or tax advice and/or a legal opinion. Always consult a financial, tax and/or legal professional regarding your specific situation.
This podcast contains opinions of the author or speaker, but not necessarily those of Guggenheim Partners or its subsidiaries. The opinions contained herein are subject to change without notice. Forward-looking statements, estimates, and certain information contained herein are based upon proprietary and non-proprietary research and other sources. Information contained herein has been obtained from sources believed to be reliable, but are not assured as to accuracy. No part of this material may be reproduced or referred to in any form, without express written permission of Guggenheim Partners, LLC. There is neither representation nor warranty as to the current accuracy of, nor liability for, decisions based on such information. Past performance is not indicative of future results.
Guggenheim Investments represents the investment management businesses of Guggenheim Partners, LLC. Securities distributed by Guggenheim Funds Distributors, LLC.