Related content

00:00 Brad

Welcome to The Flip Side podcast. I'm Brad Rogoff, Global Head of Research. And joining me today is Themos Fiotakis, our Head of FX and EM Macro Strategy Research. Thanks for having me here with you, Brad. All right. So what are we going to today? Well, we're looking at the market implications from physical AI. So our conversations going to rely heavily on insight from the 2026 equity gilt study.

00:23 Brad

And that's available to our clients on Barclays Live, in case you want to look at it after hearing this conversation. So let me cut to the chase though, with you here Themos when I speak to clients, policymakers and company leaders, they're quite worried about the implications of AI broadly in those concerns. They're often those short term and they're focused and really anticipating a world where digital AI replaces white collar jobs in the services sector, potentially leading to a very difficult job market for entire knowledge based minds professions.

00:54 Brad

What we're going to talk about more here today is what I would say is a step beyond that, and a world where machines increasingly replace humans in tasks considered to be today at least exclusively reserved for mankind. 

So, if physical AI infiltrates the sphere of craftsmanship, various different skill levels, potentially, there's another part of the job market that is suddenly at risk.

01:20 Themos

To some degree this has already started happening.

You have seen certain lines of professions, such as drivers in certain cities in the US and China, and factory workers already starting to get affected. But this could broaden out into many different lines of businesses. Think of cleaners or personal care, or craftsmen from all fields. So when automation moves from software into what you would call the real world, this stops being about coding efficiency and it starts being more about a broader range of jobs.

And I guess the question which you're also posing here is, are we underestimating the risk that this becomes economically destructive as opposed to wealth creating?

02:06 Brad

No, actually, that's usually my job to ask these big picture questions on this podcast. But hey, it's a debate, so anyone should be allowed to ask the questions. But yeah, that's the nice edge that we're talking about here.

There's a doom loop scenario. We'll talk about whether it's likely or not, but there's a doom loop scenario that if a large part of the population becomes unemployed, and if the gains mainly accrue amongst companies who will earn the income, for example, needed to buy stocks, we could even see recessionary dynamics on the back of stagnant demand. You know, counterintuitively, despite market expectations for profit growth, macroeconomic forces could instead drive corporate losses.

02:45 Themos

I mean, almost certainly there will be dislocations. And of course, the quicker these things happen. The shorter the time frame for economies and policymakers to transition in terms of job supply, reskilling of displaced workers and so on and so forth. Yet as we look at financial markets and on aggregate. Let's not lose sight of the fact that this is a textbook case of how wealth tends to get created over the years.

Well, how it has been created over the last few decades and over the last century, and how it gets dispersed then via financial markets, often market participants, they tend to view the creation of wealth and debt. These two go hand in hand, as two processes that are somewhat separate from production and income growth. But economic theory going as far back as Adam Smith identifies the genesis, the beginnings, the origins of Wealth of Nations in the decline of direct and indirect production costs, effectively the ability of humans through technology to produce more with less input from finite factors such as labor or capital.

 It is the real reason why societies can actually consume in the fourth to consume more, and why incomes proliferate, and then they transform into accumulated property. Financial markets. Then our job in financial markets is to ensure that this accumulated wealth, it becomes a supply of funding for new companies to create profit, to create jobs, and to create new opportunities based on those new technologies.

04:31 Brad

So that's all fine from theoretical point of view, I understand it. It's how we learn about economics, finance and everything. But where I push back is that in practice, these transitions, they can take decades. And perhaps more importantly, they can be messy. So even if the economy were better off on the aggregate, that does not mean the path is necessarily smooth or that markets immediately reward it.

And let's not forget the political volatility that deindustrialization created for a number of countries over the past few decades. And globalization is a good counterexample here. Despite producing substantial wealth for large groups of people around the world, it also created large groups of disaffected people who made their discontent known at the polls.

05:18 Themos

And, you know, this is while unemployment was low over this period of time.

05:21 Brad

Yeah, this time the issue could be even broader. We're not just a repetitive industrial processes. We're automating parts of both physical and cognitive work at the same time. So if labor demand adjusts more slowly than this new supply of machines capability arrives dislocations. I mean, they can now do potential benefits.

05:42 Themos

I think I think that's the right challenge.

Transition risk will need to be managed. But it is also important to highlight that some of those innovations, which are specific to physical AI, will also create broad benefits. Let's start by the proliferation of capabilities that reduce labor scarcity. It will also reduce issues linked to the operation of certain processes under conditions that are adverse for humans. And these processes will be immune to afflictions of the human race, like illnesses or forms of other types of labor.

As technologies improve, the cost of acquiring these capabilities will drop and the operating lives of that equipment will extend. 

06:28 Brad

There are certainly jobs that most humans wouldn't mind avoiding, so I get the point on that one. 

06:31 Themos

Exactly. And let's imagine, for instance, a repetitive labor process when you can potentially deploy robotic agents to augment human workers.

So production doesn't stop. It continues 24/7, or where the cost of production declines in a way that allows more space for wages in all these functions that robots cannot replace while creating new incomes, accumulating wealth for industries, and also benefiting smaller organizations and firms as well. 

I would add that physical AI could also reduce some opportunity costs that we all face in our private lives, in domains historically outside of formal markets, but very important for household prosperity.

Think about time constraints that we face due to housekeeping duties or commuting inefficiencies, or care for dependance like elderly or the infants, or even the efforts of physical risk mitigation. These could all be substantially eased. These are constraints that could be substantially eased, raising the productivity of the existing labor pools. 

07:45 Brad

As much as I'd love to debate this exact point back and forth with you, Themos, I mean, there's not an easy way to settle this argument ex-ante, but we're research analysts here, right?

So let's lean on history. And I know you leaned on history already. You took us all the way back to Adam Smith and the Wealth Nations. But let's offer some analogies and keep it maybe a bit more current. And you've done a lot of work on this, right? If we look at the most prominent kind of recent examples of technological advancements, I picked the internet boom.

 It's pretty easy to see that initially, robust equity returns did not prove sustainable. And I guess, in fact, we'd have to say a financial asset bubble of historic proportions was created. 

08:23 Themos

Yes. And I guess what you mentioned is also in the back of the minds of investors at the moment, as they see these returns for certain AI and physical AI companies start to accelerate.

But what we actually did is we looked back at three periods of rapid technological adjustment in the US the durable goods revolution, the 50s and the early 60s, the rise of the internet in the late 90s, and the cloud- data-mobile phones boom, the last decade, let's call it 10-15 years. Over this period of time, just to put some high level headline summary stats that come out from our work, the S&P advanced by 20% on an annualized basis over these periods.

 This was outpacing the S&P returns by more than 10% annually. The S&P returns in the interim periods, interestingly, over these periods of technological advancement, yields rose on average. And this is against a counterexample where yields dropped outside of those periods of advancement. And, you know, the dollar didn't actually produce much over this period of time. 

I should note here that in each of those periods, the income of non-farm proprietors, meaning the capital returns, outgrew wages by a significant amount.

I’m mentioning this because there's a pretty big critique, which you also mentioned that a big part of the AI and physical AI accrues mostly to capital gains over this period of time. Again, capital income outpaced wages, notably over three times in the 50s and one and a half time in the late 90s. Output growth over this period of time outpaced output growth over the interim periods by a significant degree.

10:09 Brad

All right. Well, I already gave you one criticism in terms of the bust after the internet bubble. But my other pushback, and honestly, I can't believe I'm the one who's going to say this to you, considering we both live, is that a lot of this is us focused. So surely there are some countries that are set to lose, even if it maybe isn't the US.

 In my mind, a number of industrial nations could underperform. I mean, you think some industrial nations have spent decades specializing in production processes and building their business model on competitive cost structures, some even explicitly relying on cheap labor inputs and loose regulatory standards. So surely they are set to get displaced? I would think, by a global reduction in production costs.

10:53 Themos

I mean, there's no way around the fact that this situation, at least as it looks now, reinforces U.S. exceptionalism. But I wouldn't generalize that result. Across all small industrial nations. There's a pretty robust conclusion in economics, going back to Ricardo, that even if a nation can produce at a less competitive cost structure than another, that nation is still better off focusing on the good and the production process they are mostly efficient in.

 So, an example of this is South Korea. For the longest time, South Korea has been facing a decline in the production cost and a decline in the price of all the goods. It was good at exporting: cars, air conditioners, all that. Still, over the same period of time, they've experienced rapid growth, rapid growth in living standards despite deteriorating terms of trade.

So despite that declining cost structure and declining prices in terms of trade, it was still a prosperous period for South Korea. 

12:04 Brad

South Korea very convenient example right now for me. I mean, considering the recent performance of the KOSPI! Outside of the obvious winners from AI infrastructure CapEx in the hardware space, though, I mean, I think the market has not broadly projected the impact of physical AI deployment across a broad set of sectors.

12:20 Themos

That's true. And this is a pattern historically. It's you know, markets are very shy in terms of projecting the rise and the growth of new total addressable markets (TAMs). In our note in our research paper, we discuss some interesting examples of how the market totally underestimated how big the markets for smartphones, cloud computing, EVs, TV streaming... the market thought it was.

There were niche products to begin with. They became enormous markets down the line.

12:52 Brad

All right. Let's bring this out of sort of the big picture and talk about sector. So when you move from macro logic to actual earnings, you know, becomes much more concrete, I think for people. And at Barclays we've done a good deal of work on AI and physical AI at a sector level in both the US and Europe.

 So the one push back that we haven't really gotten that deep into is that the benefits of physical AI, maybe much harder to realize in practice because the capital intensity is high integrations hard. And also it kind of alluded to this regulation matters. So when you checked your macro view against what the sector analysts are actually seeing, where do you think the clearest effects show up?

13:29 Themos

So our North America equity analysts have provided a clear framework to compare across sectors. And they see the clearest gains in fields like logistics, industrial distribution and manufacturing. Unsurprisingly, these are sectors where robotics are improving autonomy, flexibility and throughput. But high capital needs and complex integration are making the disruption reasonably slow. So for now, physical AI is lifting returns more than reshaping industry structure.

Environmental services also benefit robotics and computer vision, improve recycling throughput and material recovery.

Agribusiness and food processing are also benefiting from automation, better consistency, lower labor costs and higher yields. And on the flip side, if I may use that term, orders and mobility face longer term risks as autonomous vehicles could reduce car ownership and demand a very different business model.

In healthcare. Finally, in North America, AI is moving from diagnostic support towards assisting in the automating parts of fields reserved for specialized labor, such as surgery, etc. Though again, as you mentioned here or hinted, regulation is a constraint. 

14:50 Brad

Well, you're allowed to use the term flip side on this podcast, definitely. But I also promise we weren't going to be too US centric.

So what are you hearing in Europe? Anything different than that? 

15:00 Themos

There are some parallels in Europe, according to our analyst, but here physical AI is emerging as an efficiency generator but with a lag. 

Telecoms, business services, energy, metals and mining and other industrial-mainly sectors. Physical AI is enabling faster decisions, more deep automation, lower labor intensity and better recovery, uptime and efficiency in healthcare and engineering. However, it is more likely to augment than replace labor, helping ease workforce shortages.

15:34 Brad

So look, there's still going to be questions where years away from some of this stuff. But in the equity space, I feel like it's not too hard to break it down and see exactly where the impact will be in the magnitude still remaining unclear. Now, when we get to a bond world, I think things are a bit less straightforward.

So if I take your arguments at face value that we've been talking about here, stronger productivity should push up real rates, especially in the countries that adopt physical AI fastest be a stronger productivity should lift growth and probably lift the neutral rate over time. Also, AI investment its capital intensive likely to increase debt issuance, which will also likely expand the term premium and sovereign and corporate bonds.

Now we're already seeing the bond issuance, certainly especially here in the US, but we haven't yet seen that impact on corporate bond spreads, actually.

16:22 Themos 

To be sure, there is some mitigating force in the sense that lower production costs would reduce inflation and inflation expectations and compensation in bond markets. But I think overall, your analysis of impulses dominates. 

The volatility of rates has been historically a lot higher due to productivity shifts and broader growth shifts than inflation premia, inflation premia have been overall better behaved due to the achievements of central banks to be able to anchor inflation expectations.

16:58 Brad

All right. Let's take a fixed income. But now I'm going to do something risky here. And I'm going to try and think out the currency implications with our FX guru. So intuitively if adoptions uneven then first mover should benefit from faster productivity growth, higher neutral rates and capital inflows. 

So that sounds supportive for their currencies to me at least, much as the US dollar has tended to benefit during earlier technology waves.

China should also see clear gains here by the same logic with what's going on there. 

17:30 Themos

FX forecasting suits you, Brad. Thank you. Having said this, you know, once the technology spreads, those gains could fade and the laggards could benefit later on as they catch up. I would say that a little bit more near-term, as AI and physical AI builds out, some emerging market commodity currencies could benefit.

Given the nature of physical AI being pretty skewed towards metals and other kind of commodity inputs. And this is particularly important for mineral and copper exporting economies, where stronger export volumes, better terms of trade, leads to investment in the mining fields and lead to higher interest rates and productivity there. 

18:18 Brad

Well, even though you're nice and said I could be an FX forecaster, I knew I'd missed something on the side there.

But you know, the EM side, obviously interesting conclusions as well. So when I think about it, physical AI, you know, it may well be a wealth creation story in the long run, but the path there, it can still involve labor displacement, political pressure, uneven sector outcomes, and more complicated fixed income implications and plenty of volatility across a number of asset classes.

And to steal the term again as you did, maybe that is the real flip side story here. 

The more powerful the technology, the more important it becomes to distinguish the long term macro upside from the market path that gets you there. Themos, thank you again for joining me.

19:00 Themos

Thanks, Brad. This was great fun 

19:05 Brad

And thanks to all of you for listening.

For our investment bank clients who want to go deeper on this topic. As I mentioned earlier, you can read the full 71st edition of the Equity Gilt study on Barclays Live. Don't forget to subscribe so you never miss an episode. 

See you again! On The Flip Side.