The Hidden Story of the AI Economy - Who Gets Taxed When No One Does the Work?

Speaker A: M welcome to the Financial Executive's Edge, a production of the Financial Executives Journal. Here, finance meets bold leadership. Join us for sharp insights, uh, unfiltered conversations and practical strategies to elevate your thinking, drive change, grow your impact and empower your career. Uh, this isn't just insight. It's your edge. The Financial Executive's Edge.

Speaker B: Welcome to the Financial Executive's Edge. Today's podcast centers around the hidden story of the AI economy, posing the question, what gets taxed when no one does the work? I am Lynn Gargano, your host and moderator with the Financial Executives Edge and editor, uh, of the Financial Executives Journal. Artificial intelligence is often discussed in terms of productivity, automation and workforce transformation. But what happens to our tax systems when economic value shifts away from labor and moves to taxing compute power, data centers, semiconductors and energy infrastructure? In this episode we explore what happens when fewer people are doing the work, but more value is being generated than ever before. Who gets taxed and could AI fundamentally reshape tax policy at the federal, state and international levels? Join us for a forward looking conversation at, uh, the intersection of AI tax policy and fiscal policy where we explore how technology is not just changing how work gets done, but potentially where and how taxes are paid. We are joined today by Andrew Silliman, lead tax policy analyst at Bloomberg Intelligence, to explore this largely overlooked topic that may prove more consequential than than people realize. Andrew, thanks for joining us today.

Speaker C: Thanks for having me. It's an exciting topic.

Speaker B: So why don't we jump in and get started by thinking about the way taxes have traditionally worked. For decades, tax systems have been built around the idea that people work for a living, they spend money and they pay taxes. But now at AI, changing the nature of work itself, do you think we might be moving further away from a labor based tax system to a system that taxes and focuses on compute, power, infrastructure and even energy?

Speaker C: Yeah, I think that's the direction that uh, the US and eventually the global economy is moving in. I think the issue is that policymakers haven't really acknowledged this new reality. Well, I should say it's possibly beginning to sink in at the state and local level, but certainly not at the federal level. Uh, but not thinking about AI vis a vis tax is a big problem because uh, as you said, most governments arrive majority of their revenue from a source that I believe is about to diminish, uh, rapidly actually as the saying goes, um, slowly and then very very quickly. And this uh, is going to be difficult for governments because labor has been the main source of their tax remedy for most of the last century, because labor was where economic value was created. People worked, businesses hired workers, wages were paid, payroll taxes were collected, and then consumers spent their income, which generated sales tax. Uh, not only is this a very successful system, which is why it's lasted so long, um, but it's difficult to escape, uh, unlike corporate taxes or estate taxes, um, so countries like Ireland even made corporate, um, tax and other taxes, what amounts to essentially a loss leader, which was then made up for by additional wage and sales taxation produced by attracting new companies. But AI really throws all of that into chaos. I would say, uh, you could imagine, uh, two companies, Jeff, generating the same amount of output, one with 100,000 employees and the other using AI and 50,000 employees. The second company is going to generate similar or even greater profits while creating far fewer sources of taxable wages for cities, states and federal governments. So that creates a mismatch between where governments currently collect revenue and where economic value is increasingly being created. And, uh, you may look at my example and say, oh, come on, half the employees and AI is making up the difference? Come on, that's never going to happen. But, well, okay, we'll see. But, um, I mean, I think over time, businesses and people will adjust. And even with, uh, with AI, the number of workers is going to rise again someday. And then labor taxes will of course take back over, but I think that's going to take some time. And governments don't really have the time or the ability to sustain steep revenue declines like that. So Even if the second company in my example is using 90,000 employees rather than 100,000, that's still a steep wage tax decline for a government, uh, for a federal government, and more importantly for a state and city government. So, um, as AI adoption grows, I think economies become more dependent on semiconductors, on data centers, cloud infrastructure, um, electricity consumption. You might even say the compute power starts to replace labor as the critical economic input. Which doesn't necessarily mean that we're going to have a compute tax tomorrow. But it does suggest that some tax systems may have to shift away from wage taxation and towards, uh, the physical infrastructure supporting AI. And again, we're already, I think, seeing some hints of that. States are debating electricity rates for data centers. They're reconsidering property tax exemptions that are looking at infrastructure fees. Uh, these discussions aren't being framed as AI taxes per se yet, but functionally I think they may become, uh, just exactly that. So, um, it's a sort of encapsulate. I'M not saying that payroll taxes disappear, but I'm saying that over time, uh, they'll become a much smaller share of the overall tax base. And you know that that's going to happen over a very long period of time. It's likely to be much more gradual, um, than uh, than the way that our taxes have uh, have moved previously. But you know, I think over time, infrastructure, energy, capital intensive activities, they're going to be more important sources of, of revenue than uh, than labor.

Speaker B: So let's, let's take a deeper dive into this. As AI reduces the reliance on human labor and concentrates more on this AI infrastructure, what does that fundamentally look like in the way governments collect revenue? Right. And how does that impact the conversation around digital presence or physical infrastructure and data centers becoming the key focus?

Speaker C: Yeah, I think that's a great question. I think you're sort of getting to the crux of uh, the issue. Um, um, look at the US tax system for example. It's heavily dependent on labor, um, as I said, on wages, on wage taxation, on individual income taxes. In fact, if you combine individual income taxes and payroll taxes, they account for about 85% of federal tax revenue depending on the year. Corporate taxes are a much, much smaller share of tax revenue. Uh, if AI significantly reduces labor's share of income over time, governments face a real structural revenue challenge. Um, now I uh, want to make it clear that I don't think that technology, AI in this case will totally eliminate uh, work and create a post work utopia where we all collect a universal basic income. There are a lot of people that actually believe that, Sam Altman, Elon Musk, uh, among them. I'm not predicting that, but I do think that AI is going to fundamentally change the way that we work. And I think about it, uh, I think about the AI shift in the same way that I think about the Industrial Revolution or computers or the Internet. But take the Industrial Revolution because I think, I know that sounds kind of shocking, but I think that's really what may be going on here. Uh, it eliminated a massive number of agriculture jobs and it created a whole bunch of new industries, new jobs. Computers did the same thing, the Internet did the same thing. AI I think is going to do the same thing. Uh, but the problem here I think is timing, right? AI is coming in pretty fast. And um, if productivity grows faster than new forms of employment emerge, governments face a period where labor based tax revenues grow much more slowly than economic output. So um, let's take the example of farming, right? In 1800 90% of Americans were connected in some way to agriculture. By 1900 that number was 40%. By 2000 it was 2%. Um, there wasn't sort of a one to one shift from agriculture to factory work. And I think we sort of forget, uh, in the 19th century there were a lot of problems with that shift. There were a lot of displaced workers that never ended up going back to work. There was a lot of urban poverty due to unemployment, labor unrest, declining wages, and all of these people moving from the countryside into cities. And uh, eventually manufacturing picked up that excess of labor. But that adjustment took quite a long time. Generations computers of the Internet, I think are more our experience. Uh, we see them as having produced monumental changes. But the displacement was a heck of a lot more gentle. And I think AI is going to impact us less like computers than the Internet and, and more like the shift from an agricultural society to an industrial one. Um, so any major economic, societal shift that causes tax disruption and I think the pressure could be pretty significant. Um, so I published this report, um, a couple weeks ago, which I'm happy to circulate. And um, I looked at a couple of different automation scenarios. In a low, medium and high automation scenario, how much revenue the government could lose. And even in a low automation scenario in which AI leads to a displacement of 1 to 3% of labor, that still results in 50 to $100 billion in tax revenue decline annually. And in a high automation scenario where over 10 years we get 8 to 15%, uh, decline, that's a lot. Federal, uh, revenue could potentially fall by $800 billion annually before accounting for any offsetting gains, um, from corporate profits, things like that. But the point is the tax systems were designed around the assumption that economic growth and employ move together. And AI, I think challenges that assumption. Uh, so that creates pressure to find new revenue sources. Um, and where do governments look for revenue sources? They look at what's measurable, like electricity, data centers, semiconductors, infrastructure, land, power consumption. These are sort of tangible things that governments can look to. And tangible things are easier for governments to uh, um, to measure, to find and to tax. They're not going anywhere right now.

Speaker B: You mentioned that states may be starting to look at this already. And so how will states eventually tax this AI activity? Could it create a whole new definition for what creates nexus? In a state where you uh, look at where the computing is happening, where users are interaction, where data is being processed regardless of the physical presence, how would that work? And our systems even like prepared for calculating and measuring things like that?

Speaker C: Yeah, uh, I think um, as they say, states are the laboratories of, of um, of policy, of legislation. So I think if you want to know where the AI tax debate is going, you look at the states and I think they're going to feel the effects of AI on tax revenue a heck of a lot more fast than the federal government will because they have fewer uh, residents overall and less diversity of residents of companies. And so um, some states are going to feel it a heck of a lot more faster than other states. And they also rely heavily on sales tax which of course is driven by individuals. So I think those are the places you're going to see uh, experiment most aggressively with AI based tax policy because they have to. Um, but uh, some states are actually going to be very blessed. They already have a lot of data centers, they already have a lot of AI infrastructure, uh, physically located within their borders, like Virginia for example, or Texas, uh, Illinois, Ohio. So what would a more AI focused state tax look like? Uh, well, I think they already have some of the structure there. I mean property tax is obviously a big one. Infrastructure fees, electric, electricity related charges, um, and then maybe some new levies, data center specific levies, um, and I think maybe even cutting back on some of those tax uh, incentives at this point and modifications to sourcing rules for states that don't have that physical presence. And uh, you know that's, that's already taking place to a certain extent. States like Virginia, which already has the most data uh, centers in the country, they're debating many of these issues and they are talking about pulling back on a lot of those um, incentives. The interesting thing is that you have uh, some states, like I said, that are very blessed, Virginia, Texas, and then you have other states like California, New York that aren't as blessed. They don't have that AI infrastructure, the data centers. And the irony is that you have these, these physical AI locations in places that are not using as much of the AI as other states, like, like I said, California, New York. So where's the value being created? Is it where you have the data centers, where you have the physical infrastructure, or is it where the use of the A.I. uh, is centered? Um, there's no obvious answer to that question. So states are going to have to figure that out. Uh, traditionally uh, we tax based on people and physical business activity. But AI doesn't really fit neatly into those categories. And I think we could look to something like, like the Wayfair decision from, I think it was 20, 2018, 2016, something like that. If you remember that was the case about states being able to tax based on a digital marketplace's virtual presence in a state. Um, so Wayfair effectively expanded state's ability to tax economic activity without requiring physical presence. Uh, that was in the context of state tax. But that doesn't seem to really matter anymore because I think the state sort of importing those principles into, into corporate tax as well. But I think that that case becomes sort of a touchstone for AI. It becomes really sort of relevant for AI because I think you're going to see states increasingly argue that AI, uh, generated revenue should be taxable wherever meaningful economic activity occurs. That sort of virtual presence regardless of whether there's physical infrastructure in the state. And that could expand state tax exposure significantly. And then you have disputes between states about sourcing next value creation and you have discussions about double taxation and whether the sort of the water's edge technique that states have been using for decades is still the right one that they'll continue use. In fact, California is debating that right now, getting rid of that and sort of expanding their state um, tax net more broadly to bring in more revenue.

Speaker B: So, so I don't really think the question stops like within state borders. Um, I think we need to, because we're such a globally connected economy that we also need to look at this from an international perspective. And you started to touch upon that with, there might be a debate between states, but from an international perspective, could AI also challenge these long term assumptions that have been embedded in our current tax structure system with OECD Pillar 1 and 2 frameworks because they were designed before this whole new AI age? Are they already conceivably outdated with this new AI era?

Speaker C: Yeah, I think that's the case. That's a great question. I think, I think AI exposes some weaknesses in global attacks that were already emerging. Like pillar one for example. That was the global corporate wealth tax was designed uh, with, with digital services companies and the market concept, uh, market jurisdiction concept in mind, taxing where, where market activity is as opposed to where, where a uh, company is headquartered. The underlying idea was that users contribute to value creation and therefore deserve some taxing rights. But AI complicates that framework a lot because, because AI value comes from things that pillar one wasn't really designed to address like massive computing infrastructure and semiconductor capacity and training, ah, data, electricity consumption, cloud architecture and even physical investment, which wasn't something that pillar one uh, really had in mind at all. It was actually the opposite. So those inputs don't really sort of align with the concept of user markets. If value is generated by compute rather than user interaction, then pillar one's assumptions become harder to defend. So pillar one, I think, is increasingly incomplete. It's like a sheet that doesn't fit on the bed completely. It was designed before the AI boom. It addresses one version of digitalization, But AI represents a different phase entirely. So it's ironic because the international tax system has spent more than a decade debating where digital services create value. And now we're entering a world in which value is concentrated in infrastructure, physical infrastructure in a lot of cases, and compute capacity. And that's a totally different debate than the one that pillar one rose out of. And then you look at pillar two. Uh, in a sense, I think pillar two works better than pillar one because it's not nearly as dependent on user location as pillar one is. Focuses more on ensuring a minimum level of taxation. But I think even Pillar 2 still reflects assumptions from a much earlier economic model, a digital model. As AI becomes more so capital intensive, the questions emerge about where profits are generated relative to where infrastructure is located. Um, like semiconductor manufacturing, power generation at data centers. And those activities are going to increasingly become the drivers of economic value. Uh, and so the international tax rules sort of weren't necessarily designed with that shift in mind. I think minimum tax is in a sense, kind of an older concept that doesn't really work in this AI world.

Speaker B: So do you think that AI will ultimately make the tax system more efficient or more fragile? Because there just seems to be a lot of complexity and things that really have not been considered in the past and a lot of things to debate here.

Speaker C: Yeah, so I think that's another really interesting question. I think the answer is that it makes tax systems more efficient and more fragile. Uh, I think AI in a sense could dramatically improve tax administration. I think governments can use AI, uh, pretty successfully to detect fraud and to improve audits, uh, process returns more efficiently, um, identify compliance risks. And that's already happening. I mean, the IRS has been trumpeting its use of AI to improve accuracy and productivity. If you saw there was recently a hearing, uh, with the, uh, IRS CEO, um, in front of Congress, the new CEO, and uh, half of the things he was talking about were the IRS's use of AI. Um, obviously the IRS thinks it really has a lot of potential, so that's it for the efficiency side. But I think there's also the fragility side. If economic value becomes more concentrated among a smaller number of highly capital intensive firms, then governments become more dependent on a much narrower tax base. Which makes it a lot riskier for governments. Uh obviously if one of those companies goes under, there goes their tax revenue. A uh, broad labor based tax system, um, as I said earlier it's harder to evade. It's also a lot more diversified which makes it a lot safer. If one industry does poorly and another industry does well, they could sort of cancel each other out. That's not the case if you only have 6, 8, ah, 10 companies that most of your taxation is built on. And a compute based tax system at least with the companies we have today is a lot more concentrated. So AI could sort of simultaneously improve tax administration uh making uh, it much easier for governments to find revenue to tax it while making the underlying tax base a lot more volatile. But we sort of have that situation today with the Mag 7, right? That's Mag 7 comprises about 34% of the market cap of the entire S and P. And then when you add in space X, which I think is not part of the S and P yet but will very soon be, um, you know SpaceX is an AI company so you know there's even more concentration probably uh, raises it up maybe above 40%, 50%. So we already have a lot of that concentration already. But AI I think threatens to make that concentration a lot worse.

Speaker B: How would you balance the tax system then between labor based. Because there's always going to be the human in the loop and this whole new digital presence or AI era. Right, with the chips, with energy, with data centers. What do you think that balance looks like?

Speaker C: Yeah, that's a good question. Um, yeah, I mean I think what basically happens is that governments are going to have to shift um, their focus towards taxing compute and over time I think labor catches up and then they can shift back again. But uh, I think the percentage of taxation uh, is going to lean much more heavily on corporate and less on labor uh, over time so that they can sort of secure that new tax base and, and then maybe it switches back again. And we're just sort of at the awareness phase that this is something um, that governments need to do. Um, so I think initially what you're going to see is a focus on um, job loss, on uh, trying to figure out how to regulate AI and competition uh between companies. Um, tax policy I think is sort of uh, secondary consideration at this point, um, amongst governments, maybe even tertiary. Um, but I think as uh, the first sign is going to be if AI productivity really takes off and you see jobs start to tick down, that's I think when policymakers will start Seeing the first signs of a situation is they really sort of need, it's an emergency, they're sort of going to need to recalculate things.

Speaker B: Okay, so if we take everything that we've discussed today and looking ahead into your crystal ball, um, in this AI world, when do you think you started to talk about it? But when do you think we'll really see some real discussions and more debates around this awareness, government policy and eventually a longer term fundamental redesign of the US tax system to adapt to this new world? Do you think it's going to be slow and painful or do you think AI will just kind of emerge upon us full steam ahead and then the tax system will have to become reactive instead of proactive?

Speaker C: Yeah, I mean as I said, I think it's going to be uh, slow and then very, very fast. I think like I said, governments are going to see uh, AI productivity taking off and jobs sort of starting to tick down. And that's when they're going to say, oh my gosh, now we need to really rethink uh, the system that we have. Um, is it going to be able to recapture uh, the revenue we want or do we have to shift it completely? And governments have been relying on the system for more than a century. So that kind of massive change normally uh, takes decades. I think it's going to be a very tough process for governments to shift like that. And our system wasn't created overnight. Payroll taxes weren't created overnight. Um, the international tax system has evolved over generations. So I don't think it's a sudden revolution. I think it's uh, at least um, more of a gradual shift. Governments um, not sort of being able to face up to the fact that things are changing that drastically. And so you're going to have new layers added to existing systems. Infrastructure taxes are going to become more important. Energy taxes are going to become more important. Uh, states are going to place greater emphasis on data centers and compute intensive activity. You're going to see states talking more in terms of um, that digital presence type tax that we learned about from Wayfair M and international rules. I think you're going to slowly evolve towards infrastructure and resource based uh, concepts of value creation. Um, but I think sort of the key point is that that AI isn't just changing how work gets done. It ultimately changes um, what governments tax, where governments tax it and how public finance operates in the 21st century. So um, it's going to have to involve a complete rethink of the way that governments uh, tax um, and then, you know, as I think things shift back again once labor starts picking up, new industries form, uh, jobs, fill that void, then I think governments can go back to labor taxes. But the irony is that once they move in the direction of corporate infrastructure and energy taxes, it's going to be hard then to move back to labor taxes. So it's going to be a slow shift, um, towards corporate infrastructure taxes and then sort of a slow shift back again to, to labor taxes. But I think it's a huge story and I think, um, legislators, uh, haven't thought about it. I think most, um, individuals haven't thought about it. Um, and it's going to sneak up on us and then we're going to see it, and then, um, we're going to have government sort of freaking out trying to figure out what to do. But I think first we see it at the state level, then we see it at the federal level, and then I think we'll see it at the international level.

Speaker B: So that brings us to the end of the hidden story of the AI economy. Who gets taxed when no one does the work? One thing is becoming increasingly clear. AI isn't simply another technological innovation. It challenges some of the basic assumptions on which modern tax systems were built. Whether it's data centers, semiconductors, or energy infrastructure, the rules may need to change. While AI promises efficiency and productivity, it also exposes potential vulnerabilities in systems design for a labor based economy. And the answers may not emerge overnight. But the conversation has already begun here on the Financial Executives Edge with Andrew Silverman. And the decisions made over the next decade could define the relationship between technology, taxation and economic growth for generations to come. If you'd like to explore this topic further, I invite you to read How AI Is Rewriting where and How Taxes Get Paid, a Bloomberg Intelligence paper written by Andrew Silverman. Thanks for joining us today. Until next time, stay curious and keep asking the questions that shape what's next?

Speaker A: Thanks for listening to the Financial Executive's Edge. If today's episode sparked new ideas or helped sharpen your perspective, be sure to follow and review us on your favorite podcast platform. You can also visit financialexecutivesjournal.com for more insights, articles and upcoming episodes. Until next time, stay sharp, stay strategic, and maintain your edge. The Financial Executive's Edge.