The Deep View: Conversations

Ian Krietzberg:
Welcome to this rare double episode of the Deepview Conversations. Now, normally you only see me once a week. This week you're seeing me twice. Thank you very much for watching. And the reason being is that, you know, normally we record our episodes several weeks in advance. We have a long list of really cool episodes coming up the pike, and we publish them once every week. The topic we're diving into today, beyond being timely, is changing at a radically fast and kind of unpredictable rate. And so we recorded this episode this morning with the goal of getting ahead of any additional changes. And so the question that we're trying to answer today has been on my mind for at least past couple of weeks, and I'm sure it has been on a lot of other people's minds as well, is simple. What do Trump's tariffs mean for the field and the industry of artificial intelligence? And further, as we get into, what does that mean for everything else? Because AI, though it might not share that name, is a very broad term, There's a lot of technologies that fit under that umbrella, and a lot of those technologies are incorporated in just about everything that we use on a daily basis. So before we get into it, here is how things stand right now at 12.09 p.m. Eastern Time. Just a couple of weeks ago, President Trump imposed sweeping global tariffs, what he called reciprocal tariffs, around the world. These included additional tariffs against products from China, where we do a lot of manufacturing, as well as a 32% tariff against products from Taiwan. I mentioned the Taiwan one specifically because Taiwan is home to, among other things, the Taiwan Semiconductor Company, which fabricates NVIDIA's chips as well as the chips and GPUs for a number of other companies. And so as it relates to AI, the Taiwan focus is important. Shortly after that, and in between bouts of truly wild market stock market volatility, Trump announced a 90-day pause on those reciprocal tariffs. This reduced the base tariff rate around the world to 10%, which is still significant. The only country that things were not walked back with was China. So as of right now, we have 145% tariff rate against products from China. China has reciprocated. A couple of days ago, the administration noted that smartphones, computers, and computer chips are all exempt from these reciprocal tariffs when they go back into place. And the tech companies are very, very happy about that. But the administration has also noted tariffs specific to semiconductors, which would include smartphones, computers, and computer chips, are on their way. They could come as soon as this week or within the next one to two months. The scope of those and the exemptions for companies tied to those, the quantity, the size of those tariffs, completely unknown. That is how things stand. as of right now and that might well be very different depending on when you are watching to or listening to this hopefully the the discourse we have here will will kind of hold up regardless of what happens with that all said i wanted to bring on an expert in global supply chains. And so my guest is one that, if you're a longtime listener of the show, you've heard her before. She is a pleasure to have on. She's a pleasure to speak with Dr. Nada Sanders, who, among other things, is a distinguished professor of supply chain management at Northeastern University. She's also the co-author of the book, The Hue Machine, which is a very interesting read. With that all said, the question that we are answering today is, What impact will these tariffs have on AI? This is the Deep View Conversations. Nada, thank you so much for being here again. This is our, you're our first repeat guest. So very excited to talk with you today.

Nada Sanders:
Thank you, Ian. Thank you so much for having me on.

Ian Krietzberg:
Yeah, so obviously the big kind of thing on everyone's mind right now is the tariffs and the trade war. And to me, the first thing I thought was, how does that impact AI writ large? And the first person I thought to ask that of was you. And so this stuff is moving very, very fast. As you just kind of mentioned before we hit record, it's a moving target and things are changing very quickly. And so I kind of want to explore some different scenarios and kind of try and unpack the different ways that these actions could impact different elements of what's going on. But I want to start with this, you know, you study supply chains. And it might not be obvious to people how much of a supply chain AI is. And I wonder if you could kind of walk us through the extent of the supply chain that makes, you know, a chatbot work and how vulnerable it is to tariff actions.

Nada Sanders:
Thank you for asking. And there's so much to say. So just to kind of underscore the fact that this is a moving target. So this is April 15th, I believe, 11.15 a.m. you know, by the time we get off the air, it could possibly change. And right now, where we are sitting is we have US has imposed 145% of tariffs on goods from China, China has retaliated 125%, you know, and I really think once we get past 100, Right? It's like, you know, 200, you know, it gets crazy. And what I think what is also particularly significant are export restrictions that China has imposed. And the reason is that with the tariffs, at least with the right amount of money, you can access the goods. With an export restriction, you cannot. There's a limitation, there's a licensing that has to take place, which means there's no amount of money that can give you access to those goods. So kind of coming back to the question that you had asked, I'm a business professor. My area is supply chain management, and within that, I focus on forecasting and building supply chain resilience, and then the interplay between human decision making and technology, which is AI. And we have evolved, our economy, the global economy, has evolved into very much a global ecosystem over the years. We, because of free trade, because of growth, because of the need and desire to access goods, we have these super intertwined supply chains. That means that we access, when it comes to AI, when you're talking about everything from hardware to software to displays to servers to all the things that we need, We are looking at raw materials that come from many, many, many places across the globe. Some of those raw materials, in particular, such as the rare earth minerals, do come from places in the world where they are only available. And China has the primary source of these that are used in all kinds of component parts and equipment that then is made into sub-assemblies that then may be shipped to the U.S. or that might go to Vietnam and then to U.S. We are looking at not a linear trajectory of movement of goods even between countries. But we're looking at a very tangled network of these components that include everything from silicone to, again, rare earth minerals to all kinds of different components, aluminum, that get formed in everything from consoles to packaging and then go back and forth. It's not a linear trajectory, they might go back and forth across countries multiple times and until we actually see the final product where it ends up. With the tariffs imposed on it, and now especially the export restrictions, And also the fact that this is very much now a volatile moving target where it changes hourly or daily at minimum. What's happening is that everyone is scrambling. We, as a country, and I think every country, and this is certainly manufacturers and those that we, you and I and everybody listening, purchases goods from, and that includes businesses, they are scrambling. They are trying to figure out what to do. And I think, as I mentioned to you about the complexity of the supply chain, and not, we call it a chain, it's really a very complex network. It is like this very large locomotive on tracks that you cannot switch and get it to move one direction or another. There is a huge amount of structure and infrastructure that takes place from the factories from the labor, so we talked a little bit about the materials and the export restrictions and the tariffs, but there's also factories and transportation and labor skills that are all in place in order for things to move in a certain way so that you and I and everybody can go to the Apple store and get our earbuds or get our iPhone and so forth. Switching and changing this is again, just like taking this giant locomotive and just, you can't throw it on, you know, just on earth somewhere. You need to get the tracks and you need, you know, all the, if you imagine a locomotive, you know, all the stuff that's needed. What companies have been doing is since COVID, they had tried to expand, you know, Apple did this in terms of getting Foxconn to produce in India and so forth. They had already been moving and diversifying their supply chains in order to be ready. And I think that was a smart thing to do. I think without question, we know from a supply chain standpoint, relying on one source is dangerous. It's risky. So they had already begun to do that. But it takes a tremendous amount of time to do that and these tariffs had come on relatively suddenly. There has not been time to prepare. So while I think it is noble to talk about wanting to bring back production and manufacture to the United States. And I think in certain cases, I think it is important to do that. One, it takes time to do. It takes money and time. Regardless of the amount of money, you need a certain amount of time. You cannot just build a facility. I remember a couple of years ago, I don't know if you remember the contamination of baby formula that was going on. With baby formula, as a supply chain person, that is not nearly as complicated as dealing with AI, data centers, semiconductors, all of it. Even with baby formula, we couldn't just throw up and build up a facility because you need regulation and there are standards. When it comes to the kinds of products that we're talking about with AI, we're talking about the need for specific components that right now, some of them we can only access from China. So what I'm seeing companies doing stocking up. Everybody is stocking up. Individuals are stocking up. Companies are stocking up. They're loading their warehouses, their distribution centers with the goods that they have. There's lobbying going on for exceptions, exemptions of various types. There is also obviously efforts that are being made. NVIDIA is a great example of trying to really very rapidly ramp up production in the United States. Again, I think it's noble, important to do that. But I think that there is a tremendous amount of chaos. confusion and I think that if we zoom out and just say kind of what is happening and how is it going to impact AI I think without question I think at least For now, we will see prices going up. I think, I don't want to say stagnation, that's the wrong word, but I think a slowing down in innovation with regard, that rapid innovation when it comes to AI. From a business standpoint, one of the things, I'm a business professor, that I have urged businesses to focus on the AI technologies that are usable for them that meet their strategic objectives. I think it's going to force companies to focus, to really focus. That's a good thing. You had a guest on a couple months ago and I think he very notably pointed out the difference between artificial general intelligence and I think he noted on useful intelligence. And I could argue both ways because I think what is not useful at the moment but what is innovative eventually becomes useful. You have to have this innovation because we have so many innovations and those innovations may not be useful at the moment. but they will become useful. We start with that pie in the sky and then we eventually figure out how to use it in a million different ways. And it's important, because if we don't have those great innovations, we're just going to stay plateau out with what is useful. But for now, today, and in the immediate future, I think companies are focusing and targeting what is useful that's going to give them their greatest answer, their greatest needs, because they have no other choice. But I am concerned that that innovation is going to plateau out unless something changes. I'm very hopeful. On the other hand, answers to some of these scenarios are also coming about with AI, because AI capability is really pretty incredible. I mean, I think most of us use Gen AI in so many ways, and it is really remarkable in terms of, you know, take anything from clog to I'm talking about the latest versions and it's incredible in terms of how good they are and what they can do and what they can't. So I think AI can be very useful in answering and brainstorming, in answering some of these questions, how do we work around? What do we do? And I've constantly, through my career, and I maintain this, the human ingenuity with technology is the way to go. And I think AI may help business leaders right now brainstorm in terms of scenarios to address what is immediately facing them in terms of these political issues that I think are outside of the realm for You know, because we do talk about the big tech and we talk about NVIDIA, but we talk about the large companies. There are so many small and medium-sized companies that are being impacted that don't have the cloud, that can't get out there, and don't have the money to lobby and to do all that. They're being tremendously impacted. I think even on a small scale, AI can help brainstorm. What are the scenarios? My expertise, among other things, is in forecasting, and the toolbox of scenario planning is a very important one right now. AI can help with that. But this is very much an evolving, developing situation. As you know, and as everybody knows, we have a 90-day pause. That doesn't help businesses and supply chains very much because 90-day pause, 90 days isn't enough to build a factory. 90 days isn't enough to create the infrastructure needed. 90 days is merely enough to raise prices, to hoard and stock and wait to see what is going to happen. And I can keep talking, and it's just quite evolving and critical area.

Ian Krietzberg:
Yeah, definitely evolving, super critical. In terms of the idea of bringing this back, bringing this manufacturing back to the US, I think, and just to underline the point you made where, you know, Beyond the fact that this is a locomotive and once you once you cut power It takes a lot of pushing to get it back up to speed But that we don't have in the u.s. Access to you know, if you think about the the supply chain pipeline for chat GPT You know you're starting with rare-earth minerals to produce the chips and it's not just silicon. It's all these other materials to produce the chips that then go into NVIDIA's GPUs, which line the data centers, which are loaded up with hardware, server racks, etc., that come from and are manufactured in other places. And these are, you know, China has a near monopoly on certain of these rare earth minerals that are vital for semiconductor production. We don't we don't really have that. And so the idea of the manufacturing of specifically the semis, the semiconductors moving solely back to the U.S., seems not really achievable. But under the guise of that's the thing that they're trying to accomplish, and you mentioned NVIDIA, And Apple as well. In February, Apple was trying to get ahead of this, and they said we'll do half a trillion dollars in the US over the next four years. And NVIDIA says, you know, yesterday they said this, that they're producing GPUs, they're Blackwell chips at their plant in Arizona, and they're building two new plants in Texas with partners that are based in the US. But we have a very different environment here than in China. We might have a lack of available skilled workers to be working in those factories. The likely scenario, it seems to me, is that it will be much more expensive. Chips produced in America will likely be more expensive, tariffs aside, than those that the companies produce in China. They went to these other countries for a reason. And so, I guess, in a long way, circling back to this idea of, is it realistic? to actually ramp up full production several years down the line where these companies, you know, don't really need their factories in India or China or Taiwan, but they actually have a kind of vibrant U.S.-based manufacturing environment. And I guess is that realistic? And is what's going on here, could that help push us toward that?

Nada Sanders:
So, it's a really great question. I'll tell you why. I have thought, I remember during COVID thinking that there were certain things that were essential to have, to produce here in the U.S. that would, so that we would not be so vulnerable. Chips are definitely a part of it. I remember really, you know, you saw what was going on during COVID and what had happened. There was another one, it was, penicillin and I remember being really surprised thinking oh my gosh I am shocked that we, you know, how much of critical medications are sourced from China, India, Ireland, and so forth. And I remember, and I'm just a layperson in thinking these are all part of national security that I think would be really important. Having said that, it takes time to build that. So you ask me, is this realistic? I'm an operations and supply chain person. So when I think about operations, When I think about a factory or facility, automated or not, I think about innovation, I think about the actual structure of the facility, I think about machines, technology, so many cases that even when we produce items here, we actually purchase equipment for those assembly lines in China or elsewhere, or maintenance equipment. So it might even have the equipment here, but the equipment for maintenance, preventive maintenance, might be purchased elsewhere. So you have not just the raw materials, you know, this isn't just, you know, a lemonade stand where you've got lemons and you know sugar and water you're talking about packaging and assembly lines and tools and then equipment that that takes care of the tools so tools that take care of tools and then you have people that have to be trained in quality control and you're talking about items that have you know this six sigma plus quality control. I mentioned to you the baby formula because even that is so benign compared to the kinds of things that we're talking about here when we're talking about AI and then we talk about defense and aerospace and drones and aircraft and all of it. So these are very complex supply chains that have for years evolved into complex networks. So to answer your question, is it realistic? I think that possibly on a small number of items, on a narrow product line, with caveats such as that possibly in a few years innovation could get to the point where we're not using or relying so much of these on the rare earth minerals. But there's other kinds of chips or other kinds of innovations that, and I think that that is, when we look at innovation and the way that we do things, can we innovate product design? Because you have product design, you have process design. You have a process, we just talked about the process. of the way that we produce semiconductors and everything that goes into computers and all the way to data centers. And then there's the product itself and the innovation that goes into the product and those we should see and we should be incentivizing innovation in the product where maybe we are not relying on the rare earth minerals and I cannot tell you is that possible or not. I think when it comes to innovation I believe that anything is possible with the right creativity, the right know-how, incentives, money put forth towards it. But then we come full circle. We need money and time and I guess my concern is that we haven't done that and the tariffs are being put in place so we have this immediacy where we are shutting off the supply, the inbound supply of the goods that we need for the current products and processes in place, which the immediate effect is hoarding and price hikes that are there. I would love to see the monies that are being put forth to incentivize massively the innovation so that in two years, say, we are able to make these kinds of products, or at least some of them, here at home. When you talk about products, you're talking about the product itself and the product innovation. You're also talking about product variety among other things. You and I, there's so much we just talk about that. We talk about chips in this kind of a generic form, right? But how many different kinds of chips are there and what are they used for? And that's just chips, right? I'm not sure that we can do the full product variety, because every time you begin to kind of focus an area down the production process, you do product improvement, but then there's product variety, then you get into price. I think And I'm hoping that at least the listeners can get a sense of the kind of complexity that we're looking at here that goes into all of it. So I think, is it possible for us to be able to produce it here? It's possible in a few years with the right financial backing, the incentives. It's possible. I mean, anything is possible. But do I think that it's likely? It's very murky. The future is very, very murky at this point. And I think supply chains are really complex. They're entangled right now. I think companies are looking for immediacy. Their immediacy is, I already mentioned, the price hikes, the hoarding, really trying to get into Mexico, Vietnam, other countries where they're able to produce, manufacture, source from that are going to be cheaper, that are going to circumvent the tariffs. and engaging in other kinds of efforts in this very volatile time period. But I think for the average consumer you're going to see all the electronics, all the things associated with AI, I think plateauing in innovation and I think more expensive on the good side is I'm hoping it forces businesses to really begin to target what they need AI for and what's going to give them the greatest use with what they have at the moment.

Ian Krietzberg:
It's a tough nuance, and it's a tough thing to thread there for companies. I would not want to be Jensen Wong right now. But the kind of worst case scenario, I guess, which it seems like is playing out. We have the tariffs in place. The sweeping, the broader reciprocal tariffs are on pause, like you mentioned, for 90 days. With the assumption that they might go back in place We know the administration is exploring actively investigating semiconductor specific tariffs, right so so far Computers computer chips and smartphones are exempt from reciprocal tariffs, which tech companies are very very happy about but it has been promised that Semiconductor tariffs are coming, and these things would not be exempt from them. And obviously, like, we're talking specifically about AI, but, you know, as you mentioned, and I think this is another really good point, because we're not, as a people, we were never really trained to think about the full pipeline behind the product in hand, and semiconductor chips are in everything.

Nada Sanders:
They're in everything. And AI, even a year ago, two years ago, we talked about AI as this technology. It is immersed in everyday life right now. If you look at any kind of operation, it is from the marketing, the operations, procurement, managing supply chain optimization, inventory quality control. AI is embedded and is an essential partner an essential tool in the production process. Supply chains are complex, they're notoriously complex, and it is frustrating as a supply chain expert to see that some of these lessons have not been learned. I mean, for the love of God, think about what happened with stupid toilet paper during COVID that was not that long ago. Toilet paper is a great example of the inability or the difficulty in scaling. Scaling is a huge thing in terms of being able to scale production, to distribute, to switch production processes. Something so benign that has such a shallow bill of materials, meaning the ingredients, the components that go into this product, and then you take something as a server. as an iPhone, as any of the items, as a semiconductor, as a chip, and in everything that it impacts, as I mentioned, defense and aerospace and thinking of Everything that we do home appliances okay i need you know we don't have to go as far as defense and lucky martin we just look at you know our kitchen and see what that what that what that looks like so i think a is so in bed in our lives and everything that we do, But I think we have all forgotten the complexity of these supply chains to have these products and the innovations as we all see in our everyday life in terms of the capabilities. The capabilities, my gosh, of our AI assistants that we all use on our smartphones. So all of this comes from a global supply chain global ecosystem and that has been abruptly halted in ways that are really unknown. So I think that we are in this territory now where what we have AI can help us navigate and scenario plan but I am extremely hopeful that that we end up with some kind of a modifications on the one hand, a delay-halt negotiation on the one hand, in terms of these tariffs where cooler heads can prevail. And I am, and I have been for some time, hopeful that we still can continue to invest in domestic infrastructure so that we can build these capabilities, at least those that are absolutely essential and critical.

Ian Krietzberg:
Right. And to your point about AI being everywhere, I think the wave of generative AI has displaced in people's heads how embedded other forms of AI are, and this is why I don't love the term AI but it encompasses many many things and so I Now I suppose when we think of the term AI, you know, we're thinking about something like tragedy T Which is a large language model But we have adjusted to living with other forms of machine learning algorithms for a very long time everything you do on the internet is dictated by a machine learning algorithm every app and you download is loaded up with algorithms, cookies are traced, Netflix recommendations, YouTube recommendations, all across social media, it's all algorithms. Down into manufacturing, every factory has, to varying degrees, is loaded with automation technologies. And from quality control to there are robotics and robot systems that are being powered by Algorithms and so these kind of older machine learning and even older different types of algorithms are in everything And they're powered in different ways, you know older algorithms might not necessarily need to be run on GPUs They can maybe be run on CPUs which are less expensive But the construction of CPUs is also kind of comes under the the tariff stuff that's going on and so I guess, you know, you were just saying that you hope that there is a modification, which I think everyone is hoping for as well. I wonder, you kind of painted the scenario of if there is not a modification, we can expect price hikes. And I think we're already seeing inklings of this. The AI industry itself is a kind of tough one to examine because they are so funded by venture capital that they have been subsidizing the cost of generated AI using their venture capital dollars for a long time. And so people aren't really aware of how much it really costs. But now, Anthropic has a $100 monthly tier and a $200 monthly tier. OpenEye has a $200 monthly tier as well. And they're still losing money on those. And so the push to more expensive stuff, I think, is going to be echoed everywhere. And the challenge to innovation, because as it stands, a single NVIDIA GPU costs roughly $30,000. We're talking about data centers that are loaded with hundreds of thousands of GPUs, and you increase the price on one of those GPUs by $10,000, $20,000, and all of a sudden, your five billion in venture capital doesn't stretch as much. And right now, the main approach that companies have taken to powering AI innovation has been more chips, more GPUs. To your point, I think dealing with scarcity makes us more innovative, but the industry could be held back there. And so I wonder if we do get a modification, are we kind of past the point where that could have an impact? Because so much of what's going on here is just navigating unpredictability. As you mentioned, the tariffs that kind of came in place in the first place earlier this month, No one really had too much warning about what they were, who would be impacted, the extent of them. No one knew that they were going to be paused. And no one knows what's going to happen next. And so as companies are making commitments to spend hundreds of billions of dollars trying to respond to them, you know, if there was a pledge that all semiconductor tariffs are off the table for the next 18 months, get your stuff together, is that a thing that companies can take and use that time? Or does just the chaos of, but maybe tomorrow I'll change my mind on that, which is kind of just the pattern that we've seen so far, does that make it just really difficult?

Nada Sanders:
It's exactly right. I think that even if there was a pledge or a statement that we're going back, we're not going to have the tariffs anymore, I think that the level of uncertainty and volatility such that if you're running a business, if you're running a company, your priority is going to be to have as much cash on hand to be as liquid as possible, to stay afloat. You cannot afford to get yourself in a position where you are losing money or can't afford to run your business. And so I think that every business I know is really focusing on liquidity, on cash, on keeping the cash flow going. That means that, you know, raising prices as much as possible, hoarding as much inventory. That is also going to mean, I have to say, it's going to mean people losing their jobs. Because again, we're trying to really, as a business, you're trying to stay afloat. And I think that for the foreseeable future, until there is a confidence or a sense of calm, it's not going to be reversed. And I think that businesses need confidence and because of the level of volatility, one statement is not going to, would not create that level of confidence. So I think that when you had asked me, you know, could this in two years? Well, in two years, yes, I think you and I can imagine a future in two years, but I think that for the foreseeable future, for the immediacy, I think we are still looking at the very thing that I had mentioned with regard to price hikes, the hoarding, It is a very volatile environment because it keeps changing. And so I think that it's going to be a really bumpy ride and everybody's going to kind of hold on to what they have. But I think that we are looking at businesses trying to cut costs in every way they possibly can. And I think everyone I know is scrambling in in terms of supply chains. Can we, again, do it out of Vietnam or Mexico? Can we build something here? But again, you don't have the talent necessarily. You don't have the facilities. How quickly can you do that? You need cash to do that. Are there going to be subsidies? One of the answers could be if these businesses were to see large subsidies to be able to help facilitate this. One of my concerns also, and you and I have just talked about this, with regard to AI, it's embedded in everything, even for small companies, in terms of managing the market in ways that users, consumers aren't even aware of, that you and I, most people, again, they think of LLMs like chat GPT and things that we use, but they don't realize in how many ways these are embedded in the production process, in visual systems that monitor quality control and sensors and everything. And all of that costs a tremendous amount of money. Companies, again, are going to be forced to pare down and to focus on what is essential. No question about it. I am a true believer as an operations person in the 80-20 principle. I believe this for every business, so you go after the inventory items, the customers, every aspect of your business that is going to give you the greatest return and you might have to parse out and eliminate those that do not. And again, that's going to impact product variety, you narrow down, you narrow down your customers, you cut costs, all of that. Those are the kinds of things that we are going to be seeing, but I do especially worry about small in medium-sized companies. You know, there are so many of the smaller businesses that barely survived COVID, that barely got through COVID and were rebounding finally and getting their balance sheets back up. This is a really big deal for them in their ability to function and I'm extremely worried that we're going to see a lot of bankruptcies.

Ian Krietzberg:
I guess my last kind of thing on that point as well is the scale of capital required to dance between the raindrops, as it were, of whatever's going on here. Folks like Apple and Nvidia have the capital on hand to do that.

Nada Sanders:
That's right.

Nada Sanders:
Yes. And I think that, you know, we all know, I mean, competition is essential to keep the prices at a level that we can all access goods. And we've gotten so comfortable, all of us. And my concern is that it is going to get to the point where we see bankruptcies, we see a few companies dominate keeping their prices high, the access to services, goods and services, I think has the potential to become extremely limited. And I am really, really concerned that those that have are going to be, that the large companies are going to be able to afford some of these things, the Apples, NVIDIAs, and I do very much worry that the smaller ones are going to vanish. And that, I think, is a loss for all of us.

Ian Krietzberg:
Well, Nada, we're going to cut it there. And, you know, I guess you and I both will be closely watching whatever happens next. But, you know, definitely a bumpy ride. But thank you so much for coming on and trying to unpack the breadth of what's going on here.

Nada Sanders:
Thank you for having me on anytime, anytime.