Is the AI chip buildout running out of steam—or is demand for the machines that make those chips still accelerating? ASML says some customers are interested in placing equipment orders for 2029 and even 2030. That is interest, not confirmed future orders, but it reveals how far ahead chipmakers are planning.
The Daily AI Chat brings you the most important AI story of the day in just 20 minutes or less. Curated by our human, Fred and presented by our AI agents, Alex and Maya, it’s a smart, conversational look at the latest developments in artificial intelligence — powered by humans and AI, for AI news.
The Daily AI Chat. Welcome to The Daily AI Chat. This podcast was curated by our human friend, Fred, and we are your dedicated AI hosts.
Speaker 2:Yeah. It is it is really great to be here with you today.
Alex:So right now, you know, if you look anywhere in the tech world, everyone is asking you the exact same question. Like, is this massive AI spending boom an actual permanent shift in how our global economy works? Or are we just watching a spectacularly expensive hype cycle that is, you know, about to just run out of steam and crash?
Maya:I mean, is basically the defining technological question of our time right now. Every venture capitalist, software developer, and tech enthusiast out there is just desperately searching for a reliable signal to tell them what happens next.
Speaker 1:Exactly. And to find that signal for you today, we aren't going to look at the flashy software companies, you know, the ones making the chatbots.
Speaker 2:Right.
Speaker 1:You just won't find the answer in an app. Instead, we are looking at the heavy machinery. We are going to look at the company that makes the machines that actually make the chips.
Speaker 2:Yes.
Speaker 1:Specifically, we are diving into an 10/14/2026 Reuters report by Nathan Viflin and Toby Sterling. And they covered the latest financial outlook for a Dutch company called ASML. So... Okay. Let's unpack this, starting with just the sheer scale of the money moving around right now.
Speaker 2:Yeah. The numbers in this Reuters report are... Frankly, they are just staggering. They offer a really rare, completely unfiltered look into the physical supply chain of artificial intelligence. Because, you know, behind every single piece of code.
Speaker 1:There is a physical piece of silicon.
Speaker 2:Exactly.
Speaker 1:And the
Speaker 2:money physical object.
Speaker 1:And the money tied to that silicon is just mind blowing. Let's look at ASML's scoreboard for the third quarter of twenty twenty six. They just announced a massive €11,700,000,000 in total net sales for q three. Right. And out of that, their net income, so their actual profit, was about €4,960,000,000 for a single quarter.
Speaker 2:I mean, just think about that margin for a second. Keeping almost €5,000,000,000 in pure profit on 11,000,000,000 in sales over just three months.
Speaker 1:It's wild.
Speaker 2:It is an insane margin, honestly. It really just goes to show what happens when you are essentially the only company on earth that knows how to build a specific essential product.
Speaker 1:Right. But the report didn't just stop at what they have already made. They also laid out their projections. So ASML gave their guidance for the fourth quarter expecting between 14,200,000,000 and €15,200,000,000 in sales. Yeah.
Speaker 1:And because of that really strong finish to the year, they actually raised their full year 2026 sales outlook to somewhere between 44,000,000,000 and €45,000,000,000.
Speaker 2:Now wait. Before we go any further, I think it is absolutely vital that we draw a hard line for you between these two sets of numbers. Okay. Because it is a distinction that constantly gets lost in the excitement of tech reporting, but it is just crucial for understanding the real landscape.
Speaker 1:Wait. I gotta stop you there. I mean, 44,000,000,000 in Outlook versus the 11,700,000,000 they just banked in a single quarter. Isn't that basically a guaranteed trajectory at this point? Like, if they are already in q four, why are you drawing such a hard line between those numbers?
Speaker 2:I am drawing a line because we have to separate achieved results from company guidance. The q three numbers, you know, the 11,700,000,000 in sales, those are achieved results. That is money in the bank. Right. Contracts fulfilled, heavy machinery literally loaded onto planes and delivered.
Speaker 2:It is a historical unchangeable fact. But that massive 44 to €45,000,000,000 full year outlook, that is company guidance.
Speaker 1:Right. It's a corporate forecast.
Speaker 2:But isn't it usually highly accurate? Well, it is an educated forecast. Yeah. Formulated by brilliant financial minds, but it is not a guaranteed or achieved result. Guidance is essentially a company telling the market, hey.
Speaker 2:Based on our current order book and everything we see today, here is what we confidently expect to happen.
Speaker 1:But things change.
Speaker 2:Exactly. Physical realities, supply chain shocks, geopolitical incident, or just sudden shifts in global economics can always alter that trajectory before the year closes. So as we evaluate the actual tangible strength of the AI boom, we have to anchor ourselves in the achieved results while treating the guidance as exactly what it is. A confident roadmap, not a finalized receipt.
Speaker 1:That is a critical distinction to keep in mind. Yeah. Especially with the sheer volume of cash we are tracking here, which I guess brings up the multi billion euro question.
Speaker 2:Right.
Speaker 1:We have got a single Dutch company pulling in €5,000,000,000 in pure profit in three months based on achieved sales. What exactly is triggering this gold rush? Like, what physical object are these tech giants actually buying?
Speaker 2:They are buying the ability to participate in the future of computing, Literally. Without the specific machines that ASML builds, the modern AI ecosystem physically cannot exist.
Speaker 1:Okay. Let's make this tangible for everyone listening. Think of ASML's equipment as a printing press. But we aren't talking about, you know, Gutenberg pressing ink onto paper.
Speaker 2:No. Not quite.
Speaker 1:We are talking about a hyper advanced printing press that builds the literal brains of artificial intelligence. And instead of using ink, this machine uses light.
Speaker 2:Extreme ultraviolet light to be exact.
Speaker 1:Right. It uses its light to print impossibly microscopic pathways, basically the circuitry onto wafers of silicon.
Speaker 2:That printing press analogy is a really great starting point. But we do need to elevate it a bit to understand why it is so difficult. Because, you know, Johann Gutenberg didn't have to fight the fundamental laws of physics to print a book.
Speaker 1:Right. He just carved some metal letters and stamped them. What is ASML doing that is so fundamentally different?
Speaker 2:Well, it comes down to scale. The pathways we are talking about are measured in nanometers. We are talking about printing structures that are fractions of the size of a human virus.
Speaker 1:Which is just incomprehensible.
Speaker 2:It really is. You can't use physical tools to carve pathways that small because physical tools are simply too thick. So you have to use light to basically draw the circuits. But here's the problem. Normal visible light has a wavelength that is too wide.
Speaker 1:Oh, I see.
Speaker 2:It is like trying to draw a microscopic portrait with a giant thick magic marker.
Speaker 1:So the light itself is literally too fat to draw the tiny circuits.
Speaker 2:Exactly. So ASML had to harness extreme ultraviolet light or EUV. And EUV has an incredibly short skinny wavelength. But, EUV light is terribly fragile. It gets absorbed by almost everything including just the air we breathe.
Speaker 2:Wow. Yeah. So inside this printing press, the entire process has to happen in a total vacuum. And you can't use normal glass lenses to focus this light because the glass would just absorb the light.
Speaker 1:So what do they use?
Speaker 2:You have to bounce the light off a series of the most flawlessly polished mirrors ever created by human beings, made by a company called Zeiss.
Speaker 1:That is just mind blowing. They are literally bouncing invisible fragile light off flawless mirrors inside a vacuum chamber just to draw microscopic roads on a piece of silicon.
Speaker 2:Yes. And the VIP client list buying these machines is basically the undisputed who's who of global technology. The Reuters report highlights exactly who is waiting in line for these. You have TSMC, Samsung, Intel, SK Hynix, and Micron. Wow.
Speaker 2:These are the absolute titans of semiconductor manufacturing.
Speaker 1:And that client list perfectly highlights the bottleneck here. Like, why does this specific insanely complex EUV light printing process matter so much to AI? Why couldn't tech companies just use, I don't know, the older, slightly less advanced chips for their AI models?
Speaker 2:Because artificial intelligence, specifically the massive large language models and neural networks driving this current boom, well, it requires an astronomical amount of computational power.
Speaker 1:Right.
Speaker 2:And to get that computational power without literally melting down the data center or requiring the energy output of a city, the chips have to be incredibly dense. You need billions of transistors packed into a space roughly the size of your fingernail.
Speaker 1:So the pathways have to be smaller, which means the printing press has to be infinitely more precise, which brings us right back to those flawless mirrors and the EUV light.
Speaker 2:Precisely. And ASML is effectively the only company in the entire world that can manufacture these EUV lithography machines at the required scale and precision. They hold a virtual undeniable monopoly on the bleeding edge of this technology.
Speaker 1:Which is huge.
Speaker 2:It is. Therefore, lithography is the absolute choke point of the entire global AI infrastructure. I mean, if an American tech giant wants to train a massive new AI model, they need the most advanced chips. If they want those chips, they need a manufacturer like TSMC or Samsung to make them.
Speaker 1:And to make them.
Speaker 2:Right. If TSMC or Samsung wanna make them, they absolutely must have ASML's machines. If you don't have ASML, you simply do not have AI chips.
Speaker 1:It is a total funnel. The entire industry's grand trillion dollar ambitions have to physically pass through this one company's factory floor in The Netherlands.
Speaker 2:Exactly. And when you have an entire global economic shift funneling through a single manufacturing choke point, well, high demand begins to violently collide with physical reality.
Speaker 1:Okay. Here's where it gets really interesting. Because the Reuters report highlights that ASML's currently planned supply for the years 2027 and 2028 might not actually be enough to satisfy the demand forecast coming from their customers. Yeah. I mean, just put yourself in ASML shoes for a second.
Speaker 1:You hold a monopoly on the most important machinery in the modern world. You look at the next couple of years and you have to tell your clients, we actually might not be able to build these things fast enough for you.
Speaker 2:Which is a terrifying prospect for the tech industry really because the physical reality of manufacturing these machines is incredibly complex.
Speaker 1:Right. I mean, my immediate thought is just, well, build another factory, build more machines.
Speaker 2:Yeah. But it is not that simple. You cannot just flip a switch and double production. These lithography machines, they are roughly the size of public buses.
Speaker 1:Good grief.
Speaker 2:They cost hundreds of millions of dollars each. They are so large and delicate that shipping just one of them requires multiple seven forty seven cargo planes.
Speaker 1:That is insane.
Speaker 2:And remember those flawless Zeiss mirrors we talked about or the vacuum chambers? The supply chain for the parts of the machine is just as complex as the machine itself. You can't build more EUV machines if your suppliers can't build more hyper specialized mirrors.
Speaker 1:Oh, of course.
Speaker 2:So when demand outstrips planned supply, management just has to scramble.
Speaker 1:So what is the game plan here? How is ASML trying to fix this impending logjam?
Speaker 2:Well, the Reuters piece detailed the playbook directly from ASML CEO Christophe Fouquet. He outlined three primary strategies to address this 2027, 2028 capacity crunch.
Speaker 1:Okay.
Speaker 2:First, they are actively studying ways to expand their 2028 manufacturing capacity beyond their initial plans. They are looking at their global footprint, their suppliers, and trying to figure out how to physically build more machines.
Speaker 1:Okay. So that is the build more factories approach, even though we know how incredibly hard that is. What else?
Speaker 2:Second, they are heavily relying on productivity upgrades to their existing equipment. Essentially, if ASML can't magically double the number of machines they sell, they wanna upgrade the software and the mechanics of the machines that are already out in the field.
Speaker 1:Oh, I see. Making the existing printing presses run faster so their clients can squeeze more chips out of the hardware they already own.
Speaker 2:Correct. It is a smart way to increase global chip output without having to build an entirely new bus sized machine. But the third part of the playbook is the most economically significant.
Speaker 1:What's that?
Speaker 2:Fouquet openly discussed the company raising prices. They are looking at price hikes starting in 2027 with the financial effects becoming much more visible on their balance sheets in 2028.
Speaker 1:I gotta jump in here because the sheer audacity of this pricing power is just wild. Just imagine running a business where demand is so rabid and your product is so utterly irreplaceable that you are desperately trying to figure out how to build more factories than you originally planned just to keep up.
Speaker 2:Right.
Speaker 1:And because you are literally the only game in town, you can simultaneously say, oh, and by the way, we are raising the prices on these hundreds of millions of dollar machines. That is a level of leverage that most businesses literally only dream of.
Speaker 2:It is the textbook economic definition of inelastic demand. When a product is an absolute necessity, meaning the buyers must buy the product regardless of a price increase in order for their own businesses to survive, the supplier holds all the leverage.
Speaker 1:Yeah. They have nowhere else to go.
Speaker 2:Exactly. TSMC cannot afford to stop buying ASML machines just because they cost more. Because if they do, they lose their own clients to Samsung. But, you know, while 2027 and 2028 look incredibly packed with this frenzied inelastic demand, technology moves incredibly fast. We really have to look further down the road to understand the true health of this boom.
Speaker 1:Which brings us to the end of the decade. The 2029, 2030 horizon. The Reuters report mentions that ASML customers are already expressing interest in ordering equipment for 2029 and sometimes even 2030.
Speaker 2:Yes. And this is where the nuance of financial reporting and supply chain mechanics becomes critically important for us to unpack.
Speaker 1:To make sense of this, let's use another analogy. Think of it like a dinner reservation. It is like calling up a wildly popular, highly exclusive restaurant and telling the host, hey, we might wanna celebrate our anniversary there in 2029. Keep us in mind.
Speaker 2:Yeah. That's a good way to look at it.
Speaker 1:Now, that is obviously a great sign for the restaurant. Means people love the food and wanna come back. But crucially, it is not a nonrefundable deposit. You haven't paid for the meal yet.
Speaker 2:Exactly. But right now, the stock market is behaving as if that anniversary dinner has already been eaten and paid for in full, and that is where the danger lies when interpreting these reports. What's fascinating here is how the market interprets these signals versus what they actually represent mechanically. We must rigorously distinguish between expressions of interest and actual booked orders or guaranteed revenue.
Speaker 1:Walk us through the difference there. What does an expression of interest actually look like on paper?
Speaker 2:An expression of interest is basically a placeholder. It is a client like TSMC or Intel looking at their own long term models and saying to ASML, hey, We anticipate needing a certain amount of manufacturing capacity by 2030. So let's start the conversation about reserving space in your production line.
Speaker 1:Makes sense.
Speaker 2:A booked order on the other hand is a signed contract with capital firmly committed and strict cancellation penalties in place. Those 2020 '9 and twenty thirty conversations are currently just expressions of interest. They are, as you said, the dinner reservations.
Speaker 1:And yet Reuters framed this exact detail, this interest in 2029 and 2030, as something that was actively easing investor fears about the AI boom running out of steam. Investors saw people making dinner reservations for four years from now and just breathed a massive sigh of relief.
Speaker 2:Which is a deeply human reaction to uncertainty. I mean, the tech industry hates uncertainty. And right now, the greatest uncertainty is whether the hundreds of billions of dollars being poured into AI data centers will actually yield long term sustainable profits.
Speaker 1:People wanna know it's gonna pay off.
Speaker 2:They are desperate for any sign that the music won't stop. So when a foundational structural company like ASML says, our clients are already talking about 2030, observers latch onto that as absolute proof of longevity.
Speaker 1:But we need to look much closer at that. Right? Because an expression of interest doesn't actually guarantee that those AI data centers will be profitable.
Speaker 2:Exactly. I wanna step in here and really synthesize the tension we are seeing. The strong present sales, that €11,700,000,000 quarter, and the frantic scramble to increase manufacturing capacity in 2027 and 2028 absolutely support a near term expansion case. The immediate future of AI infrastructure build out is happening right now, and the scale is massive.
Speaker 1:Right. The money is being spent today. The heavy machinery is moving today.
Speaker 2:But distant interest from clients regarding 2030 does not actually prove that AI spending will maintain this blistering pace until the end of the decade. Let me be perfectly clear here. No one can guarantee the AI boom will continue forever.
Speaker 1:That's a great point.
Speaker 2:And as a vital reminder, our discussion today is an analysis of technology infrastructure and supply chain dynamics. It is not investment advice.
Speaker 1:Right. Definitely not.
Speaker 2:We are analyzing the physics and the factory floor, not telling you what to do with a stock portfolio.
Speaker 1:Absolutely.
Speaker 2:We are looking at the machinery and the bottlenecks.
Speaker 1:Exactly.
Speaker 2:And to determine if this is a durable decade long build out rather than just an overheated hype cycle, we would need to see different evidence down the line. We wouldn't just look at interest. We would need to see those twenty twenty nine conversations convert into firmly booked noncancelable orders.
Speaker 1:That would be the real proof.
Speaker 2:Right. But more importantly, we would need to look further up the chain. We would need to see the companies buying these chips, the software developers, the cloud providers, actually generating sustained massive profits from their AI services. If everyday consumers and enterprise clients don't start paying premium prices for AI software, the cloud providers will eventually stop buying new chips.
Speaker 1:And if they stop buying chips, the chipmakers like TSMC and Samsung will suddenly stop buying the EUV machines, and all those 20, 30 expressions of interest will just evaporate overnight.
Speaker 2:Precisely. It is a massive, incredibly expensive domino effect. The foundation is currently strong. The concrete is being but the house is still being built and we really don't know who is going to live in it yet.
Speaker 1:So what does this all mean? Let's bring all these threads together for you. We are looking at a landscape where ASML, the sole manufacturer of the printing presses for the AI revolution, is making nearly €5,000,000,000 in pure profit in a single quarter today. Right. Because of the insane physics of extreme ultraviolet lithography, they are the undeniable choke point for the entire global AI industry.
Speaker 1:They are stretched so thin that they are scrambling to increase their manufacturing capacity for 2027 and 2028. And they are preparing to raise prices because demand is incredibly inelastic. Yes. And while they are seeing hopeful interest from clients looking as far ahead as 2030, those are just dinner reservations, not signed contracts. The near term boom is very real and very physical, but the end of the decade reality is still just an educated forecast.
Speaker 2:You have summarized the structural reality perfectly. The machines are being built. The capacity is stretched to its absolute physical limits, and the broader tech industry is holding its breath hoping the demand remains permanent. But looking at these supply chain mechanics, this raises an important question.
Speaker 1:Oh, I like where this is going. What's the question?
Speaker 2:Well, if hardware choke points like ASML are actively planning price hikes for their machines in 2027 and 2028 due to this immense demand, how will those massive infrastructure costs ripple down the chain?
Speaker 1:That is a great point.
Speaker 2:Right. Because the chipmakers will have to pay more for the machines, so they will inevitably charge the tech giants more for the silicon chips. The tech giants will pay more for the chips, so they will charge developers more for the computing power. Ultimately, will everyday software companies and you, the end user, eventually be the ones forced to foot the bill for these multibillion dollar printing presses?
Speaker 1:That is a phenomenal question to leave off on. Because at the end of the day, no matter how advanced the technology gets, someone always has to pay the check.
Speaker 2:They certainly do.
Speaker 1:Thank you so much for joining us and exploring the hidden mechanics of the AI boom. We hope this deep dive helped cut through the noise and give you a clearer picture of what is actually happening on the factory floor, far away from the software hype.
Speaker 2:It was a real pleasure sharing this analysis with you. Farewell.
Speaker 1:Take care everyone. Come again. BNI is evolving. And we appreciate you tuning in.