SpaceX reportedly seeking $40 billion debt package for Nvidia AI hardware

Skyler Monroe (0:10)

Hey, everyone, welcome back to the AI Hardware Show. I'm your host, Skyler Monroe, and if you're into chip silicon and the hardware that's actually making the AI revolution happen, you are in exactly the right place. Big thanks to our sponsors today, Ada, helping businesses integrate AI into their workflows. Ago Consulting, that's Ago, your go-to for silicon development, from AI to SAUX and Zen Semiconductor, the umbrella powering a whole portfolio of silicon industry ventures, including Sierra RISC vCPUs, Loom AI Fabric, and Ada for deploying AI at scale. All right, we've got a packed episode today. SpaceX is reportedly looking to borrow a mind-bending amount of money to buy Nvidia hardware. Europe's semiconductor inspection market is heating up in a big way.

Russian datacenters just got hit by Ukrainian drones, and the fate of two supercomputers is unknown. GlobalFoundries and TSMC just struck a massive $2 billion partnership deal, and TSMC's September sales just absolutely exploded. Let's get into it. Okay, first story, and I want you to sit down for this one if you aren't already.

SpaceX is reportedly seeking a $40 billion debt package, and the goal is to buy Nvidia Rubin AI accelerators. Not a handful, not a rack. We're talking 360,000 of them. That is a staggering number, and it raises a ton of really interesting questions about where AI infrastructure spending is headed. So first, let's just put that number in perspective.

$40 billion, that's not equity. That's debt, meaning SpaceX would be borrowing that money. And what they want to do with it is essentially build out a massive AI compute infrastructure using Nvidia's next-generation Rubin architecture.

Rubin, for those who haven't been following along, is Nvidia's successor to Blackwell, and it's designed to push even further on the performance-per-watt curve for large-scale AI training and inference workloads. Now, 360,000 accelerators is just a wild number to wrap your head around. To give you an analogy, if each one of those accelerators was a high-performance sports car, you'd have enough to fill over a dozen NFL stadiums. That's the kind of scale we're talking about. This isn't just buying hardware for internal use. This screams ambition to become a serious AI compute provider, possibly to compete with the hyperscalers.

Here's what makes this really interesting from a hardware perspective. Rubin is built on an advanced packaging architecture that stacks memory and compute in ways that traditional chip designs just can't match.

We're talking about high-bandwidth memory, next-gen NVLink interconnects, and a whole new substrate level integration strategy from Nvidia. So SpaceX wouldn't just be buying chips. They'd be investing in an entire ecosystem of networking, cooling, and power delivery infrastructure to make those chips actually work at that scale. The debt financing angle is also worth unpacking. When you're buying this much hardware, you can't just write a check, even if you're SpaceX.

Hardware at this scale requires long procurement timelines, massive facility buildouts, and ongoing operational costs. Debt financing lets them move fast and lock in supply before competitors can snap it up. In a world where Nvidia lead times can stretch out 6 to 12 months or more, getting in line early with a commitment like this is a serious strategic move. This also tells us something important about the AI hardware market broadly.

The demand for accelerators isn't just coming from the usual suspects. The Googles, the Microsofts, the Amazons of the world, now you've got aerospace and tech conglomerates like SpaceX making 8, 9, 10-figure bets on compute infrastructure. The addressable market for these chips is expanding rapidly, and Nvidia is sitting right at the center of it. This is a remarkable story, and I suspect we're going to be hearing a lot more about it as details emerge. All right, moving on to story 2

And this one is a little more in the weeds, but trust me, it matters a lot. Europe's semiconductor metrology and inspection equipment market is forecast to grow at a compound annual growth rate of 7.4 percent from 2026 through 2033 That might sound like a dry financial headline, but the reasons behind that growth are deeply tied to AI chips and advanced packaging, and that's where it gets really interesting. So let me back up and explain what metrology and inspection equipment actually is, because it's one of those things that doesn't get nearly enough attention. When you're manufacturing a chip, especially at the cutting edge, we're talking 3 nanometer, 2 nanometer process nodes, you're patterning features on silicon that are literally smaller than a virus. You cannot see them with the naked eye, you can't even see them with a standard optical microscope. So you need incredibly sophisticated tools to verify that the chip you just made actually matches the design you intended. Metrology is the science of measurement. It's how you confirm that your transistors are the right size, your interconnects are properly formed, your layers are aligned correctly. Inspection is about finding defects, tiny little imperfections that could cause a chip to fail, either immediately or years down the line. Companies like KLA, ASML and Onto Innovation make these tools, and they are absolutely mission critical to semiconductor manufacturing. No metrology, no chips.