# The One Component That Holds Back Every New Chip

An episode of [The Hardware Podcast with Fexingo: Chips, Devices, and Electronics Engineering Conversations](https://invisiblenetwork.ai/podcast/1896799982), 30 May 2026, 8:28.

Audio: https://audio.fexingo.com/business/the-hardware-podcast/episode-0021.mp3

Every new chip design hits a bottleneck before it ever reaches a fab: the humble photomask. In this episode, Lucas and Luna unpack why advanced lithography masks now cost millions of dollars, how a single defect can scrap a whole batch, and why the mask shop is becoming the hidden gatekeeper of Moore's Law. Specific case: the multi-beam mask writer that ASML's sister company bought for $300 million, and why chipmakers are running out of ways to inspect what they can't see. #Photomask #Lithography #Semiconductor #ChipManufacturing #EUVLithography #ASML #MaskInspection #Moore #ElectronBeam #Technology #Engineering #Hardware #FexingoBusiness #BusinessPodcast #TheHardwarePodcast #Chips #Devices #SemiconductorManufacturing Keep every episode free: buymeacoffee.com/fexingo

## Transcript

Lucas: Every new chip design, no matter how brilliant the architecture, hits a bottleneck long before it ever reaches a fabrication plant. And it's a component most people have never heard of.
Luna: You're talking about the photomask, aren't you? The glass plate with the circuit pattern.
Lucas: Exactly. The photomask is essentially the master template for lithography. You take a quartz plate coated with chrome or an absorber material, you write the circuit pattern onto it with an electron beam, and then you use that mask to project the pattern onto the silicon wafer. One mask set for a leading-edge chip can run over a hundred layers.
Luna: And each one of those masks has to be perfect. Because any defect on the mask gets printed onto every single wafer.
Lucas: Right. A single sub-resolution defect—something you can't even see with an optical microscope—can kill the yield of an entire batch of wafers. And when you're talking about a chip design that costs hundreds of millions to develop, the mask becomes an enormous risk.
Luna: So how do you inspect something you can barely see? Especially with extreme ultraviolet lithography, where the wavelength is 13.5 nanometers—way below what any conventional lens can resolve.
Lucas: That's the central problem. With EUV masks, the reflective multilayers are only about forty pairs of molybdenum and silicon, each layer just a few atoms thick. A tiny bump or pit in those layers can scatter light and ruin the pattern. And because EUV masks are reflective, not transmissive, the inspection tools have to work at the same wavelength or use electron beams.
Luna: And electron-beam inspection is slow. It can take hours to scan a single mask.
Lucas: Hours per mask, and a chip manufacturer might need twenty, thirty, fifty masks for one design. That creates a real bottleneck in the ramp to volume production. In 2023, the industry shipped about twenty-eight hundred mask sets for leading-edge nodes. That's not a lot. And each mask set can cost anywhere from fifty thousand dollars for a mature node to over a million for an advanced EUV mask.
Luna: I've heard that some masks now cost three million dollars or more. Is that true?
Lucas: It is. For the most complex layers at three-nanometer and below, especially those using multi-patterning or advanced optical proximity correction, a single mask can run two to three million. And you need dozens. So the mask set alone can exceed a hundred million dollars.
Luna: That changes the economics. It means only the biggest players—TSMC, Samsung, Intel—can afford to play at the leading edge.
Lucas: It's become a barrier to entry. But there's a less obvious problem: the mask writing itself. The electron-beam tools that carve the pattern into the mask are incredibly precise, but also incredibly slow. To increase throughput, the industry is moving from single-beam writers to multi-beam writers.
Luna: Multi-beam—so instead of one electron beam scanning the mask, you have thousands of beams working in parallel?
Lucas: Exactly. The company that dominates this space is a Dutch firm called ASML—actually, it's a subsidiary, ASML Brion, but the key acquisition was in 2016 when ASML bought a company called HMI for about three hundred million dollars. HMI's multi-beam mask inspection tools are now critical for EUV. But even with multi-beam, the inspection throughput is barely keeping pace with demand.
Luna: And then there's the repair side. If you find a defect on a mask, can you fix it?
Lucas: Sometimes. Mask repair uses focused ion beams or electron beams to remove material or deposit it. But the repair itself can introduce new defects. And at these dimensions, the margin for error is essentially zero. So the mask shop has become the hidden gatekeeper. If your mask isn't perfect, your fab never gets the chance to make good chips.
Luna: It's a fascinating bottleneck. And one that doesn't get a lot of attention compared to stuff like EUV sources or high na optics.
Lucas: Right. The mask is where the design meets the physical world. It's the moment where abstraction becomes matter. And that transition is getting harder with every node.
Luna: I want to ask about something else—the pellicle. That's the thin membrane that goes over the mask to protect it from particles. But I've read that for EUV, the pellicle itself is a huge challenge.
Lucas: Huge. For deep ultraviolet lithography, the pellicle is a transparent polymer film. But for EUV, there's no transparent material—everything absorbs. So the pellicle has to be an extremely thin membrane, typically made of polysilicon or a carbon-based material, that transmits over ninety percent of the EUV light. It also has to withstand the heat from the EUV source, which can melt or warp it.
Luna: And if the pellicle fails, the mask gets contaminated. And you have to scrap it or clean it, which is risky.
Lucas: Exactly. ASML and its suppliers have been working on EUV pellicles for over a decade. They finally started shipping them in volume around 2022, but the yield and cost are still challenges. And as we go to high na EUV, the pellicle requirements become even more stringent.
Luna: So the mask ecosystem—writing, inspection, repair, pellicle—is this whole hidden industry that's essential to Moore's Law.
Lucas: It is. And it's a market worth about five billion dollars annually, growing maybe six to eight percent a year. But the interesting thing is that the number of mask shops is shrinking. There used to be dozens of independent mask makers. Now the merchant market is dominated by just a few companies: Toppan Photomask, Dai Nippon Printing, and Photronics. The captive mask shops at Intel, TSMC, and Samsung handle their own leading-edge work.
Luna: Which means the mask supply chain is concentrated. That's a vulnerability.
Lucas: It is. And we saw that during the pandemic, when mask supply tightened and lead times stretched. The industry is now trying to build more redundancy, but it's not easy when each mask shop requires hundreds of millions in capital equipment.
Luna: It reminds me of the glass substrate shortage we've seen for other industries. A single, obscure component becomes the critical path.
Lucas: Exactly the same dynamic. And with chip design complexity exploding—we're seeing chips with over a hundred billion transistors—the mask set is becoming one of the biggest cost and schedule risks in the entire development cycle.
Luna: If today's conversation gave you a new appreciation for the humble mask, it's exactly the kind of deep dive we try to do here. And quick honest thing—this show stays ad-free because a handful of listeners chip in through buy me a coffee dot com slash fexingo. That literally funds making this many episodes.
Lucas: Yeah, it's a small group, but it makes a huge difference. We don't have sponsors, so listener support is what keeps us going. And it allows us to pick topics like this, which aren't going to get millions of clicks but are genuinely important to understand.
Luna: Exactly. So if you find value in these conversations, that link is buy me a coffee dot com slash fexingo. No pressure, just putting it out there.
Lucas: And back to the mask—one more thing that's on the horizon is the move to high na EUV. That's going to require entirely new mask infrastructure. The masks themselves will need to be larger, flatter, and more defect-free than ever.
Luna: And the inspection tools will have to scale too. It's an ongoing engineering challenge.
Lucas: It is. And it's a reminder that the most important components are often the ones you never see. The mask is the unsung hero—or the unsung bottleneck—of every advanced chip.
Luna: Well said. That's a good place to leave it.

Shows are listed from Apple Podcasts' charts and directory, and Podcast Index where it is used. Every show and episode belongs to its publisher and plays from the publisher's own feed; this site hosts none of them.