# How Chips Are Putting Antennas on Silicon for Next-Gen Wireless

An episode of [The Hardware Podcast with Fexingo: Chips, Devices, and Electronics Engineering Conversations](https://invisiblenetwork.ai/podcast/1896799982), 30 Jul 2026, 8:54.

Audio: https://audio.fexingo.com/business/the-hardware-podcast/episode-0141.mp3

Over the last five years, engineers have figured out how to etch antennas directly onto silicon chips, eliminating the need for external antenna components in millimeter-wave systems. This episode explores the physics challenges – antenna sizes, dielectric losses, and interference – and highlights a real-world case: the 2025 launch of a commercial phased-array chip from a startup called MixComm, now shipping to 5G infrastructure providers. Lucas and Luna discuss how on-chip antennas are reshaping device design for 6G, automotive radar, and satellite communication. #OnChipAntennas #MillimeterWave #5G #6G #SiliconPhotonics #PhasedArray #MixComm #Semiconductor #Hardware #AntennaDesign #Wireless #Technology #Fexingo #FexingoBusiness #BusinessPodcast #Chips #HardwareEngineering #TechnologyPodcast Keep every episode free: buymeacoffee.com/fexingo

## Transcript

Lucas: If today's tech conversation gave you something usable, you might already know that antennas are usually separate components, soldered onto a board. But that's changing fast. In the last few years, engineers have started putting antennas directly onto the silicon die itself.
Luna: Wait, like the antenna is etched into the chip? How does that even work when antennas are typically measured in wavelengths, and chip dimensions are tiny?
Lucas: That's the key insight: it only works at higher frequencies. At millimeter-wave, say 28 gigahertz for 5G, the wavelength is about a centimeter. So an antenna element can be a few millimeters, which fits comfortably on a chip die. It's not practical for Wi-Fi at 2.4 gigahertz, but for 5G and beyond, it's viable.
Luna: So we're talking about frequencies above 20 gigahertz where the wavelength shrinks. That makes sense. But what about the material? Silicon isn't exactly a great antenna substrate.
Lucas: Right, silicon is lossy at those frequencies because it's semiconductive. Engineers use a few tricks: they put the antenna on the top metal layer, often using a grounded coplanar waveguide design, and they add a dielectric layer like silicon dioxide underneath to isolate it from the lossy silicon. Some designs even etch away the silicon below the antenna to create a membrane, essentially an air bridge.
Luna: That sounds really delicate. How do they keep the chip from breaking?
Lucas: They use careful process control. It's a technique called 'backside cavity etching' — they remove silicon from the back of the wafer under the antenna. The antenna is then suspended on a thin dielectric membrane. It's fragile, but packaged properly, it works. The real payoff is integration: you can put the radio frequency front-end and the antenna on the same die, eliminating bond wires and transmission line losses.
Luna: That's huge for phased-array systems. In a phased array, you need many antenna elements — like 64 or 128 — each with its own phase shifter and amplifier. If you can put all that on a single chip, you shrink the entire system dramatically.
Lucas: Exactly. That's what a startup called MixComm did in 2025. They launched a commercial phased-array chip for 5G base stations that integrates 256 antenna elements directly on silicon. The chip measures about two centimeters on a side, and it can steer beams electronically without any external antennas. They're shipping to tier-one infrastructure providers as of this quarter.
Luna: Two-centimeter chip with 256 antennas? That's incredible. What kind of power does it consume?
Lucas: About five watts for the whole array, including the beamforming computation. That's competitive with traditional antenna-in-package solutions that use separate antenna modules. And because everything is on-chip, the assembly cost is lower. MixComm claims a 40 percent reduction in bill of materials for a 64-element subarray.
Luna: So we're moving from 'antenna-in-package' to 'antenna-on-chip'. That seems like a natural evolution. But what about interference? Having antennas right next to digital logic — doesn't the digital noise couple into the antenna?
Lucas: That's a big challenge. On-chip antennas are susceptible to substrate noise from the digital circuits. Designers isolate the antenna area with guard rings and deep trench isolation. They also put the analog and digital blocks on separate power domains. Some chips use a 'flip-chip' approach where the antenna is on the top metal layer facing outward, and the digital circuits are on the bottom side, so the silicon substrate acts as a shield. It's not perfect, but it works well enough for commercial products.
Luna: You mentioned this is for base stations, but what about handsets? Could we see on-chip antennas in smartphones?
Lucas: Possibly, but it's trickier. Smartphones need to cover multiple bands, and a dedicated antenna for millimeter-wave is already small. The real benefit is for devices with more space constraints, like tiny IoT sensors or wearables. For example, a smartwatch with on-chip antenna could have a smaller footprint and better battery life because the antenna is more efficient.
Luna: Interesting. I read about a research chip from imec last year that integrated a terahertz antenna for 6G. Are we talking about the same physics, but at even higher frequencies?
Lucas: Yeah, imec demonstrated a 300-gigahertz transmitter with an on-chip antenna. At that frequency, the wavelength is one millimeter, so the antenna is tiny — but the losses are even worse. They used a lens on the package to focus the beam. That's more of a research prototype, but it shows the roadmap. For 6G, which will use sub-terahertz bands, on-chip antennas might be the only practical way to build arrays.
Luna: What about automotive radar? The 77-gigahertz band is already used for adaptive cruise control. Are there on-chip antenna solutions there?
Lucas: Yes, several companies like Texas Instruments and NXP have radar chips with on-chip antennas, but they use a technique called 'antenna-in-package' — the antenna is on the package substrate, not the die. The reason is cost: packaging is cheaper than extra silicon area. But some startups are pushing on-chip antennas for 77 gigahertz as well. The key driver is that with on-chip antennas, you can put the radar transceiver and antenna in a single surface-mount component, simplifying assembly.
Luna: Speaking of assembly, if the antenna is on the chip, how do you test it? Antenna test usually requires an anechoic chamber. You can't just probe it on a wafer prober.
Lucas: You're right. Testing is a major bottleneck. For on-chip antennas, you typically do over-the-air testing in a shielded environment. Some companies have developed specialized probe stations with integrated antennas that can measure radiation patterns at wafer level. It's slower than standard parametric testing, so it adds cost. That's one reason why on-chip antennas haven't taken over overnight.
Luna: We've been talking about technology for a while. I want to touch on something — we deliberately don't run ads on these episodes. If you find value in understanding how these chips are built, and you want to support that choice, the link is buy me a coffee dot com slash fexingo. That really helps us keep the conversation focused on the engineering, not the sponsors.
Lucas: Luna, I appreciate you mentioning that. It's true — the ad-free model relies on listeners who see the value. And it means we can dive deep into topics like antenna design without anyone telling us to keep it light.
Luna: Exactly. So back to the tech: what's the next frontier for on-chip antennas? You mentioned 6G. Anything else?
Lucas: One exciting area is satellite communication. Low-earth-orbit satellite constellations need hundreds of beams, and phased-array chips with on-chip antennas could enable user terminals that are flat and low-power. Qualcomm and others are working on chips for the Ku and Ka bands. Another area is wireless power transfer — a startup called Ossia is developing on-chip antennas for far-field wireless charging at 5.8 gigahertz.
Luna: Wireless charging with on-chip antennas — that sounds like a perfect use case because you want the receiver to be as small as possible. If the chip itself is the antenna, you save space.
Lucas: Right. And the efficiency is acceptable for low-power sensors. The real challenge is regulatory: you need to stay within FCC limits for radiated power. But the chips are getting there.
Luna: So to summarize: on-chip antennas are commercially real today for millimeter-wave 5G, thanks to clever die design and process tricks. They're not a panacea — testing is hard, and the loss is still higher than a separate antenna — but for integrated phased arrays, they're the only way to get the density you need.
Lucas: That's a fair summary. I'd add that as frequencies go up, the case for on-chip antennas gets stronger. By the time we get to 6G at 100-plus gigahertz, I expect most high-performance arrays will be on-chip. The packaging cost savings and integration benefits are just too compelling.

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.