Why Hardware Is the Hard Part of 6G

Most public conversation about 6G is about software and spectrum: which bands, which standards, which applications. That conversation matters. But the systems people picture when they say 6G, networks that sense their surroundings, locate objects to the centimetre, connect a phone straight to a satellite and coordinate fleets of drones, all rest on a layer that gets far less attention. They rest on hardware that has to work at frequencies where the ordinary rules of circuit design quietly stop applying.

5G was, at heart, a speed project. Wider channels, more antennas, higher-order modulation. The hardware was hard, especially at millimeter-wave, but the goal was familiar: move bits faster.

6G asks for something different. It wants the radio to double as a sensor. It wants the same array that carries data to image a room, track a hand gesture, or spot a pedestrian behind a truck. It wants links that survive at sub-terahertz frequencies where the atmosphere itself is hostile. And it wants all of that in devices cheap, light and cool enough to hold, fly, or launch.

Every one of those requirements pushes the frequency up, and every step up makes the physical design harder in a way that does not show up in a link budget until it is too late to fix cheaply.

Here is what that looks like in practice. At a few gigahertz an engineer can treat the chip, the board, the connector and the antenna as separate problems with tidy interfaces between them. At 100 GHz the interfaces are the problem. A transition a millimetre too long becomes a resonator. A solder bump becomes an inductor. The package that was meant to protect the chip becomes an accidental antenna, sending power in directions nobody planned.

The failure mode is strange to watch. A system can have a superb chip and a superb antenna and still perform badly, because the piece between them, the part nobody puts on a slide, is leaking a third of the power. One bad connection makes a futuristic 6G radio behave like airport Wi-Fi.

This is why RF packaging has stopped being a mechanical afterthought and turned into an electrical discipline. In a 6G front end the package is part of the circuit, and it has to be designed with the same care as the transistor.

Much of my own work comes down to one fact: loss grows with frequency, and most of it happens in the dielectric a signal travels through. A conventional circuit board guides signals through plastic laminates. At 2 GHz that is fine. At 60 GHz and above the plastic eats a meaningful share of the signal for every millimetre it travels.

Air is the best dielectric there is. Air-filled substrate integrated waveguide, AFSIW, is a family of structures that guide the signal through a channel of air cut inside an ordinary multilayer board. You keep the manufacturing you already know, the planar shape, the low weight, and you replace the lossy plastic with nothing at all. Loss drops, power handling rises, and the result is light enough to fly on a drone or ride on a satellite.

It sounds almost too simple, and the catch is in the transitions: getting a signal from a chip, a connector or a standard waveguide into that air channel and out again without reflecting or radiating it away. That transition problem took the first several years of my career and produced two patents, and it is a fair picture of 6G hardware in general. The concept is elegant. The interface is where the engineering lives.

Add one more requirement and the difficulty compounds. Many of the platforms 6G is meant to serve are not flat. Aircraft skins, vehicle bodies, wearables and deployable satellite panels all want antennas that follow a curved or flexible surface.

At low frequencies a little flex is harmless. At millimeter-wave a bend of a few degrees changes the electrical length of every element in the array. The beam moves. The phases drift. Polarization purity degrades. The match shifts. And all of it depends on how the surface happens to be bent at that moment.

A flexible array is therefore not one design problem but four at once: an antenna, a mechanical structure, a material system, and a model of how the three interact under deformation. Rigid arrays were hard enough. Flexible ones demand that the electromagnetic and mechanical design be done together, by the same people, from the first sketch.

Then there is power, which is where physics sends its largest invoice. A power amplifier at 100 GHz converts a far smaller fraction of its supply into radio energy than one at 2 GHz does, and the rest becomes heat. An array multiplies that heat by the number of elements and packs it into a footprint set by the wavelength, which at these frequencies is a few millimetres. The result is a thermal density that would be respectable in a processor, sitting behind an antenna that is supposed to be mounted on a lamp post or the back of a phone. Managing that heat without detuning the antenna it sits under is a design problem in its own right, and it is one reason the useful frontier of RF lags the laboratory frontier by a decade.

Measurement compounds all of it. Above 100 GHz there is no connector you can trust, so the classic method of plugging a device into an instrument and reading a curve starts to fail. Devices have to be characterised over the air, in anechoic chambers, with probes whose own behaviour has to be calibrated out. The instruments are scarce and expensive, the calibration standards are a research topic themselves, and a measurement that disagrees with a simulation can take weeks to adjudicate. Companies planning sub-terahertz products discover that the test bench is a larger share of the budget than the product.

Cost closes the loop. A base station antenna with a few hundred elements is affordable when each element costs cents. At millimeter-wave the element count rises, the packaging gets harder, the materials get more exotic, and the supply of foundries that can build the highest-performing devices, in indium phosphide or gallium nitride, is a short list. The consumer price that 6G assumes, phones and terminals that ordinary people buy, is not a given. It has to be engineered into the hardware from the beginning, in the same way loss and heat have to be.

Sensing raises the bar again, and it is the requirement that separates 6G from a faster 5G. A radio that carries data can tolerate a little phase noise and a little drift; the receiver corrects for them and moves on. A radio that must also measure the distance to a wall to the centimetre, or detect the breathing of a person in the next room, cannot. It needs oscillators that stay coherent over the time of a measurement, arrays whose elements are calibrated against each other to a fraction of a degree, and front ends whose own reflections do not swamp the faint echo they are listening for. None of that is visible in a data-rate specification, and all of it is hardware. The sensing story sells 6G to the public. The oscillator decides whether the story is true.

What follows for the organisations building these systems is a change in who sits at the table. In 5G the antenna team, the RFIC team, the packaging team and the thermal team could work in sequence and hand each other specifications. At 100 GHz and above the specifications interact too strongly for that. The antenna is the package, the package is the heat sink, the heat sink detunes the antenna. The teams that ship working 6G hardware will be the ones that put those people in one room from the first sketch, with one model that all of them trust, and that treat the joints between disciplines as the design rather than the paperwork.

None of this is a reason to doubt 6G. It is a reason to look at where the effort has to go. The standards will be written, the spectrum allocated, the software brilliant. Whether the systems work in the field will be decided by transitions, interconnects, packages and materials, the parts of the design that are hardest to see and easiest to underfund.

The industry learned during 5G that millimeter-wave is unforgiving. 6G will run higher still. The organizations that treat hardware integration as a first-class problem, rather than a detail for the end of the schedule, will be the ones whose 6G systems perform the way the slides promised.