Beyond AI: the next wave of custom silicon
If you only read the headlines, the semiconductor industry in 2026 is a single story about AI. The numbers seem to agree. AI accelerators are driving unprecedented demand for compute, memory and advanced packaging.
Look beyond the datacenter, though, and a quieter market is taking shape. Analog, power and sensing remain fundamental to the physical world: these are the chips that convert, measure, protect and switch. They do not train models, but nothing that moves, charges or senses the physical world works without them.
That is where a large share of our ASIC work at Move Silicon happens, and it is where we see some of the most interesting design problems of the next five years. Three trends stand out.
1. Electrification is rewriting the power chain
The shift to wide bandgap devices is no longer a forecast. Silicon carbide and gallium nitride are moving from specialised applications into EVs, charging infrastructure, server power supplies, solar systems and other high-efficiency power applications.
The power switch gets the attention, but the harder engineering often sits next to it. A SiC MOSFET switching at high speed needs a gate driver that can survive the common-mode transients it creates, react quickly to fault conditions and maintain reliable operation over the lifetime of the system. GaN pushes the problem further, with tighter gate-voltage margins, higher switching frequencies and much less tolerance for parasitic inductance between driver and device.
This is where custom silicon can outperform catalogue parts. A driver designed around a specific power device can integrate protection thresholds, sensing, control and telemetry around the actual application rather than around a generic specification. Much of our work here runs on BCD processes, where high-voltage devices, precision analog and digital control can share one die.
The same logic applies to battery management. As batteries become central to electric mobility and energy storage, accurate cell measurement, protection and state-of-health estimation become increasingly important. The quality of that information starts with the silicon measuring the cell.
2. Sensing moves to the edge, and the power budget shrinks
The second trend is less visible but just as structural. Sensors are being placed where nobody wants to change a battery: on bridges, inside industrial machinery, on the skin and in increasingly small diagnostic systems.
The value of these nodes lies less in the transducer itself than in the silicon that reads it, conditions the signal and decides what is worth transmitting.
We see this first hand through our work in structural monitoring. A node that can operate for years on a small battery, or from harvested energy, can change the economics of monitoring far more than one extra bit of resolution. That pushes the design toward a very specific set of blocks: low-noise front ends, precision converters that can sleep between measurements, always-on wake-up circuits operating at extremely low power, and power management capable of starting from very limited harvested energy.
The same principle applies to biosensing. Ion-sensitive devices, electrochemical interfaces and impedance measurements require analog front ends that are often co-designed with the sensing element itself. These are small chips by area, but some of the most demanding to design because the error budget is ultimately defined by physics and chemistry rather than by a conventional digital specification.
For a customer, a custom ASIC in this space can be the difference between a laboratory demonstrator and a product that can be calibrated, qualified and manufactured at volume.
3. The car is being rewired
The software-defined vehicle is often described as a computing story, but one of its most concrete consequences is electrical.
Moving from many distributed ECUs toward zonal architectures changes where power is distributed, how loads are protected and what kind of silicon sits at the edge of the vehicle. Smart high-side switches, eFuses, 48 V conversion and zonal power management are becoming increasingly important parts of the architecture.
These functions also bring requirements that cannot simply be added after the fact. Protection thresholds, diagnostics, fault handling and functional safety have to be considered from the first schematic.
This is another area where custom ASICs can make sense. For OEMs and Tier 1s, a dedicated device can combine power management, sensing, protection and control around a specific system architecture, while reducing board area and dependence on a collection of catalogue components.
It can also become a sourcing decision. A mature, well-supported process can provide a stable foundation for a product whose lifetime may extend well beyond the technology cycle of the latest digital node.
What ties them together
Put these three trends side by side and a common thread appears.
None of them is won simply by shrinking transistors. They are won at the boundary between silicon and the physical world, where voltage, temperature, noise and time behave badly, and where the quality of the analog and power design directly affects the quality of the product.
That boundary is where Move Silicon has chosen to specialise.
Our teams work across analog, mixed-signal and power management design, including BCD technologies, together with packaging approaches that bring power devices, drivers and sensing elements close to the rest of the system. We also design for AI hardware, where power delivery and efficiency create their own set of challenges.
The underlying skills are remarkably transferable. The same principles that make an efficient voltage regulator for a compute die can make a gate driver survive a fault, or allow a sensor node to operate for years from a tiny energy source.
The physical side of the semiconductor industry
AI will continue to attract enormous attention and investment. At the same time, the physical side of electronics is changing rapidly, driven by electrification, sensing at the edge and the need to make increasingly complex systems more efficient and reliable.
For companies building products in these markets, custom silicon is not necessarily about replacing the biggest processor. Often, the value is in the silicon around it: the power management, sensing, protection, analog interfaces and control that make the overall system possible.
The question is therefore not simply whether a custom chip makes sense, but how early it should become part of the product architecture.
If you are working on a power, sensing or automotive product and considering whether an ASIC belongs in your roadmap, we would be glad to talk it through.
