Open Source Silicon with Spherical’s Thomas Parry

What was your journey to starting the company?

I used to work for another company designing similar systems, and every day we struggled with the same issue. How to choose microchips and arrange them on a circuit board. It became clear to me that the biggest limitation in designing these systems was the microchips themselves. They were the core building blocks, and they were always the bottleneck.

At some point, I started asking a simple question: if this is the biggest leverage point, why don’t we just design our own microchips? My managers and colleagues at the time dismissed the idea, saying it was impossible. And back then, they had good reasons to be cautious since custom silicon design was prohibitively expensive and complex.

But things have changed. The rise of open-source silicon design has dramatically lowered the barriers to creating custom microchips. When I realized that the technology had evolved to make this possible, I knew it was time to go back and fix that fundamental problem. Even now, I have this gut feeling that there is a better way to do it, and that drive is what led me to start Spherical.

Tell us more about open silicon design and its development. 

The semiconductor and microchip industry has traditionally been extremely secretive. To access any information about a semiconductor fabrication facility, you need to sign strict NDAs. In many cases, even getting to that step is difficult since you have to prove your business case to analysts at companies like TSMC before they will even consider giving you access to their process design kit (PDK), which contains critical details about their manufacturing process.

On top of that, the software tools needed to design microchips are prohibitively expensive, starting at around €20,000 per person per year. This made it nearly impossible for startups, let alone hobbyists, to enter the space.

Then, during COVID, something incredible happened. Google funded an open-source process development kit and made it publicly available on GitHub. They also sponsored the manufacturing of open-source chips, allowing people to fabricate microchips for free. I was involved in the first two open-source chips ever created. This initiative acted as a catalyst, encouraging developers to build more open-source tools and software.

In the span of a year, we went from having no open-source silicon ecosystem to having thousands of people experimenting with chip design in their free time. This shift was a key enabler for Spherical. While we operate professionally at a higher level than hobbyists, this movement drastically lowered the barriers to entry, allowing us to develop custom microchips tailored specifically for the space industry.

Open design tools lower the barrier to chip development, but you still need access to a fabrication facility. How do you handle that part of the process? 

That is definitely a challenge. We do not use TSMC because it is difficult to gain access, and for the space industry, it is actually more beneficial for us to work with European fabs. Sovereignty is a key consideration, so we manufacture our chips at a German fab.

There is also an organization called IMEC in Belgium that conducts semiconductor research and runs a program called Europractice. This program acts as a bridge, helping startups and academic projects gain access to large fabrication facilities. They smooth out the process and make it easier for smaller companies to manufacture their chips. That is how we produced our first chip. Since then, we have developed a more direct relationship with our fab, but Europractice was invaluable in getting us started.

Who are your customers? 

Our first customer is a Spanish company called Orbital Paradigm. They are building a capsule that launches into space, conducts experiments in zero gravity, and then returns to Earth so customers can retrieve their samples. This is useful for biology and materials science research, where samples need to be studied after exposure to microgravity.

We sold them a power system, and the key value for them was the flexibility of our technology. Our power systems offer significantly more software configurability than traditional ones, which allows them to adapt more easily to their customers’ needs.

When you talk about a power system for satellites, what does that typically involve?

A satellite’s power system is essential for its operation because, in most cases, once it is launched into orbit, no human will ever touch it again. The system needs to collect energy, store it, and distribute it efficiently.

Satellites typically rely on solar panels to generate power. The power system regulates the energy coming from those panels and stores it in a battery so the satellite can continue operating when it is in the Earth’s shadow. That stored energy then needs to be distributed to all of the satellite’s subsystems—its computer, communications system, sensors, cameras, and anything else on board. Our power system acts as the central hub that manages and directs this power flow.

What makes our approach unique is flexibility. Most satellites require a customized power system because they have different configurations, which means traditional power systems need extensive hardware modifications to adjust things like voltage and current levels. Since we design our own microchips, we can build much of that control into software rather than relying on hardware modifications. This dramatically simplifies manufacturing and makes our power systems far more adaptable for customers.

What are the challenges involved with custom chips? 

That is a great question. While we talk a lot about our custom chips because they are our key advantage, we do not actually sell the chip itself. What we provide is a complete power management system, which is a metal box that handles energy regulation and distribution.

Our customers want flexibility in their power systems, but when they go to existing suppliers and ask for software-controlled customization, they hit a wall. I saw this firsthand in my previous job when customers would ask how to achieve this flexibility. With conventional chips, which are essentially cobbled together in a “Frankenstein” fashion, it is not possible to get the level of control they need in a reliable way.

One major issue is radiation. In space, radiation can easily corrupt memory, which is especially dangerous when dealing with software-controlled power systems. If configurations change unexpectedly due to radiation interference, it can be catastrophic. To prevent this, traditional systems require additional circuits to detect and correct errors, which makes them significantly more complex and bulky.

Since we design our own microchips, we can manipulate individual transistors at the chip level to handle these challenges in a much more efficient way. This allows us to provide advanced software flexibility in our power systems without the downsides of existing solutions. Our competitors cannot match this capability because they are working with off-the-shelf chips that were never designed for these specific needs.

When a customer comes to you with requirements for their satellite, do you design a fully custom power system each time? 

Our approach is highly modular, which allows us to assemble different modules to create a system tailored to each customer’s needs. Instead of requiring hardware modifications, we handle most of the customization through software configuration, making our systems far more flexible.

In terms of what drives purchasing decisions, performance is the top priority. Cost is a factor in a few cases, such as student-led projects, but most customers care more about reliability. Power systems need to work flawlessly in space, and our customers are willing to pay for that level of dependability.

You started with satellite power systems, but do you see potential uses beyond space, such as defense or industrial settings?

We are already seeing demand for motor control systems, which share many similarities with power electronics. These systems involve controlling voltages and currents to operate components like propulsion systems and actuators. We have a few projects in development in this area, and over time, we plan to integrate more computational intelligence into our systems, allowing for smarter decision-making in spacecraft operations.

Beyond aerospace, we are also looking at ruggedized applications in other industries. We have already been approached by companies in nuclear energy, oil & gas, and defense, all of which see potential for our technology in extreme environments. Our focus right now is on space, but we are actively exploring how best to expand into these other markets.