Black Semiconductor: Shaping the Future of Optical Chip Interconnects
For decades, the number of transistors on a CPU has roughly doubled every two years. This is known as Moore’s law. Yet, as transistors became increasingly smaller – and thus more expensive and difficult to produce – the industry started seeking fundamentally new ways to increase computing power.
Since the early 2000s, increases in computing power have mostly come from parallelization by adding cores or switching from CPUs to GPUs for more specialized computing tasks. But sharing data between standalone processors still is tedious and costly.
The brothers, Daniel Schall (CEO) and Sebastian Schall (CFO), founded Black Semiconductor in 2020. Together with Cedric Huyghebaert (CTO), they developed optical interconnects to connect chips and allow them to exchange data quickly. Using graphene to translate electronic to photonic signals and transmit signals optically between chips, their technology may lead to 100x to 1000x faster data processing. Black Semiconductor has raised funding from Vsquared Ventures, Cambium Capital, Onsight Ventures, and Project A Ventures.
Learn more about the future of optical chip interconnects from our interview with the co-founder and CEO, Daniel Schall:
Why Did You Start Black Semiconductor?
Mainly out of curiosity and also because I happened to be in an environment where this opportunity opened up – and I am an opportunity taker!
When I was eight years old, I was disassembling a transistor radio when my classmate Patrick asked me: What are you doing? Do you even know what a transistor is? All the other elements on the circuit board had two pins, but the transistor stood out by having three pins. My curiosity was sparked, so I opened the transistor with a hammer and soon figured that it wouldn’t help me understand how a transistor works. So, at the time, I simply accepted that I couldn’t know what a transistor was or what it did.
However, this experience stuck with me, and years later, I decided to study electrical engineering and learn how all these electrical things work. Focusing on microelectronics, I finally learned how transistors work. This brought up more questions than answers because I then wanted to know how microchips were made. To this day, microchip fabrication is a miracle, creating all these tiny elements and ensuring they work 24/7.
As a student, I started making my own chips at the AMO GmbH, a research institute in Aachen. The research institute focused mostly on microchips whose performance might go beyond Moore’s law, where one important ingredient was photonics. But by pure coincidence, the institute also worked with graphene, so I thought about what cool things I could build with graphene. Maybe a transistor?
We tried building graphene transistors, but it wouldn’t really work as you couldn’t switch graphene on or off due to the lack of a band gap. We ended up with half-decent results, and that was the end of the story.
But our optical results turned out to be mindblowing, especially around converting photonic to electronic signals. So, during my Ph.D., I focused on integrated graphene with silicon photonics. After a few years, we measured our first device in collaboration with Bell Labs, an expert in high-frequency measurements and data communications. We showed that we could integrate graphene on single chips and ultimately also build wafer-scale devices using 6-inch silicon wafers.
After submitting my Ph.D. thesis, I found myself wondering about what to do with this new technology. I had previously run a recording studio and worked on IT infrastructure for the German army, so I knew what it meant to work outside of academia. Autonomy was very important to me, and having these devices from my Ph.D. was visible proof that we could build something amazing. My brother Sebastian and I applied for EXIST, a German funding program for university-based startups, and once granted, we decided to found Black Semiconductor in 2020.
We started analyzing what we knew and what we didn’t and brought Andreas Umbach, who had created high-speed optical communication devices before, on board as an advisor. Together, we analyzed all the possibilities and found that we could have the greatest impact in connecting neighboring chips.
The electronics industry had looked into this for many years, but the door had always been closed. There was no technology that allowed data transfer between chips easily and could be integrated with silicon. Graphene opened that door! It accepts the same material environment and integrates with silicon chips without ruining the electronics beneath, which makes it compatible with late-stage chip manufacturing.
Graphene is the key component that allows us to convert electronic to photonic signals easily, and photonics has the highest bandwidth to transmit data. This makes connecting chips directly possible, so ideally, they wouldn’t even notice that they’re being connected.
How Do Optical Chip Interconnects Work?
Processors create data in the electronic world that another processor might need to access. Photonics is the best for data transmission as it can provide the highest bandwidths, and not only for long distances. So you need a transducer that translates electronic data into the optical world and vice versa.
Usually, this signal conversion costs a lot of energy and thus makes data transmission less efficient and costly – it just hasn’t been worthwhile before for short distances like connecting chips. Yet, graphene is crystalline by nature, forming a honeycomb lattice of a single layer of carbon atoms, and because of this arrangement, it has very special physical properties.
Converting photonic to electronic signals works a bit like solar cells: A photon gets absorbed and lifts an electron from a lower to a high energy level, specifically from the valence band to the conduction band, where it can flow freely, and a current, that is, an electronic signal, gets created.
Semiconductors like silicon have a band gap separating the valence and conduction band, so a certain amount of energy is required to overcome this band gap, and only a limited range of wavelengths with greater energy than the band gap can be absorbed and initiate a current. Graphene is special in that it has no band gap and can absorb, in principle, any wavelength.
However, graphene is just a single layer of atoms, so most of the light would simply pass through without being absorbed unless it was integrated appropriately parallel to a waveguide. Thus, integrating the graphene on a silicon chip for maximum data transfer and low energy losses is the main engineering challenge, alongside manufacturing high-quality graphene since impurities induce band gaps and hamper the graphene’s ability to absorb light.
How Did You Evaluate Your Startup Idea?
First of all, we discussed a lot about where to enter the market and decided that long-distance data communications would not be the best idea: The technology to connect data centers over several to hundreds of kilometers has been fairly established.
Everything in the optical communications industry is fairly standardized, as multiple vendors need to plug into each other, and they have a decades-long roadmap of milestones for speed and energy efficiency of data transmission. If your technology can meet those milestones, you’re good; otherwise, you’re out, as you won’t be compatible with everyone else’s technology. So betting our startup on a single attempt to meet those specifications in five to seven years didn’t seem like a good idea – if we were off even by just a few percent, we could be out as competition is fierce.
However, one important job of data centers is to sort the incoming internet data stream. For example, when you watch a YouTube video, lots of data need to be distributed within a data center and also between different chips. And that’s where we saw an opportunity for us!
Our technology is fairly disruptive and could unlock a market that doesn’t exist yet for interconnecting chips! It addresses the fundamental problem of scaling to more transistors and obtaining more performance beyond Moore’s law. Even if chips can’t get significantly more transistors, barrier-free chip-to-chip communications will be an unlock to involve more chips and, thus, more transistors in a computation.
If we’re able to meet the specifications, we might also get into data communications later on. But interconnecting chips is what is happening right now. It’s a good strategy to address a problem with strong demand that existing technologies cannot fulfill – and if you solve that problem, you can open a new door. And from there on, you can improve.
What Advice Would You Give Fellow Deep Tech Founders?
Only do it when you love it! If you consider founding a startup just for the money, forget about it, and do something else. It’s sometimes extremely stressful, and you might need to get away with only four to five hours of sleep and work 110 hours a week. If you’re horrified, I can tell you it’s like sports: you train and get used to it. Once your startup gets off the ground, you might change back to more normal hours, but for a certain time, it will be necessary to work your butt off.
You will only sustain this pressure if it’s your desire to found a company. The reward shouldn’t be the money but simply the experience of what you’re doing. I feel blessed to have the opportunity to work on Black Semiconductor; it’s my reward. I am an opportunity taker, and when I saw this opportunity, I knew it would be the one to go for. Expose yourself to many opportunities and share your ideas – don’t be afraid somebody might steal them.
Last but not least, I’d like to say that we’re very gifted in Europe to have a set of values, e.g., around health care or working conditions, that we fought for, that are outstanding and precious, and that we need to preserve. Yet, to maintain and defend these values, we need to be at the forefront of technology. New technologies move society forward, lead to increases in the quality of life, and support our European sovereignty.
Over the past few years, it has become increasingly clear that there are lots of new opportunities in deep tech and commercializing fundamentally new technologies. Yet, deep tech often also needs a lot of investment! 500M is not a lot for a semiconductor startup. It may be enough to design a CMOS node, but developing fundamentally new semiconductor manufacturing technology requires an order of magnitude more.
And it needs not only a lot of money but also a long breath. Deep tech doesn’t create unicorns within two years like B2B SAAS startups, which allow venture capitalists to cash out early. Instead, deep tech has the potential to create truly big and defensible companies, and this is the opportunity for venture capitalists to adapt and explore these new opportunities.
You can still make lots of money, but you need a different mindset. US-venture capitalists have realized this and, currently, are the ones doing the big deals! Why do we have only so few large funds with a technology focus? Funds that understand technology and are able to make long-term and large investments? If we care about preserving our European values, we should care about being at the forefront of technology.
Further Reading
Photonics: The computer chip industry’s new alternative – This article from Daniel Schall shows how photonics may change the chip industry and how Europe could be once again at the forefront of the semiconductor industry
Intel, Hermann Hauser, Google investor back Black Semiconductor – News report about Black Semiconductor joining the Intel Ignite startup program and receiving support from Hermann Hauser and Andy Bechtolsheim
Black Semiconductor: Graphene-based superchips – Portfolio overview by Vsquared Ventures
Graphene makes light work of optical signals – Nature article about early graphene photodetectors in 2013.