What if the bandwidth memory wall
didn't exist?
Microcomb is how AI infrastructure breaks the memory wall.
AI's greatest constraint is not intelligence — it is the physics of moving data. Microcomb is building the foundational photonic layer that dissolves the bandwidth memory wall, replacing hundreds of individual lasers with a single microcomb.
Using a single soliton crystal microcomb chip as the light source, we are demonstrating coherent data transmission at 25.6 Tb/s — against a commercial baseline of 0.8 Tb/s today. Same physics. Radically different architecture.
It is the proof-of-concept that establishes a clear and foundational IP position — the layer the industry needs to meet Nvidia's own specification for next-generation AI infrastructure.
The AI compute buildout is the largest infrastructure investment in history — a $7 trillion race through 2030. But it is running into a hard physical constraint. Today's optical interconnects were built for a different era: every additional wavelength channel requires its own dedicated laser. At the bandwidth levels AI demands, this creates unsustainable cost, complexity, and power consumption.
Industry calls this: the bandwidth memory wall. Traditional optical solutions cannot break through it. A fundamentally different source of light is required — one that generates hundreds of coherent channels from a single chip.
Bandwidth >10 Tb/s · Latency <500 ns rack-local · Energy <1 pJ/bit
| Metric | Today's Baseline | Microcomb Target |
|---|---|---|
| Bandwidth | 0.8 Tb/s Commercial peak, 800G |
25.6 Tb/s32× |
| Energy / bit | 5–20+ pJ/bit And getting worse at scale |
≤1–5 pJ/bitup to 20× |
| Latency | ~300–400 ns Short link, LPO |
≤500 nsrack-local |
| Light source | 1 laser per channel Multiplied across every link |
1 chip All channels |
A microcomb generates a precise frequency comb — hundreds of equally-spaced, mutually coherent wavelength channels — from a single integrated chip. No array of individual lasers. High-bandwidth coherent communications from a single chip source can only be achieved this way.
We are targeting 25.6 Tb/s using coherent optical transmission — the modulation architecture already standard between data centres, now moving within them. We are building for when the market arrives.
Fewer lasers means dramatically less power. Today's architectures burn 5–20+ pJ/bit and scale worse as bandwidth increases. Our microcomb architecture is designed to meet the <1 pJ/bit threshold Nvidia has specified as the requirement for next-generation AI infrastructure. Bandwidth and energy solved simultaneously — in one chip.
Like ARM for silicon, Microcomb owns the foundational layer. Our soliton crystal microcombs are best-in-class on reliability, stability, robustness and efficiency. The industry will build on top of this IP to 32× the interconnect bottleneck — and we hold the foundation.
A photonic interconnect demonstration using our soliton crystal microcomb as the light source — through a coherent WDM link and optical transceiver chain — instrumented against today's commercial state of the art on throughput, latency, OSNR, and energy per bit. No new research required. The science is proven. This is execution.
A decade of published research, hundreds of peer-reviewed papers, and a national centre of excellence — now entering its commercial phase.
Our soliton crystal platform delivers best-in-class performance on every dimension that matters commercially — reliability, operational stability, robustness to environmental variation, and conversion efficiency. In Professor Moss's own words: the best in the world.
Beyond the microcomb source, we hold intellectual property on our modulation approach — a second defensible layer. The combination creates a technology moat that cannot be readily replicated, even by well-equipped competitors.
This is not a laboratory curiosity. The demonstration targets the coherent interconnect requirements of next-generation AI data centres — benchmarked against current commercial state of the art on throughput, latency, OSNR, and energy per bit.
Microcomb is co-founded by the scientists who have spent a decade producing the world's leading body of work on microcomb technology — and a commercialisation team translating that science into infrastructure.
Distinguished Professor and Director of the Optical Sciences Centre at Swinburne University of Technology. Deputy Director of the ARC Centre of Excellence for Optical Microcombs. One of the world's foremost authorities on microcomb technology, with a publication record spanning hundreds of peer-reviewed papers and an h-index of 143. His research group has defined the global state of the art in soliton crystal microcombs.
Associate Professor at Monash University and founder of the Monash Photonic Communications Laboratory. ARC Future Fellow (2023–2027) and co-lead of the Information and Intelligence Theme at the ARC Centre of Excellence COMBS. His research focuses on using microcombs to support tens of terabits-per-second in optical fibre links. Lead author on the landmark 2020 Nature Communications paper demonstrating 44.2 Tb/s data transmission over 75km of standard optical fibre using a single soliton crystal microcomb — a world record at the time. PhD, University of Sydney; post-doctoral research at Chalmers University of Technology, Sweden.
Professor at RMIT University's School of Engineering and a leading authority in integrated photonics and silicon photonic platforms. Co-author on Microcomb's foundational Nature Photonics review paper alongside David Moss and Bill Corcoran. Professor Mitchell brings deep expertise in photonic integration and industry engagement, including connections to major optical networking and data centre infrastructure players.
Commercialisation lead for Microcomb, engaged through Swinburne's Founder-in-Residence programme. Stanford GSB Innovation Leadership (2025). MBA, Melbourne Business School. Swinburne alumnus.
One of Australia's leading research universities and home of the Optical Sciences Centre. Microcomb is being commercialised through Swinburne Innovation & Enterprise's Founder-in-Residence programme, based at the Swinburne Innovation Studio in Hawthorn, Melbourne.
A multi-year Australian Research Council Centre of Excellence (2023–2030). One of the world's most significant dedicated investments in microcomb science — the national research programme underpinning Microcomb's technology, spanning Swinburne, Monash, and RMIT.
Conditionally approved funding through the Breakthrough Victoria University Innovation Platform — part of a broader Swinburne–BV investment partnership supporting deep technology commercialisation from Victorian universities.
If you're working on next-generation optical interconnect, AI data centre infrastructure, or deep tech investment — let's meet at OFC. The window to establish the foundational position in this space is open.
We are speaking with hyperscalers, transceiver manufacturers, optical component companies, and deep tech investors. The benchtop demonstration is underway. The foundational IP position is being established now.