Big deal: Lightmatter and Lightelligence signal rivalry

Interesting stories emerge out of the university venturing world on a regular basis – be it the groundbreaking ability to edit genes, solar panels that boost efficiency by a third or a miraculous material that stops drugs from crystallising and makes them soluble.

Sometimes, the interesting story even includes a long-running legal battle – when it comes to the aforementioned gene-editing technology, known as Crispr-Cas9, for example, University of California Berkeley and Broad Institute have been fighting a drawn-out dispute through the US court system. That is on top of Broad Institute struggling to hold on to its intellectual property in Europe, where the European Patent Office revoked one of the patents a month ago.

But the story of Lightmatter and Lightelligence is in many ways a stranger one. Both are spinouts from Massachusetts Institute of Technology (MIT) and both secured decent funding rounds earlier this month – Lightmatter collected $11m in a series A round and Lightelligence obtained $10m seed capital.

It looks like a solid start for two spinouts out of a venerable institution, but in fact it has been a bumpy start for Lightmatter and Lightelligence, which are exploiting the same basic research. The two companies are developing computer chips that use light – or photonic – signals instead of electric signals, based on research that the respective co-founders initially conducted together.

Yichen Shen, co-founder of Lightelligence, and Nicholas Harris, co-founder of Lightmatter, co-authored a paper on the technology in Nature Photonics in June last year along with a group of nine other researchers.

Shen and Harris collaborated through the initial stages of company development, winning MIT’s 100K startup competition and taking home $75,000 in Harvard University’s President’s Innovation Challenge awards. Both entries were made under the Lightmatter banner.

What happened next is not entirely clear, though reports have emerged that Shen and Harris split over different approaches to company strategy and culture.

Both spinouts have impressive backers. Lightelligence reportedly attracted Baidu Ventures, the corporate venturing arm of Chinese internet company Baidu, as a lead investor, while Lightmatter’s round was co-led by Matrix Partners and Spark Capital.

The two rivals have not yet specified where exactly they want to take the technology, or which part of the research they are exploiting. They are not using the same patents, it appears, as Lightelligence noted in its press release that it has “exclusively” licensed research from MIT’s tech transfer office.

The technology is potentially revolutionary, something that the tale of the two spinouts distracts from all too easily. Semiconductors based on electrical signals are slowly inching towards their natural end. For much of computer architecture development over the past several decades, the number of transistors in a dense integrated circuit has doubled on a roughly regular timescale. Initially, this was every year – a principle known as Moore’s law, based on an observation made by Gordon Moore, co-founder of chipmaker Intel, in a research paper in 1965.

The rate has slowed down over the years – in 1975, Moore updated his prediction to every two years, and over the past decade it has effectively moved to three years.

The problem that companies like Intel face is simple in nature but arguably impossible to crack – more transistors mean faster speeds, but the more transistors designers want to cram on to a circuit, the smaller they need to be. Intel, for one, is currently working on 10-nanometre (nm) chips, while competitor Samsung said last October it had cracked the manufacturing process for 8nm chips.

These are unimaginably small – a human hair is 75,000nm wide and even a red blood cell is still a whopping 6,000nm to 8,000nm. A transistor that is 10nm wide consists of only 20 silicon atoms lined up next to each other.

So semiconductor producers are eventually going to hit 1nm transistors – what will happen beyond that is a big unknown but is likely to mean the end of current approaches to computer processors. It is one of the reasons researchers have been working on quantum computers.

Lightmatter and Lightelligence rely on light signals to power their chips – potentially boasting orders of magnitude in processing power over current chips and thereby being particularly relevant for data-intensive artificial intelligence (AI) applications that already exist.

Harris said: “For decades, electronic computers have been at the foundation of the computational progress that has ultimately enabled the AI revolution, but AI algorithms have a voracious appetite for computational power.

“AI is really in its infancy, and to move forward, new enabling technologies are required. At Lightmatter, we are augmenting electronic computers with photonics to power a fundamentally new kind of computer that is efficient enough to propel the next generation of AI.”

Lightmatter and Lightelligence may manage to co-exist or one may come out on top. Splitting the expertise into the technology between two companies is a gamble, of course, but even if only one spinout should survive it could prove to be a great leap forward for AI and computer technology in general.

This tale is already intriguing, but it should be particularly interesting to follow over the next few years.