For a company that has been in “stealth mode”, in the words of founder Amin Shokrollahi, Kandou Bus is grabbing more than its share of headlines. As well as winning this year’s ACES ICT award, the company made news last year for producing the 1000th invention to come out of the École Polytechnique Fédérale de Lausanne (EPFL), followed by the announcement in November 2012 that it had raised $10 million from unnamed investors to fund its growth. That’s a lot of good news for a start-up launched just two years ago.
Kandou’s technology makes it possible to transport more data on existing electronic infrastructure, ranging from long cables as encountered in DSL lines, to the tiny silicon vias that connect memory chips in high-end servers– a breakthrough with big commercial potential. The company’s patented signal processing techniques reduce the energy consumption of electronic devices by 25 to 30 per cent while increasing bit rates to peripherals such as external hard drives, printers, monitors by nearly 400 per cent, says Shokrollahi, a professor in the School of Informatics and Computer Science at EPFL.
Traditionally, chips send signals across pairs of connections, but Kandou’s approach involves sending signals across multiple wires. And as increased speed drives demand for new electronic devices, any technology that promises to support faster chips has big market potential.
The company’s name Kandou comes from the Farsi word beehive. “Like bees that work together towards a common goal,” Shokrollahi explains. “The communication wires in a Kandou signalling scheme work to compensate for errors and interference during high-speed data transmission.” In simple terms, more bits can be sent per unit of energy, or less energy can be used to achieve a given bit rate.
The opportunities for saving energy and speeding up processes lie in the connections between chips in devices. This interconnect – the “bus” that gave the company part of its name – is a “huge bottleneck”. Kandou’s solution focuses on algorithms, complex mathematical formulae that can be turned into the ‘brain power’ of silicon chips.
In some electronic devices the traditional bus already consumes around a quarter of the energy. That consumption rises rapidly as you increase the throughput or traffic on the bus. Before long the interconnects can use more than half of the power needed to run a device. With Kandou’s algorithms, energy consumption drops down to around 10 per cent.
An algorithm is born
During an academic career focused on researching error-correcting codes, Shokrollahi has amassed more than 100 papers and books and over 50 patents. He and one of his postdocs, Harm Cronie, stumbled on the technological breakthrough behind Kandou while working in the Algorithmics Laboratory in EPFL School of Computer and Communication Sciences. They were studying the digital signal lines (DSL) that many people use to access the internet over standard telephone lines.
“We realized that using a new mathematical transform, paired with novel algorithmic ideas, it is possible to transport more data on existing DSL lines,” says Shokrollahi. “One thing led to another, and we realized that the method we had found had much broader applicability than just DSL lines.”
That explains why Shokrollahi chose to work in stealth mode while patenting the breakthroughs and launching a company. “We very quickly realised that this is not something we wanted to primarily write papers on,” he said. As he adds, even potential customers can start patenting around your technology.
The Swiss Venture Kick scheme helped the inventors to develop a business plan and provided practical support and seed funding. Venture Kick then made a follow-on investment of 130,000 Swiss francs in May 2011 to support commercialisation. Shokrollahi became CEO of Kandou with Cronie the company’s chief technology officer.
Energy efficient chips
While Kandou’s goal is to slash energy consumption in electronic devices, it does not intend to make chips and devices. “We are a licensing company,” says Shokrollahi. “We design something called a ‘serial link’ for our customers. During the design process, we collect engineering fees. If the customer decides to use our designs in their products, then we collect licensing fees and royalties [for] the right to use our technology.”
There is already a well established, and profitable role model for so called ‘fabless IP companies’. The semiconductor design company ARM designs chips that end up in most of the world’s mobile telephones.
If Shokrollahi has his way, Kandou will be the first of a series of companies to exploit the heavily protected technology, with around 20 patents already filed and more in the pipeline. “The potential market for our technology is definitely over $100 billion,” he adds.
Shokrollahi already has shown a knack for bringing scientific innovations to market. In 2009, he sold his first business venture, Digital Fountain, which specialized in transmission of data on unreliable networks, to Qualcomm.
Just about anyone who is anyone in the electronics business is a potential customer of Kandou’s technology. But for now Shokrollahi won’t name names, partly because companies aren't keen to have other businesses know what they are up to.
Of course, when dealing with industry giants, small companies need to keep their wits about them. Kandou already has a couple of industry heavyweights offering guidance. One board member, Steve Papa, was the founder and CEO of Endeca Technologies Inc., which Oracle bought for $1.1 billion in 2011. And Brian Holden, the company’s VP of product management, was a co-founder of Stratacom which was sold to Cisco for $4 billion in the late 1990’s. John Fox, the company’s Senior Development Manager, is another semiconductor industry veteran.