One of the hottest topics in the tech industry right now is machine learning. Analysts believe AI-enabled devices will grow from 300 million to 3.2 billion by 2022.1 Every corner of the industry is researching or already developing products that can think for themselves. From autonomous cars to smart-home devices and automated robots, the possibilities for machine-learning devices are endless.
Given the size of this market and its projected growth, developers, innovators, and startups need to create their products as efficiently as they can to get to market quickly and to attract investors. Innovators must take their ideas further than just a piece of paper, creating proof of concepts that give investors a tangible idea. This has resulted in companies exploring custom system-on-chip (SoC) or application-specific processors (ASICs) for their designs.
Designing an SoC: Fact vs. Fiction
Several “myths” surround SoC design. First among these is that it’s prohibitively expensive, the kind of project that can only be undertaken by major companies with vast resources. For many devices, that just isn’t true. By leveraging existing, proven IP, the costs of developing an SoC can be brought down considerably. This is the electronic equivalent of not reinventing the wheel—instead, buy a ready-made wheel from an expert wheelwright and skip that step completely.
Another myth is that a company needs a large and highly skilled team to crack the SoC design process. You can go that route if you want, but the industry is now well-served by a large population of design houses that can handle all or part of the process for you. For example, Arm has established the Arm Approved Design Partner program to partner with a range of audited design services companies who are all well-versed in designing with standard IP, from Arm and elsewhere.
Finally, many startups may think developing a custom SoC can be risky, but thousands of companies are considering this path. There’s a whole ecosystem supporting companies that are building custom SoCs with documentation, tutorials, design reviews, and well-developed quality processors.
Turning a Vision into $21 Million
Hailo, a startup based in Israel, had the vision to build machine-learning capability into computing platforms used in cars, drones, smart-home appliances, and many other edge devices. The company wants to equip devices to be able to think for themselves with the performance of a data-center-class computer, operating in real-time and at reduced power, consumption, size, and cost.
By driving intelligence to the edge, Hailo could disrupt many industries for the better, from improved intelligence for a home security camera to significant enhancements for predictive maintenance in industrial applications.
For Hailo to achieve this vision, the company needed to completely rethink computer architecture. As well as its revolutionary machine-learning architecture, Hailo’s design needed a general-purpose processor, too. With Hailo’s laser-focus on costs, it was vital to be able to get the right processor at the right price. Therefore, the team could build a working prototype and demonstrate that their idea was viable to attract the attention of the investor community.
Hailo chose the Arm Cortex-M3 processor for its design and accessed the IP directly from Arm through DesignStart, with no upfront fee, allowing the company to concentrate on the truly innovative parts of its machine-learning architecture. IP companies, including Arm, appreciate the challenges faced by startups when bringing their designs to life. That’s why access programs like DesignStart have been created to help break the circle and unlock the potential of innovative ideas.
“We chose the Arm Cortex-M3 for a number of reasons—it is an incredibly capable processor in its class, with low power requirements,” says Orr Danon, CEO at Hailo. “By properly allocating what is running where and joining forces between it and our own IP, we managed to create a highly capable yet also extremely power-efficient prototype.”
And it’s not just about the hardware for Danon. He continued to say, “Creating robust hardware requires an equally robust software ecosystem with a tried-and-tested software toolchain. We’re not interested in forcing the industry to use non-standard toolsets; our toolchain combines the standard Arm tools seamlessly with our own software development kit. Arm’s extensive software ecosystem, combined with the variety of support and resources that exist, made it the obvious choice for us to ensure successful adoption in the market.”
Armed with a working prototype, Hailo looked for further funding and successfully raised $16 million from a combination of crowdfunding and angel investors. That would have been much more difficult, maybe even impossible, without being able to demonstrate that its ideas were real.
Earlier this year, Hailo increased its fundraising to $21 million with significant backing from a venture capital firm. The firm expects to have devices available for sampling very soon.
Unlocking the Potential for Innovative Ideas
For innovation to continue and for a world of a trillion connected devices to become a reality, the industry needs to make it easier and more affordable for startups, as well as existing players, to get into new markets.
More startups and innovators are pioneering new methods and ideas that have never been used before. And the best way to enhance this innovation is for IP companies to provide more opportunities and programs for smaller businesses to access the IP, tools, and software they need to innovate.
From a vision on paper to $21 million funding, Hailo was able to quickly access the IP it needed to secure investment and begin its mission to empower intelligent devices at the edge.
Chris Shore is Director, Embedded Solutions, Automotive and IoT Line of Business at Arm.
1. Source: IDC WW Embedded and Intelligent Systems Forecast, 2017-2022 and Arm forecast.