What do startups AIStorm, Esperanto Technologies, Imagination Technologies, Xnor, and Flex Logix have in common? All five are developing chips custom for AI workloads, and they’re far from the only ones. The machine learning chip sector was valued at $6.6 billion in 2018, according to Allied Market Research, and is projected to reach $91.1 billion by 2025.
Yet another player vying for market share is Hailo Technologies, a secretive Israeli company that has raised $21 million in venture capital to date. It had previously declined to disclose the details of its edge and high-resolution sensory processing technology, but today it spilled the beans.
Hailo says its smaller-than-a-penny Hailo-8, which the company is now sampling with “select partners” across multiple industries, features an architecture that enables connected devices to run algorithms that would previously have required a datacenter’s worth of computing power. Hailo-8 is capable of 26 tera operations per second (TOPs) — more than double that of chips like Gyrfalcon’s Lightspeeur 2801, which maxes out at 9.3 TOPs, and CEVA’s NeuPro, which can reach 12.5 TOPs. The company says it does this while consuming significantly less power than some competing chips and that in addition to compute it incorporates both memory and control.
Hailo furthermore claims that — thanks in part to a powerful software development kit (SDK) and a novel heat-dissipating design that obviates the need for active cooling — the Hailo-8 outperforms Nvidia’s Xavier AGX on several AI semantic segmentation and object detection benchmarks, including ResNet-50. In one preliminary test at an image resolution of 224 x 224, the Hailo-8 processed 672 frames per second compared with the Xavier AGX’s 656 frames and sucked down only 1.67 watts (equating to 2.8 TOPs per watt) versus the Nvidia chip’s 32 watts (0.14 TOPs per watt).
“In recent years, we’ve witnessed an ever-growing list of applications unlocked by deep learning, which were made possible thanks to server-class GPUs,” said Hailo CEO Orr Danon. “However, as industries are increasingly powered and even upended by AI, there is a crucial need for an analogous architecture that replaces processors of the past, enabling deep learning to run devices at the edge. Hailo’s chip was designed from the ground up to do just that. We are excited to help customers drive their intelligent devices to new limits. A new age of chips means a new age of technology.”
Hailo says it’s working to build the Hailo-8 into products from OEMs and tier-1 automotive companies in fields such as advanced driver-assistance systems (ADAS), as well as targeting industries like smart cities and smart homes. In the future, Danon believes the chip will make its way into fully autonomous vehicles, smart cameras, smartphones, drones, AR/VR platforms, and wearables.
Unfortunately for Hailo, it will face stiff competition.
Mobileye, the Tel Aviv company Intel acquired for $15.3 billion in March 2017, offers a computer vision processing solution for AVs in its EyeQ product line. Baidu in July unveiled Kunlun, a chip for edge computing on devices and in the cloud via datacenters. Chinese retail giant Alibaba said it plans to launch an AI inference chip for autonomous driving, smart cities, and logistics verticals in the second half of 2019. And looming on the horizon is Intel’s Nervana, a chip optimized for image recognition that can distribute neural network parameters across multiple chips, achieving very high parallelism.
But Eric Yang, a founding partner at Glory Ventures, expressed optimism about Hailo’s ability to continue to differentiate itself.
“We have been following the AI compute global landscape closely and found Hailo’s technology to stand out,” Yang said in a previous statement. “We are impressed with the Hailo team and their ability to execute. We look forward to continuing our relationship with them as AI becomes a ‘must-have’ technology in every camera-enabled device.”