NVIDIA (NASDAQ:NVDA) and SoftBank's (NASDAQOTH:SFTBY) Arm Holdings recently announced a partnership that will make it easier for chipmakers to add deep learning abilities to next-gen consumer electronics, mobile devices, and Internet of Things (IoT) gadgets.
NVIDIA and Arm will integrate the open-source NVDLA (NVIDIA Deep Learning Accelerator) architecture into Arm's Project Trillium machine learning chip designs. The partnership will enable IoT chipmakers to quickly integrate AI features into their products.
What this partnership means for NVIDIA
NVDLA is based on NVIDIA's Xavier AI chip, which the company introduced earlier this year as the "world's most powerful SoC (system on chip)." The SoC, which was developed over a four-year period by over 2,000 engineers, has more than nine billion transistors and processes 30 trillion operations per second.
NVIDIA aimed the Xavier at the AI and driverless car markets. NVDLA is an open architecture based on Xavier that promotes a standard way of designing AI-powered products. In other words, it contains all the instructions to create chips like Xavier for different purposes. This allows companies that don't need an SoC as powerful as Xavier to scale down its design for cheaper products.
NVIDIA offers a suite of developer tools, including new versions of the TensorRT deep learning accelerator, through NVDLA. Since the platform is open source, other tools and features can be added by researchers and developers. NVDLA's expansion reinforces NVIDIA's reputation as the top chipmaker in AI, which it gained as a growing list of enterprise customers used its high-end GPUs for machine learning purposes.
What this partnership means for Arm
Arm doesn't make any chips. It licenses its low-power chip designs to a wide range of chipmakers like Qualcomm and MediaTek. NVIDIA also licenses Arm's technology in its Tegra CPUs, which power the infotainment, navigation, and driverless systems in higher-end cars, as well as its Shield devices and Nintendo's Switch.
Arm-based chips dominate the mobile and IoT markets, which prioritize power efficiency over raw horsepower. Arm claims that its chip designs power 95% of all smartphones, 95% of all wearables, and 85% of automotive infotainment and under-the-hood systems across the world.
Those figures should scare Intel (NASDAQ:INTC), which dominates the PC and data center markets with its x86 chips -- which generally favor horsepower over power efficiency. Intel has been pushing back against Arm with its low-power Atom chips and other chips for IoT devices, but it still remains a distant underdog in the IoT market. Intel also views NVIDIA as a threat, since AI-focused enterprise customers could postpone their data center CPU upgrades in favor of more GPU purchases from NVIDIA.
Arm's Project Trillium processors are scalable chips specifically designed for machine learning and neural networks. Integrating those chips into NVDLA could bolster Arm's AI capabilities at the "network edge" -- where mobile and IoT devices reside -- to allow its low-power chips to support smarter gadgets.
Why this is a win-win situation for both companies
Deepu Talla, NVIDIA's chief of Autonomous Machines, stated that NVDLA provides "all the ingredients for somebody to make it a dish including the instructions," and that the partnership with Arm is "basically like a microwave dish."
Moor Insights & Strategy analyst Patrick Moorhead noted that the deal "could enable Nvidia ML tech to be in even smaller IoT devices like home automation and even smartphones." Moorhead added that partnering with Arm wouldn't guarantee NVIDIA's success at the "very small edge," but it "increases its chances greatly."
Karl Freund, Moor Insights' lead analyst for deep learning, called the deal a "win-win" situation for both companies. Freund noted: "Nvidia is the clear leader in AI, and Arm is the leader in IoT, so it makes a lot of sense for them to partner on IP."
The bottom line
This partnership won't immediately move the needle for NVIDIA or Arm, but it leverages NVIDIA's strength in AI to expand its reach across the IoT market. Other chipmakers, particularly Intel, should watch out -- higher demand for smarter IoT devices could cause deaths by a thousand cuts to older players in legacy industries.