What the ARM Acquisition by NVIDIA Means for the Data Science Community?
NVIDIA Acquires ARM for $40 Billion— What Paradigm Shifts this Deal Might Bring About for the Data Science & AI Communities, and What Can We Expect in the Future?
NVIDIA took the technology world by storm when they announced the acquisition of the British chip designing and licensing company ARM for $40 billion. This stake acquisition agreement, signed between NVidia and the previous owner SoftBank Group Corp., is the most talked-about news in the tech communities right because of how impactful this decision is being perceived for the future of mobile computing and AI. In this article, we are going to have a look at how this acquisition might pave a golden path for artificial intelligence in the coming years.
But before we move on to that, let us first understand what are the current standings of these two companies, NVIDIA, and ARM, and why are they so important.
NVidia controls around 70% (as per BusinessWire) of the overall GPU market, catering to the ever-growing demands of the gaming and professional markets (like the AI and Automotive industries).
This figure goes even higher if we consider just the more niche markets— Deep Learning, Data Science, and Artificial Intelligence, as these sectors primarily rely on NVidia’s CUDA technology for parallel computing. For those of you that don’t know what that is, basically NVidia’s CUDA-based GPUs allow AI models to be trained much at much faster speeds as compared to on a normal CPU. To put things into a better perspective, while some very compute-intensive neural network might take almost a month to train on a CPU, the same thing can be done on a CUDA-based NVidia GPU within a few hours. And as the saying goes, when it comes to business, time is money. In fact, the reason why Deep Learning has seen exponential growth over the last few years is that GPUs have gotten cheaper and faster as compared to, say, two decades ago. As a result, NVidia has a very strong grip over the market when it comes to the AI and related industries.
As for ARM, one can say that it’s a monopoly in its mobile CPU chip designing division. Almost every mobile SoCs (system-on-chip) maker in the market, the big names including Qualcomm, Apple, and Samsung, use ARM-licensed architectures for their SoCs. As per some speculations, ARM has a 95% (as per Wikipedia) control over the mobile chip designing sector. Since most people use their smartphones as their daily drivers to connect to the rest of the world via the Internet, ARM is basically controlling our mobile computing experience right now.
Now, as we can see, both these companies rule their respective areas of operation in the consumer markets. As a result, their coming together is obviously a huge deal. But how exactly is it affecting the AI industry in any way? And why is it so important from that perspective?
Well, here’s why.
If we look at the current scenario, mobile devices are still lagging behind the more traditional desktop computers by miles when it comes to the artificial intelligence experience. This is because the SoCs on mobile phones are not powerful enough to run some of the advanced, compute-intensive models directly on the devices itself. The traditional GPU-powered computers, on the other hand, can not only run these models efficiently, in fact, but they are also being used to create and train these technologies.
In order to compensate for this lack of compute-power, in the past, we have seen companies come up with different hardware and software-based solutions. For example, Apple uses special hardware acceleration on its SoCs called Neural Engine to assist in running real-time machine learning applications locally on the device itself. Similarly, Google also came up with the Neural Net API for its Android framework as a software-based mechanism to run AI and ML models locally on smartphones.
However, because these technologies are more of a substitute solution rather than actual GPU-based solutions, using these they require a lot of software workarounds on the developers’ part, which is obviously not very efficient and might result in frequent during the production and development stage.
Up until now, ARM was undoubtedly doing an exceptional job with its mobile chip designing. If we compare the existing flagship devices in the market from the devices launched 4-5 years back, we will find that mobile computing has changed drastically. The only department in which the mobile SoCs are actually lagging behind as compared to traditional computers is GPU acceleration.
But now, with this merger of ARM and NVidia, you can expect some major improvements coming up. For a starter, we can expect NVidia, with its years of experience in desktop GPU development, to step forward for a joint venture with the team of ARM researchers and engineers. Who knows a company that excels in graphics-based solutions and the other with expertise in mobile chip designing might be able to come up with a product, say, a CUDA-compatible mobile SoC, that might be able to run fully-fledged AI models locally on the mobile phone itself, or in the extreme cases, support for on-device model training that will allow the device to learn based on the users’ usage patterns.
If the two companies, together as one, are somehow able to pull this off, it could imply that in the future, we won’t have to rely on some unstable cloud-based workaround to run AI models on our devices. This might also mean that the users would not have to send their private data to some unknown location for the sake of getting “a tailored user experience”.
For the developers working on AI technologies, this will directly unlock the access to an entirely new area of development that till data had been either completely out of their reach due to existing technological limitations, or, it required so much work around that just was not worth it.
With this agreement between NVidia and SoftBank for the acquisition of ARM now entering into a final stage, we can start expecting some major AI-related developments in the smartphone and other mobile computing divisions coming soon. The future of how users will interact with their mobile devices, and how AI will influence their lives is definitely a golden one.
Read more about NVIDIA’s acquisition