NVIDIA and ARM: Massively Changing The AI Landscape

SHARE

The merger between NVIDIA and ARM is a potentially massive game-changer for artificial intelligence in that ARM is the most common technology used for inference, and NVIDIA’s platforms are the most commonly used for training. On paper, this merger effectively gives NVIDIA substantial control and influence over the emerging AI market. 

Let’s talk about NVIDIA and ARM this week and what it means for the AI segment.

Training vs. Inference

NVIDIA is the training king and has been for some time. Most of the large-scale AI training solutions default to NVIDIA’s technology, but on inference, where much of the volume is, NVIDIA hasn’t been as strong. But ARM is the inference king collectively, and if NVIDIA can create unique synergies between ARM inference and NVIDIA training, the resulting competitive advantage in AI will be formidable for any third party to overcome. 

This synergy should benefit both the effectiveness of ARM/NVIDIA solutions and their relative cost and will force competitors to create similar synergies. But while coming up with new inference engines is certainly possible, the core competitive advantage in Training. The path is far longer, more complicated, and vastly more expensive. 

This thorny path suggests that without a massive effort, it is unlikely NIVIDA’s market leadership can be challenged today. Once they integrate ARM into their solution, that competitive advantage will substantially increase.

Now, this also creates a more substantial opportunity for an interesting Deep Learning implementation where the inference solutions at scale contribute the capabilities of the centralized training systems, and the solution teaches itself how to be more effective and accurate over time. 

The massive potential increase in the inference footprint resulting from this merger should provide NVIDIA with another unprecedented advantage in data collection for their respective solutions and collect massive amounts of data. When it comes to Deep Learning, data is critical; the more you collect and can process, the more intelligent the resulting system becomes, and this NVIDIA ARM solution, once mature, has the potential of being unmatched in the market. 

The Smartphone Edge

Now you would think that NVIDIA wouldn’t be that interested in smartphones, which currently are almost entirely ARM-based, but each Smartphone is a massive sensor. If you can get the smartphone owners to agree to allow data capture from them, you could build an incredibly powerful predictive engine. 

This engine could do things like identifying the spread of a virus, assure social distancing, and even provide a faster warning on disasters or other localized events. And this is on top of core benefits like pointing out shopping values, identifying places you don’t know about but would uniquely appreciate, and make you aware of close by events (like rallies or demonstrations).

Also, smartphones could be the ideal terminal into the resulting smart system making the related apps more intelligent and informative. So rather than being uninteresting, the ARM penetration on smartphones coupled with an NVIDIA AI back end (either inference or training) could result in a level of intelligent capability we can only imagine today. 

In short, this should result in an intelligent assistant that is truly intelligent and can report on your child’s status, alert if a loved one is in trouble, and better keep yourself safe.  

If the solution was tied into one of the mega Smartphone vendors like Samsung, LG, or Motorola, the result could be a rather significant competitive advantage for the related phones and smart options.  

Wrapping Up:  A New AI World

While this clearly won’t happen instantly, given ARM’s heavy use in inference today, their dominant position in smartphones, and NVIDIA’s dominant position in Training, the result of this merger is effectively a potential future new AI world order. With the potential for mass data capture and mass consumption when optimized across the firm’s new ecosystem, the resulting solutions should be smarter, better able to predict future events, and provide a far more accurate real-time picture of the world around us.

While typically this would take 5-10 years for the real benefits to emerge, given ARM is widely penetrated and NVIDIA’s AI solutions dominate, the timeline is likely in the 2-5 year range. So we are talking about significant changes within a surprisingly short period if the merger is approved. 

This timeline will all result in a significant improvement to AI projects that require real-time data at scale, and the outcome will be nothing short of amazing, and we should see benefits before mid-decade.  Things just got a ton more interesting.



NewsletterDATAMATION DAILY NEWSLETTER

SUBSCRIBE TO OUR IT MANAGEMENT NEWSLETTER