This week was big for both NVidia and Microsoft. Both companies demonstrated new “custom” designs catered to AI processing. These companies know that hardware is sexy, yet most people will never touch these components. Hardware excites people and companies know that, but that’s not SaaS.
NVidia
Microsoft
Microsoft showed a Cobalt 100, 128-core ARM chip along with the Maia 100 accelerator. Then it showed rack of boards, tubes, and radiators. Satya Nadella held a square labeled Cobalt 100.
CUDA is powering many of these more fun generative models for images and video. This gives NVidia an advantage in terms of novelty and potential marketability of more creative tools and simulations. Most graphical AI efforts are already built on CUDA and these should run easily in these new platforms. Image generators have a tendency to grab the public’s attention. This may keep generators like Adobe Firefly within the public’s attention regularly.
Microsoft’s Office 365 AI additions are less sexy. But they play to Microsoft’s strengths in business data and productivity tools like Word, Excel, and PowerPoint. Enterprise businesses could offer more practical benefits in the long term for the more traditional work. We also know that Microsoft says it wants to do something with Microsoft Mesh to bring Teams into the metaverse with VR and flattened representations of 3D office spaces. Would it go viral? Less likely.
It’s cool. AI is cool. Despite Apple’s newfound hatred of leather, Jensen Huang’s leather jacket is cool. “The more you buy, the more you save,” as Huang says. And we know you miss Clippy. These brands are part of our lives. You are likely using consumer-grade technology from both Microsoft and NVidia, whether you realize it or not.
Power efficiency is critical for the business. These chips need to serve as many users as possible. The public can only grow to love and use AI habitually when we know it will be there and working quickly. The hardware must use as little power as possible to improve profit margins for these companies. Will the revenue of the subscriptions outweigh the cost of the electricity and hardware? Will AI excitement endure or is it the next bubble?