Qualcomm wants an open software platform for AI chips

Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

1. 35:01

   invented multiple programming languages, including Swift, and you bought his company Modular recently. And for people who don't know what Modular was doing, it was essentially a bid to crack the software mode that keeps the AI industry tethered to NVIDIA and CUDA, the software layer that NVIDIA uses. And he told me, and this was several months ago, that there's this coming wave of different kinds of chips, which you were just talking about, the CPUs, all these things used for AI, and developers need a way to use all of them.

2. 35:29

   And I'm curious how much of that strategy that he had at Modular is now the Qualcomm chip and data center strategy. You know, I met Chris and we're so fortunate to have he and his team in Qualcomm. I'm sure he had options to go to very different companies if he wanted to. but I think what made modular and Qualcomm a perfect fit is kind of we share we we had we share the same vision about what is happening with the industry first of all a company like Qualcomm look when we invented some of the most incredible technology in the world we license for everybody right so so a company like Qualcomm we believe that you gain a lot more innovation and scale if the technology is widely deployed and you have more companies contributing to the innovation. Today, you do have a control point on the software side, and there is an

3. 36:25

   industry need to have an open alternative to run AI, because AI is not going to be unique to the data center. It's going to go everywhere. You talk today about a multitude of hardware in the data center. Think about the multitude of hardware when you put cars and PCs and phones and wearables, all of this. Developers need to target all of this compute to run models, big and small, and they need it to have a common platform. And it doesn't exist in AI, what has been mature in traditional data centers like Linux as an example, or even if you go to the cloud, think about the Google Kubernetes, or even in our space, when you think about the horizontal phone space is Android as an open platform. Those things don't exist for AI. So the vision was modular.

4. 37:14

   Chris and the modular team was thinking, we need to develop this. Developers need this. The industry need this. We're going to develop. It's a modern. It's have higher performance and runs on any hardware. And I said, okay, let's do this at Qualcomm and let's offer to everybody. And to the point that we at Qualcomm will celebrate, we will celebrate every single modular deployment on NVIDIA, on AMD, on Broadcom, on any other chip, on Intel, because we actually believe the industry will benefit from a standard, common, open platform.

5. 37:49

   We said that we're going to do Apache 2.0 license. It's going to be open source. What it means to the Qualcomm chip, once the platform is open, we can't compete on hardware. That's what we've done our whole life. That's our mobile business. That's everything that we do. So we're not afraid. So compete on the merits of the hardware versus software logic. We compete for the matter and the harder. That's going to allow harder innovation.

6. 38:09

   If you have a closed system, I may have a harder innovation, but nobody will ever be able to use it. So for us, since we're new to the data center, we enter in the space, then we're happy because we'll create an opportunity to compete in the harder. The other part of the equation is the conversation we had before about AI smartphones and PCs and cars. The edge is going to be part of this compute continuum when everything is going to be running AI and you have the same problem.

7. 38:41

   We need the platform that scales. So we're going to be, like I said, doing everything possible to make sure that every hardware company can use it. Modular is going to be agnostic. Modular is going to be run as a Qualcomm company. And we're just going to be supporting modular chips and try to build great hardware. Let's end on something that I am also trying to figure out which is 6G. So I've covered multiple Gs.