Now Playing
Qualcomm's CUDA Challenge, China's Cloud Loophole & Samsung Foundry Pricing
Audio is available on Spreaker — see link below.
Qualcomm open-sources the Modular AI platform in a direct strike at Nvidia's software moat, while Chinese AI labs exploit a legal remote-access loophole to run restricted Nvidia chips. Plus: Samsung raises foundry prices 15% as AI demand restores leverage to fabs.
Audio is available on Spreaker — see link below.
Qualcomm just open-sourced the entire Modular AI platform, including the Mojo compiler, under an Apache two-point-zero license. That's the move that turns what looked like a three-point-nine billion dollar acquisition into a direct, public challenge to Nvidia's software dominance.
Here's what the platform actually does. Modular provides a unified software stack that runs across hardware from Qualcomm, AMD, Nvidia, Apple, and Google.
The signal from the open-source release is that Qualcomm understands the actual problem. Nvidia's moat isn't primarily hardware.
While that software competition develops, there's a parallel pressure on Nvidia's dominance coming from a very different direction. Chinese AI firms including Moonshot AI and DeepSeek are accessing Nvidia's GB three-hundred chips remotely through cloud providers in Thailand, Malaysia, and Japan.
One more shift worth tracking. Samsung has raised SF four and SF five process pricing by ten to fifteen percent for new orders, with the steepest hikes in China and the US.
The three things that matter most from here: Modular's developer adoption rate over the next two to three quarters, Senate movement on RASA and whether BIS rule-making accelerates, and whether Samsung can sustain these foundry price increases without losing design wins to TSMC. The infrastructure layer of AI is under more competitive pressure than it looks from the chip headlines alone.
Chapter summary auto-generated from the verified script. Listen to the full episode for the complete content.