Tiny Corp Asserts AMD Is Close to Bridging the ‘Software Gap’ with NVIDIA; Team Red Poised to Excel if Competitor Falters in AI Chip Production

AMD is making significant strides in its software capabilities, prompting leading AI startups to assert that the gap between AMD and NVIDIA’s CUDA platform is narrowing.

AMD’s Evolving Software Stack Challenges NVIDIA’s CUDA Dominance

NVIDIA’s success in the AI inferencing market can largely be attributed to its robust software stack driven by CUDA. While AMD has made substantial gains in performance that increasingly rival Team Green, it has historically struggled to penetrate the software ecosystem. However, recent developments suggest that this dynamic may be shifting. Notably, Tiny Corp, a startup known for its consumer-focused AI solutions, has indicated that AMD is making significant progress in closing the software gap with NVIDIA.

AMD is closer to NVIDIA than most people think, we’re working hard on the software gap. Similar to Intel with CPUs, if NVIDIA stumbles for a generation, AMD can capture the majority market share. This is easier than gaming for market share to shift.

— the tiny corp (@__tinygrad__) August 18, 2025

According to Tiny Corp, a misstep by NVIDIA could provide an opportunity for AMD to take a lead in the AI sector, similar to its recent gains in data center CPUs. At the Advancing AI event in June, AMD unveiled a new version of its ROCm software stack, which includes support for cutting-edge frameworks like vLLM v1, llm-d, and SGLang. The updates focus on delivering optimizations in areas such as Distributed Inference, Prefill, and Disaggregation.

AMD ROCm 7 presentation on AI innovation and developer productivity features. — AMD ROCm 7 Announcement | Image Credits: AMD

The enhanced ROCm 7 version prioritizes inferencing workloads, demonstrating improved performance metrics like DeepSeek’s R1 FP8 throughput and competitive training capabilities, which AMD claims surpass those of NVIDIA’s CUDA. Furthermore, AMD plans to extend ROCm compatibility to Ryzen-based laptops and workstations by the end of the year, alongside comprehensive support for Linux and Windows platforms. This initiative reflects AMD’s commitment to making its software stack widely accessible to various user groups.

Should AMD successfully achieve parity with NVIDIA’s CUDA, it could redefine the competitive landscape in AI, positioning the company as a formidable challenger to NVIDIA’s long-standing dominance in the market. The evolving situation warrants close attention as both companies continue to innovate.

Source & Images