AMD Lags AI Chip Race; NVIDIA and Broadcom Lead Amid Negative Flywheel Risk

D.A. Davidson analyst Gil Luria warns AMD is falling behind in the AI chip race due to over-reliance on third-party networking suppliers and a weaker software stack, leading to lower real-world performance and utilization than spec. Luria initiated coverage on AMD with a “Neutral” rating and a $220 price target, noting MI300X and upcoming MI350X lag in deployment efficiency and end-to-end solution capabilities versus NVIDIA’s Blackwell architecture. The analyst highlights NVIDIA and Broadcom boast more comprehensive product portfolios and ecosystems that shorten client deployment timelines, while AMD’s reliance on third-party networking could lengthen integration and amplify efficiency loss, increasing per-unit cost. AMD may see limited adoption in high-demand AI, including by OpenAI, amid a “negative flywheel” reinforcing NVIDIA’s CUDA ecosystem and narrowing AMD’s penetration. Supply constraints could further favor larger leaders as TSMC may allocate advanced process capacity to higher-volume manufacturers. AMD’s shares fell over 13% after its Q4 report, with the outperformance largely driven by $3.9B in China sales not included in earnings guidance—leaving investors focused on non-China results. Long-term, AMD aims a $1T data center addressable market by 2030 and 60% CAGR for data center revenue from 2025–2030. Luria sees the mid-2026 OpenAI deployment of AMD M1450 as a key test of AMD’s ability to catch up in AI frontiers.