英偉達黃仁勳：年內將尋機進行資本運作

🔑 增強重點摘要

• •Nvidia completed a $20 billion deal with Groq, structured as a non-exclusive licensing agreement for inference IP and engineering talent, allowing key personnel like founder Jonathan Ross to join Nvidia while bypassing traditional acquisition regulations.[1][3]
• •Nvidia invested $5 billion in Intel for a roughly 4% stake, enabling Intel to develop custom x86 CPUs for Nvidia data centers and a system-on-chip integrating Nvidia RTX GPU chiplets.[3][6]
• •Nvidia participated in nearly 70 venture capital deals in 2025, including up to $100 billion committed to OpenAI, $10 billion in Anthropic, $2 billion in Synopsys, and $1 billion in Nokia, as part of its ecosystem expansion.[3]

🛠️ 技術深入

• •Groq's Language Processing Units (LPUs) specialize in high-performance, low-latency inference, potentially integrating into Nvidia's CUDA software platform.[2][3]
• •Market speculation post-Groq deal suggests Nvidia may explore replacing HBM with SRAM in future LPUs for inference-optimized chips, to be detailed at GTC 2026.[5]

🔮 前景展望AI analysis grounded in cited sources

⏳ 時間線