Nvidia has struck a non-exclusive licensing agreement with AI chip competitor Groq. As part of the deal, Nvidia will hire Groq founder Jonathan Ross, president Sunny Madra, and other employees. CNBC reported that Nvidia is acquiring assets from Groq for $20 billion; Nvidia told TechCrunch that this is not an acquisition of the company and did not comment on the scope of the deal. But if CNBC’s numbers are accurate, this purchase is expected to be Nvidia’s largest ever, and with Groq on its side, Nvidia i ...

Group 1 - Nvidia is acquiring AI chip startup Groq for $20 billion, marking its largest acquisition to date, which will enhance Nvidia's dominance in chip manufacturing [1] - Groq specializes in a new type of chip called a language processing unit (LPU), which claims to run large language models (LLMs) at 10 times the speed and with one-tenth the energy consumption compared to existing solutions [2] - Groq recently raised $750 million at a valuation of $6.9 billion, experiencing rapid growth by powering AI applications for over 2 million developers, a significant increase from approximately 356,000 developers last year [3]

Core Insights - AMD has entered a significant agreement with OpenAI to supply tens of thousands of GPU chips, amounting to 6 gigawatts of computing power over five years, marking one of the largest chip acquisitions in the AI industry [2] - The AI computing race is shifting focus from training large language models to inference, which is crucial for real-world applications, leading to increased demand for efficient computing solutions [3][4] - Morgan Stanley projects approximately $3 trillion will be invested in AI over the next three years, with a significant portion likely directed towards inference, potentially surpassing training in revenue and GPU units shipped [4] AMD's Position - The partnership with OpenAI positions AMD as a serious contender in the inference market, offering competitive performance and cost advantages compared to Nvidia [7] - AMD's MI series chips are becoming attractive alternatives for organizations that cannot afford Nvidia's top-tier GPUs, providing solid performance for inference tasks [7] Nvidia's Market Dynamics - Nvidia is expected to maintain its leadership in the AI market due to its established software ecosystem and partnerships, although its market share may decline as competition increases [5][6] - The company's dominance in training with its H100 and A100 GPUs may be challenged as the focus shifts to inference, which requires energy efficiency and hardware availability [3][4] Competitive Landscape - Intel is positioned to capture a share of the inference market with its diverse portfolio, including CPUs and accelerators, despite lagging in cutting-edge GPU technology [8] - ASICs are gaining traction for large-scale inference workloads due to their cost and energy efficiency, with companies like Marvell and Broadcom poised to benefit from this trend [8] Hyperscaler Strategies - Major tech companies like Amazon, Alphabet, and Meta are developing their own AI chips to reduce costs and gain supply control, which may decrease their reliance on Nvidia's GPUs [9] - Chinese companies such as Alibaba and Baidu are also enhancing their AI chip capabilities, with Alibaba planning to launch a new inference chip to support its cloud division [9] Infrastructure Demand - The growth of AI inference workloads will drive demand for supporting infrastructure, emphasizing the need for fast and reliable networking solutions from companies like Arista Networks and Cisco [9]

Investment Ratings - The report maintains a Buy rating on Aspeed, MPI, and a Neutral rating on WT Micro [2][3][4]. Core Insights - The semiconductor industry is experiencing significant growth driven by the rise of AI ASICs, which is positively impacting companies like Aspeed, MPI, and WT Micro [1][2][3]. - Price objectives (PO) for Aspeed, MPI, and WT Micro have been raised due to improved earnings outlooks and market dynamics [1][8][10]. Summary by Company Aspeed - Aspeed is expected to benefit from a favorable capex efficiency due to the rise of ASIC-based AI servers, leading to a long-term optimistic outlook [2]. - The company’s 2025/26/27E EPS estimates have been raised by 6%/7%/15% respectively, with a new price objective of NT$5,300 [2][12]. MPI - MPI's growth is anticipated to be driven by AI ASICs, with an expected increase in pin count and robust growth momentum [3]. - The 2025/26/27E EPS estimates have been increased by 5%/16%/12%, and the price objective has been lifted to NT$1,000 [3][16]. WT Micro - WT Micro is expected to see improved growth post-transition to TPU, with a more positive outlook for 2H25 [4]. - The 2025/26/27E EPS estimates have been raised by 5%/9%/10%, with a new price objective of NT$135 [4][21].