Is the selling of DeepSeek bringing a buying opportunity? Morgan Stanley continues to support NVIDIA Corporation (NVDA.US) as its top pick.

onGPUAI NVIDIA CorporationAINVIDIA CorporationAmpere A100A40T40 Morgan StanleyNVIDIA Corporation(NVDA.US)DeepSeekNVIDIA CorporationThe lower-priced reasoning solution has mostly been replaced by Hopper and the current comprehensive training version of Blackwell. Analysts are confident that NVIDIA Corporation, as the largest beneficiary of long-term reasoning workloads, still maintains a strong position in the reasoning field.In the past few years, analysts believed that cheaper specialized products for inference (such as T4, L40) would be more popular. However, as these workloads become more complex, they continue to view high-end GPUs as the main source of inference computing. The expansion of computing test times further increases the computational requirements for inference, which is a good sign for high-performance solutions, and NVIDIA Corporation is the company that provides such solutions. According to ML performance benchmarks, the token per second provided by B200 on LLama270b is 2.5 times that of H200, and in a multi-chip situation, it should be even higher as they have transitioned to using the fifth generation nv-link by Blackwell, which doubles the bandwidth compared to the fourth generation. In other words, for every dollar invested, performance may at least double, which is significant for products that only started shipping a few quarters ago. Analysts believe that this performance has set a high benchmark for ASICs, and NVIDIA Corporation is expected to release the Rubin chip in about a month, which may further raise the bar for ASICs to catch up. IV. Stock market preference is expected to reverse in the second half of the year, with a bright future for the GPU market Although the stock market has recently favored ASICs over GPUs, analysts expect this preference to reverse in the second half of 2025. They continue to predict that GPU revenue will accelerate in the second half of the year, while ASICs will not see the same growth. In the long run, both markets have great potential, but analysts believe that the market sentiment shift towards NVIDIA Corporation is greater in the second half of the year. Analysts believe that the rise of DeepSeek has led to the outstanding performance of ASICs because people are increasingly focusing on service costs and believe that ASICs are more suitable as low-cost inference platforms for many NVIDIA Corporation customers. They confirm that optimized products are more suitable for specific workloads, which is reasonable. However, according to their observations, what truly changed the cost curve in 2025 was Blackwell, not ASICs. Furthermore, there is a viewpoint that ASIC projects have more lasting spending relative to NVIDIA Corporation products. Analysts believe that this viewpoint is somewhat correct, as there are commitments to enhance ASIC capabilities in areas with currently low penetration rates. But if training demand weakens, capital commitments related to training and inference may waver. Broadcom Inc. states that the company's $600-900 billion Serviceable Addressable Market (SAM) is primarily related to training. Analysts believe that when the training market growth slows down, NVIDIA Corporation's current advantage in training is a strength because when training is completed, it automatically releases GPUs typically used for inference. Without initial purchases, the competition threshold may be lower, but they see this as a long-term risk. In the short term, these GPUs may replace inference ASIC projects, as NVIDIA Corporation's flexibility can achieve a higher degree of overall unitization of cloud, internal, inference, and training workloads. NVIDIA Corporation is also constantly adjusting its software stack and providing these updates to customers for free, some of which have significantly improved performance in the past. Even if overall AI capital expenditure shrinks, analysts would not be surprised if NVIDIA Corporation gains market share, especially compared to first, second, or third generation ASIC products.