Barclays: LLM computing demand far exceeds expectations, AI chip spending wave has not yet peaked.

Barclays has stated that spending on artificial intelligence (AI) chips must increase to meet the minimum expectations of model developers. Barclays analyst Tom O'Malley said that since the release of NVIDIA Corporation's (NVDA.US) financial report at the end of August, the sell-off of AI chip stocks has caused a ripple throughout the entire AI ecosystem, with increasing doubts about whether the AI chip spending trend is "approaching its peak." Analysts believe that this view has not properly taken into account future computing needs. The analyst added that at least nine independent organizations are developing cutting-edge, large language models (LLMs) with massive numbers of parameters. Due to various reasons (return on investment, funding, training data limitations, roadmap issues, etc.), most of these companies cannot guarantee they will continue the next round of model development. However, just a few iterations of models require an incredible amount of computation. This demand far exceeds current industry expectations. The analyst believes that the following three factors are key to their coverage of stocks. First, the gradual achievement of expected production capacity is needed. Analysts estimate that by 2027, nearly 20 million chips will be needed for training to provide power for just three cutting-edge models with approximately 50T parameters. One of the major reasons for high demand is that the growth rate of new model computing requirements is expected to far exceed current expectations. Secondly, there is a way for businesses and consumers to achieve a win-win situation. Analysts believe that in terms of AI accelerators, a dual approach is needed, with NVIDIA Corporation and AMD's (AMD.US) commercial solutions being more suitable for training and inference of cutting-edge models, while custom chips will be used for more specialized workloads within data centers. The analysts added that they see this development unfolding as expected overall, with a few exceptions such as Apple Inc. (AAPL.US) using TPU. They continue to expect that this market differentiation will play a role in the future. Thirdly, the inference market will be very strong. O'Malley stated that NVIDIA Corporation recently claimed that about 40% of its chips are used for inference, and with other accelerator providers refocusing on the inference market, this provides support for the development of this emerging market. Inference will become the main channel for monetizing cutting-edge models currently being developed. Analysts stated that investing more resources to reduce inference costs will help increase the return on investment in model development.