Serenity: Amazon has the clearest supercomputing AI investment logic, but supercomputing profits being eroded by semiconductors may trigger a turning point risk in capital expenditures.

BlockBeats reported on June 24 that market analysts have recently discussed the sustainability of AI capital expenditures by supercomputing manufacturers, with opinions showing clear divisions.

Serenity believes that Amazon is currently the most persuasive supercomputing investment target. The reasoning is as follows: Amazon has a large workforce of roughly 1.57 million employees, and AI capital expenditures can be directly translated into internal cost reductions—automation of labor through large language models, optimization of delivery with autonomous driving, and warehouse robots reducing operational costs. All of this, combined with the expansion of AWS computing power (self-developed Trainium chips) and potential chip sales revenue, make the commercial pathway clear.

Google is listed as the second choice, with AI investments backed by multiple pillars such as defending its search moat, Google Cloud revenue, and advertising optimization. However, its physical AI deployment is relatively weak. Microsoft and Meta, on the other hand, are considered not to have fully explained the necessity of their capital expenditures to the market. Microsoft's Maia chip progress is lagging, and its AI strategy is constrained by its investment in OpenAI, leading to subdued market sentiment.

However, skepticism cannot be ignored. Some argue that the ultimate beneficiaries of the current large-scale capital expenditures by supercomputing manufacturers are, to a large extent, semiconductor companies represented by Nvidia, while the profit margins of the supercomputing companies themselves are being continually squeezed. Should supercomputing manufacturers lose the motivation to increase spending, or if downstream model companies such as Anthropic report lower-than-expected revenue growth, the market valuation system currently built on the narrative of AI capital expenditure could face a sudden reevaluation risk.

Serenity emphasized that all supercomputing manufacturers are now vying for a “leading large model effect”—the platform with the strongest model will continue to accumulate training data and expand its capability advantages. This structural gap may already be solidifying, which is the underlying reason all players are racing to accelerate construction.