Serenity: Amazon has the Most Coherent Supercomputing AI Investment Logic, but Supercomputing Profits Are Being Eroded by Semiconductors or Triggering Capital Expenditure Inflexion Risk

BlockBeats News, June 24th. Market analysts have recently engaged in a discussion on the sustainability of AI capital expenditure by supercomputing companies, with significantly diverging views.

Serenity believes that Amazon is currently the most compelling target for supercomputing investment. The reasoning behind this is as follows: Amazon has a massive workforce of about 1.57 million employees, where AI capital expenditure can directly translate into internal cost reduction—through large language model automation of labor, autonomous driving for optimized delivery, warehouse robots for operational cost compression, along with the expansion of AWS computing power (using their in-house Trainium chip) and potential chip sales revenue, presenting a clear business path.

Google is ranked second, with AI investments supported by defensive search moats, Google Cloud revenue, and advertising optimization, but with a relatively weaker physical AI presence. Microsoft and Meta are seen as not having clearly articulated the necessity of capital expenditure to the market, with Microsoft's Maia chip progress lagging, and their AI strategy hampered by the investment in OpenAI, leading to subdued market sentiment.

However, voices of skepticism are also significant. Some argue that the primary beneficiaries of the current large-scale capital expenditure by supercomputing companies are mostly semiconductor firms represented by NVIDIA, with the profit margin of supercomputing itself being continually squeezed. If supercomputing companies lose the drive to expand spending or if downstream model companies like Anthropic experience slower-than-expected revenue growth, the current market valuation system built on the AI capital expenditure narrative will face a significant risk of abrupt revaluation.

Serenity emphasizes that all supercomputing companies are currently racing to seize the "leading large-model effect"—the platform with the most powerful model will continue to accumulate training data, expanding its competitive advantage. This structural gap may be solidifying, serving as a deep-rooted driver that all players are rushing to accelerate their development efforts.