Model Moats: The Vanishing Weight and the Price of the Watt
Yahoo Finance ·
Anthropic has committed nearly $90 billion to secure compute infrastructure through Nscale at the Monarch Compute Campus in West Virginia, including $44.6 billion confirmed in a September SEC filing and an additional $45 billion in advanced negotiations reported on October 19. As model weights commoditize, the industry battleground has shifted to power control. On October 20, DeepSeek open-sourced DeepInfer, and llm-d 1.0 reached general availability, with IBM reporting a 5-10x cost reduction for GLM-5.2 on 544 H100s. Meanwhile, Chinese giants ByteDance, Alibaba, and Tencent have committed roughly $35 billion for over 3 million units of Nvidia's RTX PRO 5500, facing a structural bottleneck with fill rates stuck at 5 to 10 percent. Ultimately, the true defensive assets are no longer software models, but physical campuses, chips, and power agreements.
AI 시장 분석
As Anthropic invests approximately $90 billion to secure computing infrastructure, the AI market paradigm is shifting from software to physical power and hardware. With DeepSeek unveiling its inference engine and launching llm-d, the software layer is commoditized, making companies that control power and chips the core competitive players. From an investor perspective, attention must be paid to shrinking AI software margins and rising valuations in the infrastructure and power sectors.
상승 영향
- Semiconductors — Anthropic's $90 billion infrastructure investment and massive chip orders from Chinese big tech are expected to drive strong revenue growth for core hardware manufacturers like NVIDIA.
- AI — The commoditization of the software inference layer is sharply elevating the strategic value of companies that monopolize physical data center campuses and power infrastructure.
하락 영향
- Software — The launch of DeepSeek's open-source inference engine and llm-d service layer has caused commercial API prices to drop 5 to 10 times, sharply plunging margins in the software sector.
DYAX 전담 분석
As AI model weights are commoditized and inference costs decline, high-margin areas are shifting from software to hardware and power infrastructure. Anthropic is pouring massive capital into securing gigawatt-scale power and NVIDIA chips in West Virginia, moving to preempt physical infrastructure.
The bullish scenario involves infrastructure suppliers and power-related companies enjoying monopolistic positions and recording sustained revenue growth, while the bearish scenario is that excessive infrastructure investment costs act as financial burdens. Key indicators to watch are the signing speed of large-scale data center power acquisition contracts, NVIDIA hardware shipment volumes, and the alleviation of supply bottlenecks.
AI가 생성한 분석으로 투자 자문이 아닙니다.
DYAX Investor Sentiment
Bullish (Long) 51% · Bearish (Short) 49%
287 participants
Related News
- Why JPMorgan Predicts SpaceX Will Triple by Targeting a $28 Trillion Market
- Judge Orders Meta to Pay $25 Billion to New Mexico in Cambridge Analytica Lawsuit
- CoreWeave Plunge Sends Fresh Warning to the AI Boom
- Hop-on Subsidiary Digitalage Targets Full Commercial Launch Before End of 2026
- Elon Musk Suggests Potential Merger Between SpaceX and Tesla
- SpaceX Provides 20 Percent Bonus Pay to Accelerate Memphis Supercomputer Launch
