Amazon Raises Total Investment Cap in Anthropic to $25 Billion
Multiple English media outlets, citing company statements and informed sources, report that Amazon has agreed to increase its investment in Anthropic by up to $25 billion, significantly raising the total investment cap for the large model company. The funds will be closely tied to AWS cloud services and its self-developed Trainium chips. This new commitment expands upon a previous investment of up to $8 billion, with Anthropic promising to invest over $100 billion in AWS over the next decade for the computing power and chip resources needed for model training and inference.
According to reports, under the new framework, Amazon will initially inject about $5 billion in immediate funding, with the remaining approximately $20 billion to be disbursed in phases based on subsequent "business milestones." The overall structure will still involve a combination of convertible bonds and equity to acquire a minority stake, while fully integrating the Claude model into the AWS and Amazon Bedrock ecosystem. Wall Street analysts estimate that in an optimistic scenario, Anthropic's business could contribute about $25 billion in revenue by 2027, adding several percentage points to AWS's overall revenue and becoming Amazon's third-largest single investment after Whole Foods and MGM.
Source: Public Information
ABAB AI Insight
This "up to $25 billion" structural commitment is essentially Amazon's long-term power lock-in in the AI infrastructure race, rather than a simple VC-style financial investment. By tying the funds to the requirement of "spending $100 billion on AWS computing power," Amazon is effectively pre-booking high-quality AI workloads for many years to come, transforming Anthropic from a potential multi-cloud customer into a "deeply integrated AWS exclusive engine." In the cloud computing competition, this is akin to upgrading model providers from traffic entry points to "power consumption bases," directly providing certainty in demand for its own chips and data centers.
From an industry perspective, this move pushes the vertical integration of large models, cloud, and chips into a new phase. The Microsoft-OpenAI and Google-DeepMind/Anthropic combinations have already demonstrated the paradigm of "models as the traffic front end of cloud"; meanwhile, Amazon, previously seen as the "third pole with a slightly slower start," is attempting to narrow the gap by doubling down on Anthropic and betting on Trainium, ensuring that Claude models become one of the default choices in the enterprise market and developer ecosystem for AWS. This will further squeeze the survival space for independent clouds and smaller model companies, as the integrated package of "top models + leading cloud + self-developed chips" is becoming the default configuration for large enterprises.
Financially, expanding the investment in Anthropic from single-digit billions to potentially $25 billion indicates that Amazon views AI as "infrastructure-type capital expenditure," rather than a "high-risk innovation project." This type of investment is increasingly aligned with the logic of building data centers and high-speed logistics networks: significant upfront asset accumulation, but once scaled, marginal costs drop significantly, recovering in the long term through high-margin cloud services and model APIs. For shareholders, this represents a repricing of future cash flow discount models and a re-anchoring of the "non-AI business P/E ratio"—the market will be more willing to assign Amazon a valuation multiple closer to that of an "AI infrastructure platform" rather than "traditional e-commerce + cloud."
On a deeper level, this transaction also accelerates the process of "financialization of model companies." Anthropic, as an independent company, still appears to be raising funds from multiple sources and maintaining options for collaboration with various cloud vendors; however, with a commitment to AWS for over $100 billion in computing power consumption within ten years, and simultaneously introducing multiple rounds of funding from Amazon through a convertible bond structure, its capital structure and strategic freedom are effectively evolving towards "semi-endogenous assets." This suggests that the future fate of top model companies may not simply be an IPO or acquisition by a giant, but rather embedding themselves into the balance sheets of several super platforms through complex convertible bonds, revenue-sharing, and computing power agreements, becoming a new form of "quasi-public infrastructure" in the global technology capital landscape.