Amazon and AI company Anthropic have entered into a new major agreement where the e-commerce giant is injecting $5 billion into the company behind the Claude models. In return, Anthropic has committed to purchasing cloud services from Amazon Web Services (AWS) for a staggering $100 billion, according to TechCrunch.
This is not the first time the two companies have entered into such a reciprocal agreement, and the transaction follows a pattern that has become increasingly common in the AI industry.
Money in a Loop – How the Deal Works
Structurally, this is what analysts call a «circular AI agreement»: Amazon invests in Anthropic, and Anthropic uses a large portion of that capital to purchase Amazon's own services. The result is that the money largely flows back to the investor – in this case, Amazon.
The enormous asymmetry between the size of the investment and the purchase commitment – a twentyfold difference – is striking and underscores how closely intertwined the fates of the two companies now are.
Industry Experts Warn Against «Circular Financing»
This type of agreement is facing growing criticism from market analysts. Jacob Bourne from eMarketer has previously pointed out that this trend indicates the AI industry is not particularly diversified, and that such constructs do not necessarily reflect real, organic demand for AI services.
When business-critical automations run exclusively on one vendor's models, API changes, price increases, or service limitations can halt operations overnight.
The criticism follows several lines. Firstly, questions are raised about whether such agreements contribute to artificially inflating market values, in a way reminiscent of the situation before the dot-com crash. Secondly, the close ties can create systemic risk: if one of the major AI companies struggles, it could, according to analysis firms, trigger a chain reaction throughout the entire network of reciprocal agreements.
Risk of Vendor Lock-in
For companies building solutions on Anthropic's models via AWS, the agreement brings to light a broader discussion about vendor lock-in. When an actor is deeply integrated into one platform's proprietary APIs and data infrastructure, switching to alternatives becomes costly and complicated.
Industry analyst Rebecca Wettemann has described it this way: Vendors are betting that high switching costs involved in rebuilding solutions on another platform will make customers «sticky» – that is, locked in.
Market Power and Competition Law
The agreement between Amazon and Anthropic could also attract attention from competition authorities. Exclusive ties and closed ecosystems could, in principle, hinder the ability of new players to compete, and in the extreme, conflict with competition law.
It is worth noting that the AI industry has generally generated limited profit relative to the enormous capital invested. OpenAI, according to available forecasts, estimates operating losses of up to $74 billion in 2028 alone – a figure that illustrates how immature the profitability picture still is.
Anthropic and Amazon have not yet commented on the criticism regarding the deal's structure beyond public press releases, according to TechCrunch.