Decoding the Massive Amazon-Anthropic Deal: A Structural Analysis of the 'Next Move' in AI Investment
On April 20, 2026, news of another massive deal hit the AI industry. Anthropic announced that Amazon has agreed to invest an additional $5 billion, bringing Amazon's total investment to $13 billion. Meanwhile, Anthropic has committed to spending over $100 billion on AWS over the next 10 years, securing up to 5GW of computing power for the training and operation of Claude.
At first glance, it appears to be a simple investment, but digging deeper into its structure reveals a clever business model that integrates 'investment' and 'customer lock-in.' This article breaks down the full scope of the deal and explains the key points that investors and business professionals should grasp.
1. Overview of the Deal: The 'Exchange' of $5 Billion Investment and $100 Billion Expenditure
1-1. Scale of Investment and Milestone Conditions
In addition to the immediate $5 billion, this investment includes an option for up to $20 billion in additional funding tied to 'certain commercial milestones.' This means Amazon could potentially invest a total of up to $33 billion in Anthropic, combined with its existing $8 billion.
This initial investment was executed based on Anthropic's latest valuation of $380 billion.
1-2. Anthropic's 'Consideration': A 10-Year, $100 Billion AWS Commitment
Anthropic has stated it will spend over $100 billion on AWS technology over the next 10 years, including current and future generations of Amazon's proprietary AI chip, Trainium. Additionally, it will secure up to 5 gigawatts of capacity to train and deploy Claude models on AWS.
This structure is not merely an 'investment-to-return' model, but a circular model where 'invested funds flow back into the company's own cloud services.'
2. Amazon's Proprietary Chip Strategy: The 'Heart' of the Deal
2-1. Graviton and Trainium
At the center of this deal are Amazon's proprietary chips: the low-power CPU 'Graviton' and the Nvidia-competing AI accelerator chip 'Trainium.' The contract covers Trainium2 through Trainium4, though Trainium4 has not yet been released to the market.
Furthermore, Anthropic will also utilize tens of millions of Graviton CPU cores.
2-2. Rapid Growth of Amazon's Custom Chip Business
In his shareholder letter, Amazon CEO Andy Jassy stated that the company's custom chip business is growing rapidly, with an annual revenue run rate exceeding $20 billion, doubling from the previously disclosed $10 billion.
Regarding this deal, Andy Jassy stated, "Our custom AI silicon offers high performance at significantly lower cost". This partnership strengthens AWS's proprietary silicon strategy, and Anthropic's large-scale usage reinforces the positioning of Trainium chips as a substitute for third-party GPUs.
3. The Meaning of a 'Circular Deal': Similarities to the OpenAI Case
3-1. Comparison with Investment in OpenAI
This deal is reminiscent of the investment Amazon made in OpenAI just two months ago (contributing $50 billion to a $110 billion round, with a pre-money valuation of $730 billion), which also featured a structure combining cloud infrastructure services.
In short, Amazon is simultaneously playing the dual roles of 'investor' and 'infrastructure provider' for AI companies, building a scheme where invested capital flows back as AWS revenue.
3-2. The Reality of Multi-Cloud Strategy
What is interesting is that Anthropic has signed parallel contracts with multiple cloud providers. In November 2025, Microsoft agreed to an investment of up to $5 billion, and Anthropic announced it would purchase $30 billion in compute capacity on Azure. Furthermore, in April 2026, it expanded its partnership with Google and Broadcom, securing capacity on the scale of 'several gigawatts'.
While designating AWS as its 'primary' cloud provider (2023) and primary training partner (2024), Anthropic has also signed contracts with competing providers such as Microsoft and Google.
4. Anthropic's Business Growth and $800 Billion Valuation
4-1. Rapid Revenue Expansion
Anthropic's annualized revenue has surpassed $30 billion. This is a threefold increase from the previous quarter, reaching a level that exceeds OpenAI for the first time.
Looking back at the valuation trend, it has followed a parabolic curve from $61.5 billion in March 2025 (Series E), to $183 billion in September 2025 (Series F), and then to $380 billion in February 2026 (Series G, $30 billion raised).
4-2. The $800B Valuation and the Path to an IPO
Multiple VCs have proposed new investments to Anthropic at a valuation of over $800 billion, but the company has not accepted these at this time. This level is more than double the February round (pre-money $350 billion).
Meanwhile, Anthropic is in preliminary discussions with Goldman Sachs, JPMorgan, and Morgan Stanley regarding an IPO that could take place as early as October 2026, with banks estimating that it could raise over $60 billion in an IPO.
5. Three Points Investors Should Watch
5-1. Evaluation of Amazon's 'Investment Cycle' Model
Amazon's investment is a structure that is recycled into AWS revenue, and it can be called 'customer acquisition cost' rather than pure 'cost.' Amazon expects capital expenditures of approximately $200 billion in 2026 alone, centered on AI infrastructure, and Anthropic's $100 billion commitment holds significant meaning as a recovery scenario for that.
5-2. Competitiveness of Custom Chips
Amazon is steering toward building AI infrastructure ahead of demand, prioritizing early access to power, land, and silicon. However, there is a risk of utilization issues if demand does not materialize as expected. How close Trainium can get to Nvidia will hold the key to the success or failure of the deal.
5-3. Sustainability of AI Valuations
An $800 billion valuation against $30 billion in annualized revenue represents a revenue multiple of approximately 27x. The path from $30 billion in revenue to the profitability required to justify an $800 billion enterprise value is not guaranteed, and there is also the question of whether the valuation of AI as a whole has diverged from fundamentals.
