As ties between OpenAI and its allies grow increasingly fraught, Microsoft has dropped a shock. Microsoft announced it will join AI chip leader NVIDIA to invest a total of $15 billion in Anthropic, OpenAI's rival.
Under the agreement, NVIDIA will invest up to $10 billion and Microsoft about $5 billion. This round pushes Anthropic's valuation to an astonishing $350 billion. In return, Anthropic signed a pledge: it agreed to purchase $30 billion of Microsoft Azure computing capacity and to prioritize deployment of NVIDIA's Blackwell and next‑generation Vera Rubin chip systems.
The deal is deeply ironic—Anthropic's largest backer had previously been Amazon (AWS), and Anthropic rose in part to resist a Microsoft–OpenAI alliance. Now Claude (Anthropic's model) will run across Amazon, Google and Microsoft cloud platforms, becoming in the truest sense an AI 'arms dealer'.
In‑depth analysis: Fourfold contest behind the successive investments — from 'symbiosis' to 'harvest'
This $15 billion financing agreement is far from a simple financial investment; it exposes four harsh truths about Silicon Valley's AI game:
1. Microsoft's hedge: OpenAI is no longer the only partner
Microsoft and OpenAI have long been seen as one of the tech industry's most stable alliances. But as Sam Altman in 2024–2025 pushed to free OpenAI from Azure compute dependence (seeking Oracle cooperation and amid rumors of building its own data centers), Satya Nadella showed seasoned CEO tactics: if you are no longer loyal, I won't bet on just one tree.
Investing in Anthropic is Microsoft's most forceful defensive move. It not only hedges against governance turmoil or technical bottlenecks at OpenAI but, by bringing Claude into the Azure ecosystem, ensures that whoever wins the large‑model war, Microsoft's role as a 'rentier' (cloud provider) remains unshakable.
2. NVIDIA's 'exclusive' strategy: strangling in‑house chip competitors
NVIDIA's $10 billion outlay is staggering. The logic is defensive locking. Amazon (Trainium), Google (TPU), and even Microsoft (Maia) are racing to develop in‑house AI chips to escape the 'NVIDIA tax'. As Anthropic is the largest compute consumer besides OpenAI, if Amazon fully bound it, Anthropic could migrate at scale to AWS's chips. Jensen Huang used $10 billion to secure Anthropic's long‑term commitment to NVIDIA hardware — a 'buying a moat' strategy to ensure promising model companies don't become testbeds for in‑house chips.
3. The 'left hand gives to the right' revenue magic (Circular Economy)
Note the numbers: Microsoft and NVIDIA invest $15 billion, but Anthropic will in turn buy $30 billion of Azure compute (including NVIDIA cards). This mirrors OpenAI's model:
- Big players fund startups (investment).
- Startups return the money to big players (by purchasing cloud services/chips).
- Big players report surging cloud revenue and stock gains.
- Startups see valuation spikes and big players show paper gains.
This 'cloud round‑tripping' model, while legal, creates an illusion of prosperity. Capital circulates among a few companies; while it drives technological progress, it casts doubt on the authenticity of AI industry revenues. It is not just investment but a financial engineering exercise to sustain high share prices.
4. Amazon's embarrassment and Silicon Valley's 'Game of Thrones'
Amazon is the biggest loser. Amazon previously invested billions in Anthropic, expecting it to be a trump card against Microsoft/OpenAI. Now Anthropic, holding Amazon's money and resources, has embraced Microsoft's Azure. This shows that in the compute‑constrained 2025 environment, compute supply capacity (who has the most H100/B200 clusters) outweighs equity agreements. To survive and catch up with GPT‑5/6, Anthropic had to bow to the Microsoft–NVIDIA alliance that holds the strongest compute reserves.
Conclusion: The endgame of monopolization
The deal marks the AI industry entering an era of tight oligopoly control. Independent AI startups have largely vanished, leaving OpenAI and Anthropic as the two super‑agents. The controllers behind them remain the familiar faces of Microsoft, NVIDIA, Amazon, and Google.
For ordinary observers it's a technological feast; but for regulators and markets, this closed loop of 'I invest in you, you buy my products, we pump valuations together' is hardening the AI bubble and making it more dangerous.