Amazon is in discussions to make another multibillion-dollar investment in OpenAI competitor Anthropic, as reported by The Information on Thursday. However, there's a catch: Amazon wants Anthropic to use its custom AI chips—a deal that could change the course of the investment.
Why it matters: The talks highlight a tension between major cloud providers and AI developers. On one side, Amazon wants Anthropic to use its Trainium chips to reduce dependency on NVIDIA's popular GPUs. On the other, Anthropic prefers Nvidia’s tech for its maturity and compatibility. How this negotiation plays out could affect the size of Amazon's total investment.
What’s happening:
- Amazon previously committed up to $4 billion in funding to Anthropic, a deal that included access to Amazon Web Services (AWS) infrastructure, including some Trainium-powered servers. Anthropic, however, mainly relied on Nvidia chips.
- This time, Amazon is pushing Anthropic harder to adopt more of its Trainium servers. The request is significant—Amazon’s chips aren’t as established as NVIDIA' CUDA-backed GPUs, which AI developers are accustomed to using.
- Amazon’s motivation is clear: if Anthropic uses more Trainium chips, Amazon reduces its dependence on NVIDIA, whose hardware has been in high demand as AI development booms.
The stakes: Anthropic’s decision to accept or decline Amazon’s request could affect the amount Amazon is willing to invest. The discussions reflect a broader dynamic in the industry where partnerships between cloud giants and AI firms come with compromises. Anthropic’s flexibility to use both Google and Amazon as cloud providers might give it an edge, but adopting Amazon’s chips could limit this flexibility in the long term.
Zoom out: These negotiations underscore how the race to dominate AI development is reshaping relationships in tech. Amazon needs more AI-driven customers like Anthropic to adopt its technology to lessen reliance on third-party hardware providers. For Anthropic, balancing the promise of funding with the technical preferences that best support its AI training efforts is a careful act—one that might determine the company’s trajectory as it aims to compete with OpenAI.
The bottom line: Anthropic’s reliance on NVIDIA chips is understandable given their established ecosystem and tools. However, if Amazon offers sufficiently attractive terms, Anthropic may reconsider, particularly as it seeks to secure funding at a valuation of up to $40 billion. The key question is whether the trade-off in technical flexibility is worth the potential influx of billions in investment.