Amazon ploughs another $4 billion into Anthropic

Amazon has invested another $4 billion into OpenAI rival Anthropic, bringing the hyperscaler's total investment to $8 billion.

The announcement didn't reveal Anthropic's valuation, merely stating that Amazon remains a minority shareholder.

There is some information floating around out there, courtesy of The Information (paywall), which reported in September – citing an unnamed existing investor – that Anthropic was in talks to raise capital that would value it at between $30 billion and $40 billion. That seems a lot more grounded than the $157 billion valuation of ChatGPT maker OpenAI, following its most recent funding round.

Eye-watering sums of money aside, what is perhaps more significant regarding this transaction is the strengthened partnership that comes with it – one that promises deeper integration between Amazon Web Services (AWS) and Anthropic's Claude large language model (LLM).

Under the agreement, Anthropic will work closely with AWS' Annapurna Labs to develop and optimise future iterations of AWS' in-house AI accelerator – called Trainium – to handle demanding AI workloads.

"Through deep technical collaboration, we're writing low-level kernels that allow us to directly interface with the Trainium silicon, and contributing to the AWS Neuron software stack to strengthen Trainium," said Anthropic. "Our engineers work closely with Annapurna's chip design team to extract maximum computational efficiency from the hardware, which we plan to leverage to train our most advanced foundation models."

"This close hardware-software development approach, combined with the strong price-performance and massive scalability of Trainium platforms, enables us to optimise every aspect of model training from the silicon up through the full stack."

Up to this point, Nvidia has been by far the biggest beneficiary from AI and its unquenchable thirst for high-performance silicon. This tight integration between AWS hardware and Anthropic's Claude AI suggests that there is a desire for a bit more choice in the chip market.

On the software side, Claude is one of many LLMs supported by Amazon Bedrock, its suite of AI application development tools, which is made available as a managed service. BT announced in September it will use Bedrock to create its in-house generative AI (GenAI) platform.

"Through Amazon Bedrock, Claude has become core infrastructure for tens of thousands of companies seeking reliable, practical AI solutions at scale," Anthropic said, noting that the likes of Pfizer, Intuit and even the European Parliament are making use of it.

Indeed, as much as AWS pitches Bedrock and its related AI solutions as LLM-agnostic, its financial interest in Anthropic might just put this neutrality to the test.

Similarly, Anthropic was born out of a desire by former OpenAI employees to promote responsible AI, and mitigate the potential pitfalls of unabated AI development.

While AWS doesn't call the shots at Anthropic, it could quite easily wield the kind of influence that might just dilute Anthropic's well-intended approach to AI. After all, AWS is one of multiple hyperscalers currently engaged in an AI land-grab – one of the others being OpenAI backer Microsoft. It might prefer it if Anthropic moves more quickly than cautiously.

Mutterings from this year's exodus of talent from OpenAI has created the impression that the company is prioritising the capabilities of its GPT LLM over any guardrails that might be required to keep it under control.

For all our sakes, let's hope that as Anthropic becomes more successful, it maintains a watchful eye on its moral compass.