Amazon Doubles Investment in AI Startup Anthropic To $8 Billion

Amazon said on Friday it will invest an additional $4 billion in AI startup Anthropic, following earlier investments of $4 billion made in September and March. As part of the deal, Amazon Web Services will become Anthropic's primary training partner, with the AI firm committing to use AWS's Trainium and Inferentia chips for future model development.

Anthropic operates the Claude large language model.

Anthropic

[rossdee]

Anthropic? Isn't that the company responsible for Climate Change? (Anthropic Global Warming)

Re:

[backslashdot]

3/10 .. joke too contrived.

Re:

[Qwertie]

You're thinking of anthropogenic (=human-caused) global warming. Anthropic is named after things like the Anthropic Principle (the constraint that an observer must find themselves somewhere that supports intelligent life, otherwise they would not be an observer).

long term thinking

[awwshit]

If Anthropic is even slightly successful then Amazon makes its money back in fees alone.

Re:

[larryjoe]

If Anthropic is even slightly successful then Amazon makes its money back in fees alone.

Anthropic doesn't even need to be successful because it will pay Amazon lots of money for cloud services long before its success or failure is determined.

Interestingly, several decades ago Lucent did arguably a similar thing that was considered illegal. It would book billions of dollars of sales to a national telecom after giving that same telecom billions of dollars of loans. In light of today's trend, what Lucent did wrong was failing to structure the loans as an "investment."

Re:

[awwshit]

Anthropic could fail before Amazon makes $8B in fees. They need some level of success.

$21.28B

[anonymous scaredycat]

You are correct but they need a lot of success(or to find investment from elsewhere) for Amazon to make there money back . AWS costs Amazon money to provide i.e. cost of power, hardware, personnel, buildings, internet etc. I just did a quick check on google which for 2024 found "The AWS operating margin widened to 37.6%". So for Amazon to get the money back though fees Anthropic would need to pay AWS $8B/0.376=$21.28B.

nope, it's Claude

[Provocateur]

And we didn't claim to move everything to the Claude

I knew we didn't need that interpreter, just to make our announcements.

AI is crap.

[rlwinm]

I just tried (at the suggestion of a friend) asking an AI to fix a SW issue I'm facing. It's useless. Every answer it gave me was not only off topic but flat out wrong when it got close. My friend writes websites. I'm sure for something like that it's fine. I'm debugging startup code on an ARM CPU that works differently depending on the host glibc (the cross compiler binary is the same). And that's today's "simple" task for me.

I don't understand where all the hype comes from. Maybe if AI is used by peopl

Re:

[Rei]

This.

Claude 3.5 Sonnet (note: not Haiku) is the leading model for code today.

Ask a garbage model, get a garbage answer.

Re:

[Rei]

Also, AI models aren't *gods*. If there's a complex task that might take you days to do, maybe don't ask the AI to solve it perfectly single shot. E.g. let it instrument code with debugging to track down problems and feed it back the debugging data, or whatnot.

nope, it's Claude

[Provocateur]

And we didn't claim to move everything to the Claude

I knew we didn't need that interpreter

\o/

[easyTree]

So Amazon have determined <somehow /> that Anthropic's use of AI hardware is likely to grow more than other AI company which is amenable to motivation by their investment (interesting in itself in my opinion) and used the investment to tie the growth in uptake of their AI hardware to Anthropic's use of AI hardware.