HN Reader

Curious if anyone knows the logistics of these cloud provider/AI company deals. In this case, it seems like the terms of the deal mean that Anthropic ends up spending most of the investment on AWS to pay for training.

Does anthropic basically get at cost pricing on AWS? If Amazon has any margin on their pricing, it seems like this $4B investment ends up costing them a lot less, and this is a nice way to turn a cap ex investment into AWS revenue.

They certainly need the money. The Pro service has been running in limited mode all week due to being over capacity. It defaults to “concise” mode during high capacity but Pro users can select to put it back into “Full Response.” But I can tell the quality drops even when you do that, and it fails and brings up error messages more commonly. They don’t have enough compute to go around.

Anthropic should double down on the strategy of being the better code generator. No I don't need an AI agent to call the restaurant for me. Win the developers over and the rest will follow.

I think Claude is actually superior to ChatGPT and needs more recognition. So good news, I guess

Anthropic will be the winner here, zero doubts in my mind. They have leapfrogged head and shoulders above OpenAI over the last year. Who'd have thought a business predicated entirely on keeping the ~1000 people on earth qualified to work on this stuff happy would go downhill once they failed at that.

Anthropic gets a lot of it's business via AWS Bedrock so it's fair to say that Amazon probably has reasonable insight into how the Claude usage is growing that makes them confident in this investment

This makes sense in the grand scheme of things. Anthropic used to be in the Google camp, but DeepMind seems to have picked up speed lately, with new “Experimental” Gemini Models beating everyone, while AWS doesn't have anything on the cutting edge of AI.

Hopefully this helps Anthropic to fix their abysmal rate limits.

> Amazon Web Services will also become Anthropic’s “primary cloud and training partner,” according to a blog post. From now on, Anthropic will use AWS Trainium and Inferentia chips to train and deploy its largest AI models.

I suspect that's worth more than $4B in the long term? I'm not familiar with the costs, though.

I had to switch from Pro to Teams plan and pay 150 USD for 5 accounts because the Pro plan has gotten unusable. It will allow me to ask a dozen or so questions and then will block me for hours because of „high capacity.“ I don’t need five accounts, one for 40 USD would be totally fine if it would allow me to work uninterrupted for a couple of hours.

All in all Claude is magic. It feels like having ten assistants at my fingertip. And for that even 100 USD is worth paying.

The status pages of OpenAI and Anthropic are in stark contrast and that mirrors my experience. Love Anthropic for code and its Projects feature, but OpenAI is still way ahead on voice and reliability.

Claude is absolutely incredible. And I don’t trust openAI or Microsoft so it’s nice to have an alternative.

Will Amazon leadership require it's new Gen AI to physically move itself to an office to perform valid work?

I love Claude 3.5 sonnet and their UI is top notch especially for coding, recently though they have been facing capacity issues especially during weekdays correlating with working hours. Have tried Qwen2.5 coder 32B and it's very good and close to Claude 3.5 in my coding cases.

AWS is achieving 2 objectives:

1/ Best-in-class LLM in Bedrock. This could be done w/o the partnership as well.

2/ Evolving Tranium and Inferential as worthy competitors for large scale training and inference. They have thousands of large-scale customers, and as the adoption grows, the investment will pay for itself.

I've been playing with Alibaba's Qwen 2.5 model and I've had it claim to be Claude. (Though it usually claims to be Llama, and it seems to think it's a literal llama, i.e. it identifies as an animal, "among other things".)

related coverage

- https://www.anthropic.com/news/anthropic-amazon-trainium

- https://www.aboutamazon.com/news/aws/amazon-invests-addition...

- https://techcrunch.com/2024/11/22/anthropic-raises-an-additi...

Can someone with familiarity in rounds close to this size speak to their terms?

For instance: i imagine a significant part of this will be “paid” as AWS credits and is not going to be reflected as a balance in a bank account transfer.

I' m not sure how they make it back. The guardrails in place are extremely strict. The only people who seem to use it are a subset of developers who are unhappy with OpenAI. With Bard popping up free everywhere taking away much of the general user crowd and OpenAI offering the mini model always free and limited image generation / expensive model. Then you have to do it yourself crowd with llama. What is their target market? Governments? Amazon companies?There free their offers 10 queries and half of them need to be used to get around filters I don't see this positioned well for general customers.

So we have

Microsoft -> OpenAI (& Inflection AI) Google -> Gemini (and a bit of Anthropic) Amazon -> Anthropic Meta -> Llama

Is big tech good for the startup ecosystem, or are they monopolies eating everything (or both?). To be fair to Google and Meta they came up with a lot of the stuff in the first place, and aren't just buying the competition.

Rival? They kick you out after a few messages and ask you to come back later. Gpt doesn't do that

Some of these investments sound big in absolute terms.. However not that big considering the scale of the investor AND that many of these investors are also vendors.

MSFT/AMZN/NVDA investing in AI firms that then use their clouds/chips/whatever is an interesting circular investment.

I often hear people praise Claude as being better than chatGPT, but I’ve given both a shot and much prefer chatGPT.

Is there something I’m missing here? I use chatGPT for a variety of things but mainly coding and I feel subjectively that chatGPT is still better for the job.

How much of that is converted back into AWS credits?

I tried out Claude once, I found it was alright but not much better than ChatGPT for what I was doing at that point

Then I thought I'd try it again recently, I went onto the site and apparently I'm banned. I don't even remember what I did...

I must be missing it. How is anthropic worth so much when open source is closing in so fast? What value will anthropic have if competitors can be mostly free?

How many employees will lose their lifelihood to pay for this again?

Does anyone know how they are going to make money and turn a profit one day?

As someone who doesn't really follow the LLM space closely, I have been consistently turning to Anthropic when I want to use an LLM (usually to work through coding problems)

Beside Sonnet impressing me, I like Anthropic because there's less of an "icky" factor compared to OpenAI or even Google. I don't know how much better Anthropic actually is, but I don't think I'm the only one who chooses based on my perception of the company's values and social responsibility.

its a better experience, prints out token responses faster, and doesn't randomly 'disconnect' or whatever ChatGPT does

I hope they're also cooking up some cool features and can handle capacity

Gotta protect those H100s from rusting

> Amazon does not have a seat on Anthropic’s board.

Insane

Great move. The value to easily deploying content, code, anything digital to AWS is immense.

what does this say about their internal teams working on the same thing?

I much prefer Claude over ChatGPT, based on my experience using both extensively. Claude understands me significantly better and seems to "know" my intentions with much greater ease. For example, when I request the full file, it provides it without any issues or unnecessary reiterations (ChatGPT fails after me repeatedly instructing it to), often confirming my request with a brief summary beforehand, but nothing more. Additionally, Claude frequently asks clarifying questions to better understand my goals, something I have noticed ChatGPT never did. I have found it quite amazing that it does that.

So... as long as this money helps them improve their LLM even more, I am all up for it.

My main issue is quickly being rate-limited in relatively long chats, making me wait 4 hours despite having a subscription for Pro. Recently I have noticed some other related issues, too. More money could help with these issues, too.

To the developers: keep up the excellent work and may you continue striving for improvement. I feel like ChatGPT is worse now than it was half a year ago, I hope this will not happen to Claude.

Mmm. Amazon lays off thousands of workers but drops 4Bil$ into another company. Mmm.

Is this really a $4B investment, or credits on AWS?

AWS margins are close to 40%, so the real cost of this "investment" would be way less than the press release.

AWS is just playing copycat with msft. They rarely have any good original ideas. Other than IaaS and online retail.

I know that if Nvidia did this lots of people on twitter would be screaming about fraud and self-dealing.

If they are over capacity, does that mean they have significant revenue?

McAfee like investing

Is there an implied valuation? Or not enough details released?

They certainly need the money, not sure how many users will pay for monthly subscription.

Claude pro is a joke. Limited messaging and token lengths

Claude Sonnet 3.5 is simply amazing. No matter how much I used it I continue to be amazed at what it can produce.

I recently asked it what the flow of data is when two vNICs on the same host send data to each other and it produced a very detailed answer complete with a really nice diagram. I then asked what langue the diagram uses and it said Mermaid. So I then asked it to produce some example L1,2,3 diagrams for computer networks and it did just that. So it then asked it to produce Python code using PyATS to run show commands on Cisco switches and routers and use the data to produce Mermaid network diagrams for layers 1,2, and 3 and it just spit out working Python code. This is a relatively obscure task with a specific library no one outside of Networking knows about integrating with a diagram generator. And it fully understands the difference in network layers. Just astonishing. And now it can write and run Javascript apps. The only feature I really want is for it to be able to run generated Python code to see if it has any errors and automatically fix them.

If progress on LLMs doesn't stall they will be truly amazing in just 10 years. And probably consuming 5% of global electricity.