Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
Together AI raises a $102.5M Series A (together.ai)
70 points by bratao on Nov 29, 2023 | hide | past | favorite | 24 comments


The growth of together.ai makes a lot of sense given that 99% of current AI infrastructure spend is in training (vs inference). As far as navigating the product maze, this is where you want to be in AI right now.

The training revenue stream can sustain the company over the next couple of years as it develops other product lines. They get 2 big bonuses as part of this too: 1. training/finetuning llms for a bunch of companies exposes them to the big problems in this area that customers are willing to spend $ on 2. they're building a big distribution network with AI buyers

This is defensible too. They have a horde of big customers/spenders that are likely incapable of undertaking these efforts on their own and have very high switching costs. Something that you can not say about the big llm providers. Together is sort of flying under the radar serving a great market segment/need while there is a lot of expensive battles being fought everywhere else in the AI space.

This is definitely a company to watch. I wouldn't be surprised to see together.ai becoming a top 3 player in this space.


> I wouldn't be surprised to see together.ai becoming a top 3 player in this space.

I'd like some more detail on how they differ from the other 50 or so inference providers currently out there.

I always figured inference was something GCP/AWS would eventually get into their platforms


Even though they offer inference, training is their primary focus.

Inference hasn't really picked up revenue-wise (across the space) comparing to training and it's not a great market to be in. As you mentioned, it's crowded and the barriers to entry are minimal. Anyone with experience in spinning up containers and scaling them can offer this servicer. Paradoxically, it's also the market where the big cloud providers are very well positioned to dominate. Spiky and unpredictable workloads is where their bread and butter is. Their whole economic and infrastructural model is pretty much tailored to this traffic pattern.

Training is a totally different ball game. It is a model that is disruptive to big cloud providers given that it follows very different traffic patterns. Training LLMs involves spinning up 100-1000s of machines for a relatively short period of time and with interconnect that doesn't typically exist in data centers. That is a very unique workload. Additionally you need significantly more specialized ML knowledge in tensor parallelism, optimizations, CUDA etc. That is not as common as scaling a container based workload..

Fun fact: Oracle is surprisingly well positioned in terms of their interconnect fabric. Even Microsoft is partnering with CoreWeave for GPU clusters because they dont have as much capacity interconnected in the right way.


This makes sense, but I find it hard to believe the big cloud players won’t have the datacenter skills to compete…

I agree that supply is an issue, but paradoxically the fact that these GPU cloud providers (CoreWeave et al) are partnering with the big cloud players says that the big cloud players are where people would prefer to buy. Once supply constraints are solved, these providers would need some novel offering beyond “we have hardware”, e.g. some specialized distributed training framework. But MS/Google/AWS are also building their frameworks so…

And then the elephant in the room is: compute spend so imbalanced on training vs inference. Why? Is it that there arent enough real use cases? Is it that improvements are so frequent it makes sense to toss out older versions? Is it that privately trained models are a requirement for the highest spenders? My impression is that a lot of corporate spend at scaleups is purely speculative r&d to evaluate capabilities but thats a small sample from friends


The interconnect comment here is spot on.


People keep saying that training is tricky, which is true, you need people with experience, but they're not so rare that this will present a moat forever since you can very much hire the people who know how to do this.

My experience with ML projects is that while there is churn in the modeling, most of the effort for a long lived system still goes into the data, but engineers really want to work on the modeling/infra pieces much more than data quality.

Which is to say, I have a lot of skepticism that this is a long term business.


I wonder how NVIDIA gets to report their investment here from an accounting perspective?

Could they book the money TogetherAI spends on GPUs as revenue now, and then when TogetherAI flops and returns five cents on the investment dollar write it off in a different line of business?


Nvidia spending money on nvidia hardware is less annoying than the model of raising money from VCs to give to nvidia. It's an interesting question though - does it still count as a high risk investment worthy of tax advantages if lots of the investment is spent on the hardware the invester sells?


There are well-defined accounting rules for that, which also applies for joint-ventures, holdings and any holding structure.

Many examples, but this was the first relevant link on DDG:

https://www2.deloitte.com/us/en/pages/audit/articles/a-roadm...


NVIDIA probably recognizes revenue upon delivery of product. The investment write down would probably impact their P&L and have some associated tax benefits there.


Likewise Microsoft's $10 billion OpenAI investment has a large (majority?) portion in Azure cloud credits. Microsoft has around a 70% margin at Azure.


> Microsoft has around a 70% margin at Azure.

I find this impossible to believe, given they are far from a monopoly here. Where are you getting the figure?


72% gross margin on Azure: https://www.microsoft.com/en-us/investor/earnings/fy-2023-q2...

None of the cloud providers are monopolies, but there are few enough they can collude through tit for tat, though at times I think some were run for growth at a loss.

I believe all have multi thousand percent margins on egress bandwidth.


I don’t find it surprising that SaaS list price margins could be that high. The big purchasers do volume discounts to get massive discounts while the smaller players are forced to pay more.


If MSFT and AMZN are making a 70% margin then why wouldn't google accept a 30% margin and eat their lunch?


No because the overhead and uncertainty involved with smaller customers means they cost more and are less loyal / locked in and thus represent more risk. The smaller customers represent very little revenue in the short term and if one grows to any real size negotiations with sales begins.


the price is pretty low: https://www.together.ai/pricing

4B MODEL, PRICE 1K TOKENS: $0.0001

register with an email, test account has 25$ credit, python API as well, the model is smaller, but good to have some fun with the API integrated with other systems, haha...


Not really fair to compare a 4B model and a 1.7T model.

Per flop the 4B model here is slightly more expensive than GPT4.


The theory that GPT-4 is 1.7T also posits that GPT-4 is composed of eight 220B experts, meaning once you've loaded the model, inference costs aren't akin to a 1.7T model, but instead to a 220B model.

If prices scale linearly, we reach 0.03 / 1K tokens (lower end of GPT-4's price range) at about 0.0001 * 300, or about 1.2T parameters (and that's dense - no MoE here).


I guess the question is how the mix of experts works. Do they predict one token from each model? If so you’re still doing 1.7T worth of computation.


clause 2.1 is $8/million tokens for prompt or .008/1k tokens. interesting


Anybody using them got feedback?


I've explored most of the inference as a service platforms, theirs is particularly nice. (Clear/simple UI, good pricing, good speed.)


So this is where all the VC money in FinTech has gone to.




Consider applying for YC's Fall 2026 batch! Applications are open till July 27.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: