Decoding Cohere: Unraveling the $2.1B AI Startup's Breakthroughs

0:00Breaking news out of AI land, Cohere, which is an AI startup that is now valued at over$2billion has just raised a round of$270 million. Today on the podcast, we're going to break down what they plan on using that money for, who is giving them that money, and what it is that makes Cohere so valuable. I think overall, this is just a sign that there's still a lot of money to go around in generative AI. But what does Cohere actually do? So Cohere is developing an AI model ecosystem for corporations and the enterprise. today they announced that this 270 million dollars they raised is part of an official series c round so routers actually reported earlier this year that cohere was in talks to raise you know quote-unquote hundreds of millions of dollars and they actually quoted that their valuation was going to be over six billion dollars so if there's any um i guess evidence that that reporting was accurate cohere appears to have missed that valuation mark substantially because they are now apparently, according to TechCrunch, they've raised their money at between 2.1 and 2.2 billion dollars.

1:10Now, that doesn't necessarily mean that valuations have dropped. Potentially, that just means that a previous, you know, deal fell through or that it was an unrealistic expectation. And I actually believe I heard that they were at one point in talks with Google to raise 200 million dollars from them. And perhaps they had a higher valuation, but it would have come at the cost of, you know, them being forced to use Google Cloud to run their all of their AI tools. And so sometimes you'll see deals like that where the valuations may be a little inflated because companies like Amazon, Microsoft or Google really want you to use their cloud platform and they know they're going to make a lot of that money back.

1:49So that's just a little caveat on that. In any case, the COO Martin Kahn, he was quoted as saying the new capital will fuel continued development of Cohere's AI platform, which is focused on enterprise customers, allowing companies to use their preferred cloud provider to increase data privacy and make implementation simpler. The latest round allows us to invest in compute, grow our team, engage with more of the world's leading enterprises and further advance our world leading AI, ultimately empowering companies to build incredible products while keeping their data private and secure. So I think they're doing a couple things that are kind of impressive.

2:29Number one, they're allowing you they're essentially they're cloud agnostic. So you can switch between any cloud provider, you can take their AI models, and you can host them on Google, Microsoft Azure, or AWS. And this makes it really powerful because by not being locked down to one particular provider, you're able to move around if you're getting deals or discounts from a specific cloud platform, you can move around where a lot of these guys, Google Cloud in particular, is going to try to lock you in on their platform and make it really hard for you to move to another platform once you're kind of locked in.

3:03So I think that this is a smart idea on their end, especially because they're building tools that enterprises will need. So they really need to have that flexibility. Enterprises use a lot of different cloud providers, and it would, I think, stunt the growth of the company for one, saying, look, if you want to use us, you have to be on AWS. And so that just offers a lot more flexibility. Aiden Gomez, Yvonne Zhang, and Nick Frost, who was one of the first employees at Google's Toronto AI Labs, all co-founded Cohere back in 2019. And before starting Cohere, Gomez co-authored the paper, Attention is All You Need, which essentially that was the paper written by the Google researchers that introduced the transformer model, which is what OpenAI uses to train chat GPT.

3:49So it's really kind of the architecture behind what large language models are today. And so obviously, Cohere has, you know, some of the top scientists and researchers in this AI space that are on their founding team, which offers them a lot of credibility. They've been doing this since 2019. And I think that's the reason why today, right, this isn't just an AI startup that popped up out of nowhere, and all of a sudden could raise 270 million. This is one where, you know, this team made this bet back in 2019, and they've been focused on this. And now that, you know, this whole AI boom is happening, they're perfectly positioned.

4:23And they're, you know, some of the leading minds in this space to really take advantage of this. Khan actually joined in early 2023. So his previous role was as the CFO at YouTube. And so, you know, obviously, they're bringing in some really big players, bringing in people from, you know, high levels at YouTube, bringing in some people that were high level researchers at Google. Cohere, which essentially has developed multilingual language models trained on data from native speakers, among other AI, really aims to kind of stand out in this, you know, really rapidly expanding field of AI startups by focusing on the enterprise use case.

5:04So I guess that's kind of like their niche, that's kind of their angle, is they're really focusing on the enterprise use case. their platform as we mentioned earlier is cloud agnostic which allows them to better serve that and they take a really hands-on approach when working with their corporate clients to help corporations create custom llms based on their proprietary data so they really go in there they help these corporations they take their data and so it's you know beyond just a software solution they're also kind of giving these companies the the manpower that's incredibly expensive and hard to come by right now to figure out how to build these things out.

5:43Khan who just joined recently from being the YouTube CFO said Cohere was founded to create a platform that empowers all enterprises to transform their company and products with world leading AI that's cloud agnostic accessible customizable and data secure. Our mission is to allow enterprises worldwide to leverage this transformational technology. So really, I think one thing that is interesting that I would love more data on, but it's really hard to get is the customer number. So Cohere has not released any numbers on that. But they do claim to be working with a number of startups on their website, it says they work with HyperWrite, Spotify, Longshot, Jasper, Helvia, Bamboo, HR, Glean, amongst some others.

6:27So they obviously have some big companies that they're currently working with. And I think that this just kind of lends to the fact that they're kind of an all in one solution to really help these enterprises get started with LLMs. And because I think one of their biggest, I believe one of their biggest likes selling points right now is the fact that they're going to they're willing and they're able to help these enterprises get this set up. So for enterprises that may not have the budget, or perhaps don't even know what they need to do to get started building out these LLMs, and it's really expensive and hard to find these AI data scientists, Cohere is able to help them integrate that.

7:03And I think you can see that by the fact that in this most recent round, Cohere really partnered with a whole bunch of different firms and software companies, including Salesforce Ventures, and which they say they're partnering with them to help break down and analyze and summarize lengthy text using machine learning algorithms. So Khan said, generally, we can share what works that we're experiencing high demand for major enterprises, both customers and partners. And they currently have around 180 employees. Now, this is the part that I think is incredibly interesting. Cohere to date has raised is in the third place for the AI company that has raised the most amount of money and they are not far behind number two so obviously OpenEye is number one having raised 11.3 billion dollars most of that coming from Microsoft and Anthropic which is you know known as number two they recently closed around with Google they've raised 450 but Cohere to date has raised 445 million, only 5 million short of what Anthropic has done.

8:11And that is quite a bit ahead of some of their other rivals, including inflection AI that's at 225 million and adept that is at 415 million. So cohere takes the cake takes the number three spot the bronze for the most money so far raised by the one of these AI startups. And just in case you know, you're worried that this is seeming really excessive. Training these AI models tends to be quite capital intensive, especially when we're when they're building them out to be so big and so robust, especially as we're kind of building some of these. And you know, these these companies, we can see exactly what they're raising and what they're spending, you know, Anthropik and OpenAI and Cohere.

8:54But the ones that we don't really get to hear about as much is like, what did Google spend on Google barred. Not just training, but like all of the employees that helped put that thing together. I would bet that it is an astronomically large amount of money, perhaps probably rivaling what OpenAI has ever spent in their, you know, the 11 billion they've raised. I would assume Google has well is able to outspend them. And it does cost more because of just the organization they have to support that supports them. So I think that, you know, raising money, especially when Cohere is going up against, you know, the giants like OpenAI and Google, they really do need to raise this fresh capital to stay relevant and to have the manpower to continue to do this.

9:33Khan said that when it comes to AI, building and training language models requires a lot of capital, but we're mindful and intentional about what we actually need and committed to securing funds that will ensure we can continue providing the best possible solution to our customers. He also said that they've been intentional about maintaining a diversity of global investors and not taking one big check from one company, especially a cloud provider, which may limit our ability to remain independent. And I think what they're really saying there by saying, you know, especially not a cloud provider, I think he's kind of taking a dig and poking fun at OpenAI and Anthropic.

10:10OpenAI obviously taking a lot of money from Microsoft and picking them as their default cloud provider. And Anthropic just freshly taking$200 million from Google and using Google cloud as their main, you know, cloud provider. So it's interesting to see if they'll be able to stay, you know, cloud agnostic and stay independent. Obviously, that is kind of the direction they're going, even though I do believe someone recently reported in the Wall Street Journal that they were, you know, in talks with Google about getting a 200 million dollar, you know, investment, and evidently they were able to get it from other places.

10:45Perhaps it's just a concept they entertained, you know, it'd be better to take it than let their company go bankrupt. But obviously, it's kind of not what their company's core value or focus is. So they were able to avoid that. Regardless of whether, you know, or where their future capital is going to be coming from, Khan says that he sees search and retrieval as one of the next kind of core areas of growth that cohere. And this is really interesting because Google just released Google app, the Google app builder, the Google AI app builder. And part of that and the biggest part of that is actually just the ability for enterprises to create really robust search functionality.

11:28They Google just recently as of yesterday, I believe partnered with a major hospital Mayo Clinic. And so I think this is apparently where they see the biggest growth coming from search and retrieval, but I'm not actually sure if they're going to be able to get there if they're not already there. Google seems to have the search industry completely nailed down their AI and algorithms. They've been doing that for, you know, the last 10 years or more, the last 20 years, that's what they do. And so I'm not sure if they're going to be able to compete there, but that's apparently where they said they're going to be spending the most amount of money in the future and the most amount of resources where they where they think they'll see the most growth um they did say that today chatbots don't have access to the world uh they don't know about what happened 10 minutes ago and you know they're they're saying that they're going to be able to help you to cite sources um so that users don't have to blindly trust a model and and all this stuff um and that also um you know we're seeing that happen with uh chat gpt once you do the web browsing it's able to cite sources google bart is able to cite sources so once again i'm not sure if they're like they're not going to be the first ones to be able to do either of those two things that they're talking about.

12:35So I'm curious to see kind of what their next plays are. Looking even further ahead, though, than those kind of features that already we're kind of starting to see, Cohere does plan on building models that can take action to actually do work for customers, like book a flight, schedule a meeting, or file an expense report on a person's behalf. So in that way, I think that it's kind of chasing after some of its other competitors like Adept, Inflection, and OpenAI, all of which are currently building and I and you know I'll say they're all taking very different approaches to this but it would seem they're all building different systems to connect AI with third-party apps and services and products and you will see this with OpenAI doing their plugins and Adept Inflection doing this in other ways so I think despite the competition Khan really thinks that Cohere is in a position of strength he said we're differentiated as the independent cloud agnostic AI platform for enterprise.

13:25We are solely focused on enabling our customers to create proprietary LLMs capable of leveraging data and creating strategic differentiations and business value. I think it's going to be really impressive to see. I wanted to quickly list off who invested in this round because I do think that it is a little bit telling and is quite interesting. So Innova Capital led this round, which by the way, was oversubscribed. And they had participation from Nvidia, Oracle, Salesforce Ventures, DTCP, Miri Assets, Schroeder's Capital, Sentinel One, Thomvest Ventures, and Index Ventures, who I believe invested in a previous round.

14:03So in any case, it is interesting, right? We're seeing people like Nvidia, Oracle, Salesforce, and a handful of others that all of those companies are obviously, you know, doing deals here to, or not, you know, they potentially could be doing deals to help within their own company, use this within their own company. And people like Nvidia, obviously, they're just going to benefit overall when new AI companies are coming up and trying to compete. I think Nvidia in particular knows that the more competition in this space, the more these companies are gonna have to spend on computational power to make their models better and better.

14:38So I think this is a genius move by Nvidia to back someone like this, that isn't necessarily the front runner. And then we see companies like Oracle, maybe they're making a play to kind of get back at Microsoft. Salesforce kind of talked about some ways that they were already integrating it. So I think it's gonna be really interesting to see. But I think this is like definitely one of the top five AI companies to keep your eye on. Because right now they have raised the third most amount of money of any AI company. They have one of the biggest AI pioneers that helped write the transformer paper.

15:09So I definitely think cohere is one you're going to want to keep your eyes on into the future.