E12: Inside VC Deals - Elon Musk's xAI ($6b), OpenAI Fund ($15m), Perplexity ($250m) + 4 more

I'm Matt Joseph. And today is a very special day on the pod because today we are diving into the top funding rounds of artificial intelligence startups over the last month.

Now, anyone who has been watching AI over the last couple of years knows that the amounts of money that these companies have been raising is ridiculous.

And some of the startups we're going to look at today, continue that trend. But I do think that there is a significant difference between where we were a year ago and where we are today.

One of the top investors in the space is Sequoia capital Sequoia is thinking of this in two acts of generative AI, they have act one and act two. Here's what they say about it.

Act one came from the technology out. We discovered a new hammer, foundational models, and unleashed a wave of novelty apps that were lightweight demonstrations of cool new technology. Act two will solve human problems end to end. They tend to use foundation models as a piece of a more comprehensive solution rather than the entire solution.

examples of companies entering act two include Harvey, which is building custom LLMs for elite law firms; Glean, which is crawling and indexing our workspaces to make generative AI more relevant at work; and Character and Ava which are creating digital companions.

Sequoia is telling us that the technology in generative AI is really cool, but it has to be more useful. Now, that's not to say that some of these startups aren't making a ton of money.

Here's what they say.

Make no mistake. Generative AI has already had a more successful start than SaaS, software as a service, with greater than a billion dollars in revenue from startups alone. It took the SaaS market years, not months, to reach the same scale.

SaaS is led by companies like Salesforce. Microsoft. but there's a ton of other SaaS companies out there. When they call out a billion dollars in revenue, they're not talking about users. They're talking about actual money that people are paying to use it.

That doesn't mean it necessarily has staying power, but it does mean that these are actually good businesses that these AI startups at the top of the market are creating.

So now that we have that context, let's actually jump into the startups that raised money this month. First up, xAI has raised 6 billion at an 18 billion pre money valuation.

Now, xAI is Elon Musk's competitor to OpenAI. Elon Musk was actually a cofounder of OpenAI, so this story is especially juicy. xAI actually emerged out of Twitter.

Now, you'll know if you've been following the news at all, that Elon renamed Twitter to X and xAI is the AI platform trained on all of Twitter's data.

Now, the source in the TechCrunch article said,

We all received an email that basically said, It's now $6B at an $18B pre money valuation. And don't complain because a lot of people want in. Presumably, that came from Elon himself.

This is a big deal for a lot of reasons, and mainly because of the way that Elon plans to train xAI's large language model, here's what the article says.

xAI's marketing literature already makes clear that the outfit's ambition is to connect the digital and physical worlds, but it may not be widely understood that Musk plans to do this by pulling in training data from each of his companies, which includes Tesla, SpaceX, his tunneling outfit, boring company, and Neuralink, which develops computer interfaces that can be implanted into brains.

Of course, another of Musk's companies is X. The social media platform has already incorporated xAI's months old chatbot Grok into the platform as a paid add on. Yet that's just one piece of what Musk tells investors will become a sprawling virtual cycle.

So xAI is not just going to be building a chat bot. They're actually incorporating data to be able to work in the physical world. They're going to bring in data from Tesla and SpaceX and all these other companies and try to actually build an AI that can do both digital and physical tasks. Now, there was still a lot of questions and speculation about what exactly this money is going to be for.

There was an interesting quote from a Fortune article that talked about this. It said,

The giant sum is a clue that Musk could be planning to go full stack. That is, to control all of the elements of an AI system from the LLM software to the computing power hardware necessary. Instead of relying on Alphabet, Amazon, or Microsoft data centers, Musk could build, own, and operate his own data center with 6 billion, at least for a few years.

It's too small a price for chip production, but too big of a price for an LLM training run. He said, you can do a lot of training with 6 billion, but it's not enough to have your own chip foundry.

So all that says that it's not quite clear what Elon's going to do with this 6 billion. But what is clear is that you have to pay attention to this startup because you know that with Elon involved, it's going to do something big.

Up next, Perplexity is raising over 250 million dollars At a two and a half to $3 billion valuation for its AI search platform. Perplexity is a prosumer product. It's essentially a paid search product that competes with Google search. Now the difference for Perplexity is that it isn't bound by the same constraints as Google. Google makes its money from what some people call 10 blue links.

They put a bunch of sponsored search results all over the page and when people click on those results, that's how they make money.

Perplexity, because it's a paid product, doesn't have to do any of those things. And so what you end up with is a much richer analysis based on the question that you ask it. Now, Perplexity is operating in a very crowded space, which includes Google search and its own product, Gemini.

here's what the article has to say about that.

Unlike other enterprise tools for knowledge work Perplexity enterprise pro is also the only enterprise AI offering that offers all the cutting edge foundation models in the market in one single product.

That's open AI, GPT four, anthropic, Claude Opus, Mistral, and more to come. Now it's interesting when you evaluate Perplexity to return to that idea that these new AI companies are actually making a lot of money.

Perplexity is making some money, but it's hard to justify a 3 billion valuation, even with the money that they're making. So here's what the article says.

In the case of Perplexity, the startup offers its tools on a free and enterprise paid tiers. And so far it has processed 75 million queries this year and is currently on ARR north of $20 million according to Bloomberg.

$20 million might be worth a billion dollar valuation depending on the growth. It's hard to justify a $3 billion valuation at that amount using public market comps. Many investors think the Perplexity is going to displace Google Search.

That because it's not constrained the same way that Google is, ultimately consumers will wanna switch.

I have doubts about that because ads are essentially the best business model in the history of technology.

It's hard to imagine how Perplexity would become a huge company if it didn't actually offer ads of any kind. So maybe that's in the works. But unless they do that, it's hard to see them making the kind of money that Google makes and also trading at the same valuation as a public company.

That said, if you've used this product before, it's pretty great.

And I highly recommend it just to get a sense of how these new age products are working. It's definitely more useful than a lot of the products that have come before it and maybe it's even more useful than Google search is.

Up next, Cognition Labs has raised 175 million from Peter Thiel's Founders Fund, and it's only six months old.

Cognition made a big splash earlier this year when it released Devin, which was an AI software engineer that went viral. It went viral for good reasons and for bad reasons. The good reason was that the flashy demo that they released made people believe that you could already replace software engineers with artificial intelligence. The bad reason is that once software engineers dove into how they came up with this demo, they realized that it didn't work the way that the company said it did.

So let's take a look at that announcement video from Scott Wu, who's the CEO of Devin.

Hey, I'm Scott from Cognition AI. And today I'm really excited to introduce you to Devin, the first AI software engineer. Let me show you an example of Devin in action. I'm going to ask devin to benchmark the performance of Llama on a couple of different API providers. From now on, Devin is in the driver's seat. First, Devin makes a step by step plan of how to tackle the problem.

After that it builds a whole project using all the same tools that a human software engineer would use.

Devin has its own command line, its own code editor, and even its own browser. In this case, Devin decides to use the browser to pull up API documentation, so that it can read up and learn how to plug into each of these APIs.

Here, devin runs into an unexpected error. Devin actually decides to add a debugging print statement, reruns the code with the debugging print statement, and then uses the error in the logs to figure out how to fix the bug.

This demo set the software engineering world on fire. why don't we dive into the backstory of the CEO, Scott Wu? Cause I think that's just a really interesting story.

So Scott was a child prodigy. He won a bunch of math competitions and coding competitions as a kid. He went to Harvard and he partnered with a bunch of other coding competition legends to build this.

And this from the article,

Scott believes that the proficient background of his team gives the startup an edge over its competitors. It's almost like this game that we've all been playing in our minds for years now, and now there's this chance to code it and turn it into an AI system, he said.

But we need to dive into what some of the criticism is saying about it because actual software engineers don't believe this system works well at all. Here's a software engineer who actually dove into the example that Devin showed and demonstrated that it actually didn't work the way that they said it did.

This is the Internet of Bugs.

It's a lie. So this video is in three parts. First, we're going to talk about what should have been done,

actually did, and how did it, and 35 years. I am not anti AI, I really am anti hype, that's why I'm doing this. Devon was introed not quite a month ago now, and it was touted as the world's first AI software engineer, and I don't believe that it's the first software engineer, and I already made a video about that, I'll put links in the description.

But today is about the specific claim that's the first line of the video description, which says, Watch Devon make money taking on messy upwork tests. That statement is a lie. You cannot watch that in the video. It does not happen in the video. It does not happen.

What's worse though is that the hype and the fear, uncertainty, and doubt from people repeating and embellishing on that claim because they're trying to get clicks or they're trying to go viral or they just want to be part of the Zeitgeist.

The hype around Devon in general is just crazy and that statement seems to be what a lot of it is pinned on. For the record, personally, I think generative AI is cool. I use GitHub Copilot on a regular basis. I use

So long story short, what Carl and these other engineers are saying is that Devin didn't actually do the task that it was supposed to do, and it did it in a way that was incredibly inefficient. This really highlights the limitations of the hype cycle around AI, that there's some really cool ideas and there's some smart technology, but these tools don't actually do what they promise they can do just yet. But let's be clear with 175 million, they're going to have a lot of time to figure out exactly how to build this and they seem to have a great team of people who are working on it.

Up next CoreWeave has raised 1. 1 billion at a 19 billion valuation.

The New Jersey based company rents out chips housed in data centers across the US that customers use to create and deploy AI systems. It's part of a new crop of companies offering cloud computing tailored for AI, setting it apart from big tech companies, such as Microsoft and Amazon, which offer a wider range of cloud services.

So basically CoreWeave is in the infrastructure stack and they are a way for developers to get access to compute without actually having to build data centers. It's just like AWS, except it's very focused on AI. And that just means that it's data centers are optimized for AI use cases as opposed to being optimized for general data use cases.

CoreWeave was founded by this guy, Michael Intrator. He's actually a hedge fund manager who pivoted into AI when he saw the boom happening. He wasn't even an investor primarily in AI before this. Here's what his LinkedIn says.Intrator was s a co founder and CEO of Hudson Ridge Asset Management.

Hudson Ridge was a fundamental systematic natural gas hedge fund.

he really timed his approach to the market well, and he partnered with some people who were really good to get his company off the ground. Here's what he had to say about the latest round.

This new funding will propel us forward and allow us to continue investing in product offerings to meet the explosive demand we're seeing for our specialized cloud infrastructure.

Part of what makes this company interesting is how it was founded There's kind of a running joke amongst investors that all of the crypto guys pivoted into AI. But this is actually a case where that genuinely happened.

CoreWeave was founded in 2017 as Atlantic Crypto, an Ethereum mining company that bought NVIDIA graphics processing units, that's GPUs, both to mine its own crypto and rent out GPU servers to other crypto miners. Early in 2019, Atlantic Crypto changed its name to CoreWeave and pivoted to providing GPUs on demand for generalized computing purposes rather than focusing on crypto.

So here we have an example of a company which was doing crypto, realized that crypto was crashing and got out of the game.

But that move actually turned out to be very prescient for them because of the growth that it gave them. Sacra estimates that CoreWeave hit 465 million in revenue in 2023. Up 1760 percent or about 19 X from 25 million in 2022 off the rapid acceleration of demand for GPU compute from cloud providers, LLM companies, and every app looking to integrate generative AI features.

CoreWeave projects 2. 3 billion of revenue in 2024 with signed cloud contracts in excess of 7 billion through 2026 up from 5 billion in early 2023.

So CoreWeave is making real money and they expect to make even bigger amounts in the years to come. Now, one interesting note from their revenue line,

a significant portion of that 7 billion in contracts came from Microsoft, which agreed on a multi year deal for CoreWeave to supply it with GPU compute amid rising demand from Azure cloud customers that Microsoft has not been able to meet.

So what we're seeing here is that even Microsoft, which has its own cloud, cannot meet the demand that it has for GPUs and that's a huge opportunity for CoreWeave.

The thing that I would be concerned about if I were CoreWeave is whether this demand is actually durable. They're raising money as if the demand is going to continue to increase, but as we know, much of this demand is actually coming from startups, which are using their venture capital money to build their products. If those startups prove to not be durable and not last, you might see a crash in the market.

This didn't matter for the cloud providers of the past era. That's companies like Microsoft and AWS, because this wasn't their core business model. But for a company like CoreWeave, this is all they do and they might actually be in trouble if the demand doesn't stand up the test of time.

Up next SiMa AI has raised 70 million to produce a multimodal gen AI chip. Here's what the article had to say.

According to Gartner, the market for AI supporting chips globally is forecast to more than double by 2027 to 119 billion compared to 2023.

However, only a few players have started producing dedicated semiconductors for AI applications. Most of the prominent contenders initially focused on supporting AI in the cloud. Nonetheless, various reports predicted a significant growth in the market of AI on the edge, which means the hardware processing AI computations are closer to the data gathering source than in a centralized cloud.

SiMa, named after seema, the Hindi word for boundary, Strives to leverage this shift by offering its edge AI SOC to organizations across industrial manufacturing, retail, aerospace, defense, agriculture, and healthcare sectors.

The general claim here, the idea of a multimodal chip, is just that the chip can interpret multiple types of information.

So you could think sound, text, images, videos. It can process all of that right there on the chip. And the edge component is that it'll process information directly at the source as opposed to being in the cloud.

So you could imagine a facility like a factory that just wants to process all of its information right there in the factory and not send information up into the cloud.

That's really where SiMa AI would come in.

The argument there is going to be around security. You don't want your data sitting up in the cloud with Amazon or Microsoft or some other firm because you don't know what's happening to it when it's up in the cloud. So if you have a very secure AI application, you'd probably want to use a chip like Sima's to facilitate it.

Now SiMa was founded by this guy, Krishna Rangasayee. Krishna has an interesting backstory because he was the COO of a company called Groq, which is not the same Grok as Elon's chatbot competitor to ChatGPT, but did work on machine learning applications. He started his career as an engineer at a company got acquired by Intel, and then he shifted over into executive roles at Zillinix, where he led things like sales teams. Zillinix was a leader in the semiconductor space.

Here's Krishna talking a little bit about his journey into starting SiMa.

generally it takes a lot of money to get into chips and anytime you're competing with NVIDIA, you're in a really difficult spot, but they have a manufacturing partnership with TSMC, a big Taiwan producer, and it seems like they're on the right track.

Up next, Parloa, a German conversational AI platform, has raised 66 million to expand internationally.

Now, the reason I wanted to cover Parloa is because it signals a shift in the overall market. We're seeing a rise of these international conversational AI platforms. the American platforms, while they are moving to serve international markets, are being crowded out by local competitors like Parloa.

Kate Clark, who's a venture capital reporter for the information, actually touched on this in an article this week. Here's what she said.

Khosla Ventures, an early OpenAI investor, is also backing similar LLM developers, including some located in different countries, such as Japan's Sakana AI and India's Sarvam.

The early stage firm expects LLM developers that are native to a country or geography to succeed because they'll adapt better to language and cultural differences than overseas rivals, as well as receive more local government support, said Khosla partner, John Chu.

So the trend that we're seeing is that there's going to be a multi LLM strategy that's run by AI companies that are trying to build their own applications. They're not just going to say, I'm just going to sit down with chat GPT.

They're going to say, let's use all of these models and combine them and that's really where some of these local competitors might do really well abroad.

Parloa was founded by this guy, Malte Kasub. Now Malte previously founded a conversational AI agency, and apparently he used some of the information that he learned in running that agency to start Parloa cause he was running both of those companies at the same time for a period.

I think that was probably before Parloa started raising a lot of money and becoming big.

Now, here's a statement from Malte on his LinkedIn announcing the round.

I'm incredibly excited to share big news today. We've raised 66 million in our series B funding round led by Altimeter, marking the largest series B funding round in the European contact center AI ecosystem. Behind all this is a team of now 250 Parloans of whom I could not be prouder. Together. We have tripled our revenue onboarded great customers and partners and launched countless new features.

But above all together, we have laughed, dreamed, fought, and celebrated. And for that, I can't thank you enough. You are incredible.

He's drawing an interesting distinction there, trying to narrow down the round to focus on the AI contact center ecosystem in Europe.

I don't imagine that's particularly big ecosystem, but we'll give him credit for that. Congratulations Malte on the round.

Up next, the OpenAI Startup Fund has raised 15 million to continue investing in AI focused startups.

Now, this is a really interesting one because of all of the complexity around the formation of this fund. The fund is actually not affiliated directly with OpenAI, but it kind of is.

Here's what the article says.

The fund, whose portfolio companies include legal tech startup Harvey, Ambience Healthcare, and humanoid robotics firm Figure AI, came under scrutiny last year after it was revealed that OpenAI CEO Sam Altman had long legally controlled the fund.

While marketed like a standard corporate venture arm, Altman raised capital for the OpenAI Startup Fund from outside limited partners, including Microsoft, and had the final say in the fund's investments. And here we have this very interesting and surprising nugget. Neither OpenAI nor Altman had or have a financial interest in the OpenAI Startup Fund, but critics nonetheless have argued that Altman's ownership amounted to a conflict of interest. OpenAI claimed that the general partner structure was intended to be temporary.

So this is a fascinating story because Sam Altman is known for having a lot of interests from his time as president of Y Combinator, where he invested in a ton of different startups.

I mean, for example, when Reddit went public, it was revealed that Sam owned 8 percent of the company,

which was worth, at the time, 462 million. So no one quite knows what all of his ownership interests are and there are questions about where his ownership interests influence his leadership of OpenAI.

This is just another example of him drawing lines without people really knowing about it that then caused them to question his motivations.

A few more details on the fund. As of last year, the open AI startup fund had 175 million in commitments and held 325 million in gross net asset value. That's basically saying that they own 325 million of startup shares. It has backed well over a dozen startups, including Descript, a collaborative multimedia editing platform valued at 553 million last year, language learning app, speak AI power, note taking app, Mem and IDE platform, AnySphere.

Now it's important to note that Sam no longer controls this entity.

In April, Altman transferred formal control of the open AI startup fund to Hathaway, previously an investor with a VC firm, Haystack who played a key role in managing the startup fund since 2021.

It's hard to think about the AI space without thinking about OpenAI and its founder, Sam Altman.

He's essentially the most important figure in artificial intelligence today. And everything he does is under scrutiny. The concern here is that he is looking at how much these startups are using OpenAI and then using that to determine who he invests in.

Technically it's not illegal, but it may just be a little bit unseemly and create a perception problem for that fund. I believe that's why he kept it separate. But overall, I think he's making the right move to divest this interest and clean up the lines here a little bit so it doesn't seem like there's any self dealing going on. He really doesn't need the money and he certainly doesn't need the negative press here.

So who's leading the open AI startup fund, if not for Sam Altman? Well, it's this guy,

Ian Hathaway. Now Ian was an investor at Haystack, a firm, which has done incredibly well, by the way, and now he is in charge of this fund. It's not clear what exactly his ties are to Sam Altman, but he is now completely in control of the fund.

Now that said, it's hard to imagine Sam not being involved in some capacity. I have to imagine that he and Ian are very close, especially given that the fund is focused entirely on OpenAI,

but the optics here are good for the fund. There are a lot of AI focused funds out there. This one has very unique positioning in the market, and I think they're going to continue to do really well.

And that is it. Now, I want to give a quick honorable mention here to Augment, the startup that raised 292 million

that we covered in our last video. They're building a competitor to GitHub Copilot. It's worth checking out if you want to learn more about it.

But we've got a pretty good sense of the kinds of startups that are raising big dollars in today's AI market.

There's a lot of infrastructure, chips, developer tools that are getting funded. And the space is obviously still in a hyper growth phase, although we're not seeing any more of the classic AI for this kinds of startups that we saw in the last cycle. It's not to say that won't continue and won't happen in the future, but it does seem like Sequoia has hit the nail on the head that they are now moving towards end to end applications.

And that's what we expect to see more of in the future. Thanks for watching. Thanks for listening. And we will see you again soon.