AI startup investment is on pace for a record year

The startup investing market is crowded, expensive and rapid-fire today as venture capitalists work to preempt one another, hoping to deploy funds into hot companies before their competitors. The AI startup market may be even hotter than the average technology niche.

This should not surprise.

In the wake of the Microsoft-Nuance deal, The Exchange reported that it would be reasonable to anticipate an even more active and competitive market for AI-powered startups. Our thesis was that after Redmond dropped nearly $20 billion for the AI company, investors would have a fresh incentive to invest in upstarts with an AI focus or strong AI component; exits, especially large transactions, have a way of spurring investor interest in related companies.

That expectation is coming true. Investors The Exchange reached out to in recent days reported a fierce market for AI startups.

The Exchange explores startups, markets and money. 

Read it every morning on Extra Crunch or get The Exchange newsletter every Saturday.

But don’t presume that investors are simply falling over one another to fund companies betting on a future that may or may not arrive. Per a Signal AI survey of 1,000 C-level executives, nearly 92% thought that companies should lean on AI to improve their decision-making processes. And 79% of respondents said that companies are already doing so.

The gap between the two numbers implies that there is space in the market for more corporations to learn to lean on AI-powered software solutions, while the first metric belies a huge total addressable market for startups constructing software built on a foundation of artificial intelligence.

Now deep in the second quarter, we’re diving back into the AI startup market this morning, leaning on notes from Blumberg Capital’s David Blumberg, Glasswing Ventures’ Rudina Seseri, Atomico’s Ben Blume and Jocelyn Goldfein of Zetta Venture Partners. We’ll start by looking at recent venture capital data regarding AI startups and dig into what VCs are seeing in both the U.S. and European markets before chatting about applied AI versus “core” AI — and in which context VCs might still care about the latter.

Hot, expensive, crowded

The exit market for AI startups is more than just the big Microsoft-Nuance deal. CB Insights reports that four of the largest five American tech companies have bought a dozen or more AI-focused startups to date, with Apple leading the pack with 29 such transactions.

The business data and market analysis firm also said that in the first quarter some 626 AI startup deals were struck globally. The transactions were worth $17.7 billion. For context, in all of 2020, those figures were 2,334 and $35.4 billion. We can see that the AI startup market was, through the first quarter, on pace for a modest increase over 2020 results in deal terms, roughly tying 2019’s all-time high, while in dollar terms, AI startups are on track to set all-time records by a huge margin.

The final deal volume numbers may wind up higher than we might expect from simply looking at Q1 2021 data, however. Glasswing Ventures’ Seseri told The Exchange that her firm is “seeing about 2x-3x more AI-related deals this year compared to last year” since the start of the second quarter. That implied acceleration in deal activity could bolster H1 2021 round counts.

But it’s not just investor interest that is driving the deal activity; Seseri also said that “demand is up as enterprises have either accelerated their planned adoption of AI products” or have finally gotten to work with their digital transformation efforts. That puts AI tech inside the mix of startup niches that are enjoying a tailwind from companies’ modernization efforts.

The rising demand for AI rounds — and the ensuing larger rounds that we can infer from CB Insights’ data — is likely due to what Seseri described as “AI [driving] outsized results in VC because of its transformational and measurable nature.”

Outsized results in the era of megafunds must be attractive.

But while Seseri reports that from a U.S. vantage point AI-focused venture capital activity is blistering, in other markets it may simply be as frenetic as other startup categories. Atomico’s Blume, for example, told The Exchange that the “pace of deals during Q2 across venture capital has been very high, and AI is no exception,” but added that he would not “put AI down as one of the most ‘hyped’ technologies this quarter.”

What’s going on? Blume said that rapidly growing “SaaS companies with early revenue traction continue to command very premium multiples, and those using AI as part of their product are no exception.” From this perspective, AI startups are often simply part of the SaaS market and are thus part of a cohort of startups that are, in aggregate, favored.

Back in the United States, Blumberg told The Exchange that “most, if not all, of our last 20 investments are in companies using AI to deliver utility by transforming raw data into actionable insights as a major part of their value proposition.” Even more, he said that Blumberg Capital anticipates that its “next 20 investments [will] also principally deliver value by leveraging algorithms and data sets for the benefit of humanity.”

Where is competition most fierce? Blumberg said that inside of verticals is where the battle for deals is “focused.”

To sum up the data and various venture perspectives, global AI deal volume could be on pace for record deal and dollar volume, and either AI rounds are particularly hot or at least as hot as the larger SaaS market, which is itself enjoying historically rich multiples as more capital pours into the aggregate startup world.

Why, though?

Investor interest in a particular startup category often follows market demand for its goods and services; most money went into neobanks after they saw early market traction, to pick a single example. And API-delivered startups became more popular after Twilio blazed their trail, akin to how Salesforce cleared brush for SaaS startups.

AI startups are not only solving real problems but are seeing their market come to them over time. Blume said that he’s “more optimistic about the pace of AI adoption [today] than in early 2020” thanks to a decline in fear concerning AI. This is allowing companies to use “AI as an appropriate technology choice to solve a real commercial problem” without having to explain why AI is an acceptable idea. The result is one of easier “adoption and accelerating growth,” he said.

Accelerating adoption is happening as bullshit levels decrease. As we reported last year, VCs including True Ventures’ Rohit Sharma were seeing “significantly less AI-washing in startup pitch decks.” The trend hasn’t receded, Blume said. “Since early 2020, we’ve seen AI firmly take its place as an enabling technology of great software products, rather than as something that every company feels the need to talk about in order to attract investor attention.”

Perhaps as importantly, AI is becoming more and more common as something that companies are using as opposed to building their entire identity around. “The most high-potential AI companies we see today,” Blume noted, “are those who start with a meaningful commercial problem that needs solving, and then pick an AI approach as the most appropriate tool to solve it.”

In this context, AI is much more than a nice-to-have; it is a solution to a problem and a competitive advantage companies can’t afford to skip, Seseri told TechCrunch. “AI has become indispensable for both tech and nontech companies in driving everything from cost efficiency to superior user experiences at scale.”

Applying core AI

For all the interest around applied AI, there seems to be less enthusiasm in the VC community for more abstract AI, Blume reported. “Early-stage ‘core AI’ companies with teams developing innovative technology but without clear application or commercialization plans have perhaps had the least competition, with many investors still less comfortable taking this kind of risk.”

However, there is one type of “core AI” company that VCs are curious about: those that can make other uses of AI more efficient. This already contributed to the rise of MLOps, and it ties back to the “AI gross margin debate” we reported about last year — i.e., whether AI startups can be as marginally profitable as SaaS upstarts that lack a major AI component.

As you may recall, this was a conversation that a16z started in 2020 and which the firm arguably resurfaced with a new piece on “the cost of cloud.” Cost of compute is indeed one of the key variables in the AI startup discussion, and it is clear from our conversations that VCs still think about it a lot.

However, they seem eager to see early-stage startups track their economics rather than actually move away from the public cloud to help tune their economics. “At a pre-Series A or B stage, it starts with awareness of cloud spend and then typically involves optimizing the cloud workloads,” Seseri said.

What later-stage companies should do is still open for debate, and once again, technology might be the savior everyone is hoping for. For starters, it can help the cloud become less expensive, Blume noted. Beyond what we have already witnessed, “there is even more cost improvement to come here in the next few years, as access to AI-specific processors such as Graphcore’s become more widely accessible.”

The optimization space is also promising — and is also a category investors are ready to bet on. “We recognize that applications of AI have the best chance of success if they are fueled by efficient and effective data sources — and we are investing in companies that are helping businesses deal with these costs,” Seseri told TechCrunch.

As is often the case in our world, problem and opportunity are synonyms, and this is also the take that Jocelyn Goldfein at Zetta favors. “We haven’t begun to exploit the potential of techniques like active learning and unsupervised learning, better data quality tooling, automation of labeled and synthetic data, more efficient data infrastructure, to say nothing of falling costs of compute and storage. In other words: In our opinion, high data costs are just one more market opportunity for savvy entrepreneurs.”

We’ll have more data in a few weeks, once Q2 2021 venture capital data becomes generally available. But we will not be surprised to see more record-leaning results. The next question will be what portion of the AI-focused startups that raised during today’s active market will make it on their own, what fraction will sell to a larger company, and which slice will die. Investors are betting as if we’re going to see many more winners than we might from a similarly sized set of historical technology startups. We’ll see.