There's a funding bonanza in Silicon Valley right now for technology that has seemed on the verge of a breakthrough for decades.
The stories are everywhere: Blockbuster seed or series A funding rounds. Established venture capital firms like Scale Venture Partners losing out to bids as much as 10x higher than their own. Easy access to capital, along with an expanded pool of investors, is fueling a massive surge in funding for startups of all types and sizes. That's very clear in the artificial intelligence market.
So far in 2021, investors have poured $29.5 billion into AI startups, according to data compiled for Protocol by PitchBook. That already dwarfs 2020's $27.8 billion total with several months, and likely more deals, still left.
"I don't think I've seen a PowerPoint without the word AI in it in the last two years," Venrock partner Brian Ascher told Protocol. "It's almost like software equals AI."
There are many converging trends driving the increased interest in AI and machine learning. Long considered an emerging technology, the pandemic helped show organizations the benefits of automation and how it could turn areas of their businesses that were once considered a cost drain into a value driver.
After years of failure by many enterprises to deploy AI, there are more operational use cases that demonstrate its potential. Companies that were hesitant to move to the cloud are also now rushing to catch up to rivals, prompting new strategies around how to use AI to drive operational gain from all that stored data. And industry-specific applications are exploding, including in sectors like health care that in the past weren't as much of a focus among software developers.
"The percentage of our investments that are AI-first … has gone from 5% of the fund to over 50%, and is probably continuing to rise," said Scale Ventures Partners' Andy Vitus. "It feels as though every single thing we look at these days has an element of machine learning. Some things tend to be a flash in the pan. This one's not."
'Thin tech with a veneer of AI'
The definition of what AI really means is expanding significantly — to the chagrin of some industry veterans. But it goes beyond just the end application.
One increasingly attractive area is data management, with startups like Fivetran and Databricks trying to tackle key infrastructure challenges that will ultimately be necessary to support the user-facing AI tools. Overall investment in the sector through mid-July 2021 was $2.2 billion, according to PitchBook. While that's just slightly above 2020's $2.1 billion, the lower deal count — 43 to date in 2021 versus 84 in all of 2020 — is evidence of the industry's maturation.
"Applications are overdone," said Shruti Van Dyke Gandhi, general partner and founding engineer at Array Ventures. There need to be companies "in the middle creating better pipelines for you or cleaning the data up for you, so you can start analytics around it," she added.
But the increased attention on the AI sector overall is leading to entrepreneurs trying to exploit the bonanza of funding. Some startups are packaging commonly available tech as something more groundbreaking to generate marketing buzz, according to VCs. Compounding that problem is a proliferation of open-source tools, which makes it difficult to establish any common industry benchmark around key aspects of AI like the labeling of the data, particularly as some commonly used toolsets are found to be error-ridden.
"There's a lot of very thin tech with a veneer of AI whitewashed over it," said Vitus. "The key figure of merit is always: Can you do this as well or better than a human? There are a lot of AI companies out there that cannot."
Now, particularly as more AI-based startups go public, there's a bubbling movement among founders focused on end applications to be more transparent about the robustness of the underlying algorithms. Some are even releasing research ahead of wide-scale product launches to show the power — and more importantly, the accuracy — of the systems.
And as the industry matures, there's broader recognition that many of the hyped-up promises around AI are decades away — if not impossible for humans to achieve. In a nod to how expectations have changed, there's a growing number of startups that are seeking to augment — not replace — humans in many job-specific tasks, according to Venrock's Ascher and other VCs. It's part of an effort by startups and large vendors alike to provide more tailored products instead of general AI applications that are marketed to a broad range of use cases.
"Over the last five years, entrepreneurs have gotten a much better sense of the problem they are trying to solve," said Ascher. "There are people who just throw AI in a pitch deck to put lipstick on a pig. And then there are others who have figured out the right problem and built the right solution."
The chaos continues
Still, there are many areas where AI tech has advanced significantly. Chatbots, for example, promised to replace human agents. That has yet to come to fruition, according to industry insiders, though it remains a quest for some. But despite that setback, AI-backed call-center software — like real-time transcription services and post-call analysis tools — are used across many companies; it's one reason why the space is seeing so much activity.
"That tech is now at an accuracy level that is better than human beings," Five9 CEO Rowan Trollope previously told Protocol.
The current environment seems poised to continue so long as capital remains easy to get. While such a surge usually brings about questions of a bubble, it's clear enterprises are only going to deepen their use of AI. And with such a huge market potential, expect to see even more entrepreneurs look to build startups around the tech.
"It's complete chaos at the moment," Vitus said. "Until you see interest rates returning to normal, it's going to carry on."
