Self-driving vehicles aren’t expected to be a reality for at least a couple more years. But a growing coterie of venture-backed “deep technology” companies are hitting new milestones in developing components for advanced driver assistant systems (ADAS) that are needed to analyze data collected by cameras on all sides of a vehicle in real-time and make safe navigation choices.
Just as cellphone cameras keep improving, so does the resolution on cameras that will be a mainstay of autonomous vehicles. The shift to 8 megapixel cameras means cars “have to process a tremendous amount of data so you can build an environmental model around the vehicle in real time to be able to detect cars, trucks, pedestrians, bicycles – everything around you including lanes, traffic signs – and classify them,” says Marc Bolitho, CEO of Recogni, a company based in San Jose, California, that builds high-performing AI inference software to crunch that data.
“To do that you need this high compute [capability],” says Bolitho. “You need to do this very fast and process this data quickly so you can leave time to make a safe driving decision. And you also need to do that efficiently. Today’s solutions aren’t efficient from a power perspective.”
Adding self-driving systems to electric vehicles will increase their energy usage and dramatically reduce how far and how long they can drive before needing a recharge.
Increasing the compute power of ADAS while also making them more energy-efficient would extend EVs’ range. It’s one of the engineering challenges that define much of what’s known as deep tech. The typically longer product development cycles are seen by some as a bad fit for many VC investors, who are more comfortable with technologies that are easier to scale and bring to market.
While some investors are put off by deep tech’s challenges, others see them as a way for start-ups to gain a competitive advantage and create a barrier to entry for future competitors. Boston Consulting Group estimates that investment in deep tech more than quadrupled from $15 billion in 2016 to more than $60 billion in 2020.
Deep tech’s “economic, business and social impact will be felt everywhere because deep tech ventures aim to solve many of our most complex problems,” BCG wrote in an overview of its January 2021 report Deep Tech: The Great Wave of Innovation. “The great wave encompasses artificial intelligence, synthetic biology, nanotechnologies and quantum computing, among other advanced technologies.”
Technologies that fall into the “other” bucket include tech for autonomous cars, climate technology, robotics and space technology, say deep-tech investors.
Recogni is considered part of deep tech because it designs high-performance computing products for AI workloads in automotive electronics that enable safe and sustainable autonomous driving. The company is still securing commitments to its Series C round, targeting an undisclosed amount of capital. It had previously raised $73.9 million over three funding rounds from a dozen investors, including Celesta Capital, Mayfield and the corporate venture arms of BMW, Continental Tyre Group and Toyota, according to PitchBook. The $48.9 million it raised in a Series B round in February 2021 enabled Recogni to double the size of its team and start applying its energy-efficient AI inference architecture to its first device, Bolitho notes. In early January, Recogni unveiled the Phoenix ADAS/AD ECU system, which combines high processing capacity for high-resolution sensor data with low latency and low power consumption for autonomous vehicles.
“We expect to start seeing design wins for some of our companies that are working on next- generation automotive-type applications”
Nicholas Brathwaite, one of the founding managing partners at Celesta Capital, which invested in Recogni’s Series B round, says the start-up’s computational ability is “an order of magnitude superior to the competition” and consumes one-fifth to one-eighth of the power used by competing devices.
“It’s an essential ingredient to overcome the compute bottleneck being seen by the industry to realizing and advancing on the path to autonomous vehicles,” Brathwaite says.
At the recent Consumer Electronics Show in Las Vegas, where there was plenty of buzz around advanced automotive software and hardware, Recogni demonstrated its electronic control units and held meetings with many OEMs and industry analysts.
This year “we expect to start seeing design wins for some of our companies that are working on next-generation automotive-type applications,” Brathwaite tells Venture Capital Journal. While the mass deployment of electric vehicles has already begun ramping up, notable numbers of vehicles with autonomous features aren’t expected to come to market until 2026. “That means you need to be getting design wins this year so you can get into [OEMs’] 2025 platforms.”
Mass production of ADAS can wait for a couple of years, but “automotive [companies] make decisions a couple of years in advance of when the actual product comes out,” Brathwaite adds.
Bringing some of those products to market this year is a big target “because for some companies, quite frankly, if they’re not able to commercialize in 2023, it puts pressure on their ability to raise additional funding,” says Brathwaite.
Gartner expects the automotive semiconductor market to double to almost $120 billion by 2030, translating to a compound annual growth rate of 11.5 percent, according to PitchBook’s Q3 2022 Mobility Tech report. Gartner estimates that as automobile OEMs move aggressively to form tighter ties with chipmakers and bring design in-house, half of the top 10 automobile OEMs will design their own chips and build strategic relationships with chip companies as soon as 2025.
ABI Research, a market-foresight advisory firm providing strategic guidance on the most compelling transformative technologies, expects 3.76 million consumer vehicles that ship in 2025 to feature Society of Automotive Engineers Level 2 and higher capabilities, which let drivers shift safety functions to the vehicle under certain conditions.
From an investment perspective, deep tech can present a conundrum. On one hand, most of the innovation is in hardware rather than software, which much of the broader venture ecosystem centers around. Hardware generally takes longer to develop than software, lacks software’s easy scaling potential and doesn’t generate usage data early on once a prototype exists, as software does.
On the other hand, because much of the risk in deep tech is in the technology development instead of the product-market fit, once the technology risk itself is resolved “very often the market demand is already proven,” says Benjamin Joffe, a partner at deep tech VC fund SOSV. “If you manage to build something that’s 10 times better than what’s out there,” there is already a market for it, he says.
SOSV expects the final close on its $400 million Fund V around the middle of this year. The firm has been using a $100 million growth fund that closed in June 2021 to provide follow-on funding to portfolio companies whose product development and market strategy are going well.
SOSV focuses on two major verticals: climate tech and health tech. What differentiates the global firm from most other deep tech VCs is its pre-seed programs that provide office and lab space, initial funding and in-house subject experts who can help entrepreneurs develop their products, anticipate risks and figure out go-to-market and fundraising strategies.
SOSV’s IndieBio start-up program opened a new 25,000-square-foot facility in midtown Manhattan in December to serve founders in food and agtech, bio-engineered materials, therapeutics and drug discovery. Aiming to be a hub for life sciences development in New York, IndieBio NY has secured pledges from Empire State Development and the Partnership Fund for New York to invest $25 million over the next five years. IndieBio launched its first such open office facility in San Francisco several years ago.
IndieBio partner Parikshit Sharma eschews the term “accelerator” because most accelerator programs have short-term engagement with start-ups, which typically ends when start-ups graduate.
Joffe says there is a misconception that investors need a PhD to understand and invest in deep tech and that all deep tech projects are very expensive and have longer development curves. “They can be extremely capital-efficient,” he says. “Some go to market within a couple of years. Some others are really slow.”
For example, some of the synthetic biology start-ups backed by IndieBio have already entered into commercial partnerships despite being fairly early in their fundraising efforts. MycoWorks, which makes synthetic leather from mushrooms, has a contract with the fashion company Hermes for one of its handbag brands. Another IndieBio start-up, Gozen Bioworks, which is also producing animal-free leather, already has a partnership with a product manufacturer and distributor of high-end leather products.
“For every dollar LPs have given us, we’ve given LPs back on average five dollars”
Yet another IndieBio investment, The Every Company, which creates alternative proteins that can replace eggs in foods, is partnering with Ingredion, a leading food ingredients company, and with BioBrew to scale up and manufacture proteins.
Early commercialization isn’t confined to biotech products. Nalej, which has built an infrastructure platform for edge computing, so far has not had to fundraise beyond its $3 million seed round in 2019 thanks to lucrative commercial and public sector contracts, says Yanev Suissa, a managing partner at Virginia-based SineWave Ventures, which invested in Nalej’s seed round.
“They have a massive contract with the Air Force and the Department of Defense that we originally struck with [the Defense Advanced Research Project Agency] to provide edge infrastructure and solutions to those groups,” Suissa notes. “That’s made them quite successful. They have double-digit millions in revenue.”
Nalej may need to raise another round in the future, but for now “they’re in great shape for a seed company that’s built a great technology,” he adds.
One reason that Shabodi, a 5G infrastructure provider in SineWave’s stable, was quick to raise a $10.3 million Series A round last year on the heels of a $3.3 million seed round that closed in December 2021 was to accommodate investors with ties to strategic players that couldn’t get into the Toronto company’s competitive seed round.
SineWave led the Series A round along with CEAS Investments, a new investor whose experience with 5G networks and relationships with strategic partners and customers Suissa says are “particularly valuable” to Shabodi.
Shabodi doesn’t “need the money from an investment perspective. They want the business partnership,” Suissa explains. The company asked for business contracts from interested strategic investors before allowing them to invest.
“A lot of sectors of deep tech have accelerated significantly, so they’re raising larger rounds earlier just because the technology is able to get to market faster”
Instead of individual 5G telecom providers such as AT&T, Ericsson and Verizon having their own platforms and applications, Shabodi’s architecture allows any application to be layered on top and enables simple integration with any of the various cloud service providers.
In addition to commercial partnerships, some deep tech companies benefit from government support. Infinite Elements, a bio-mining start-up backed by IndieBio, received a $500,000 grant from the Advanced Research Projects Agency-Energy for a project that runs from August 2022 through August 2024. The company uses biological processes to extract, concentrate and refine critical metals and rare earth elements from electronic waste and low-grade mineral ores at a cost that is said to be competitive with conventional mining.
Government programs are significant because they offer start-ups non-dilutive support, not just for research – as the Small Business Innovation Research program does – but also for pilot facility manufacturing, says Sharma at IndieBio.
Creative Ventures, whose deep tech investments are in AI, robotics, synthetic biology and advanced materials, says its focus on macroeconomic trends, including shifting demographics, has resonated with limited partners. The firm, based in Oakland, California, began fundraising for its third fund near the end of last year, targeting $200 million.
Among the demographic changes driving Creative’s portfolio decisions are an aging population, which is driving up healthcare costs and exacerbating labor shortages among younger workers. “If you think about it as a productivity equation, there’s fewer young people to throw at manual labor problems relative to an older population,” says James Wang, one of Creative’s general partners.
That has informed Creative’s investment in two AI-based start-ups: Path, whose proprietary AI system enables robots to understand and weld parts without the need for skilled welders or robot programmers, and Caper, which uses AI to replace cashiers.
In addition to funds of funds and family offices, Creative’s LPs include strategic investors such as Singha Ventures, the corporate VC fund of Singha Corporation and Boonrawd Brewery Group in Thailand. Singha Ventures anticipates supply chain disruptions related not only to changes in agriculture and climate change but also projected labor shortages, says Wang.
“They see [the need for these technologies] within their own business, and it’s similar with other strategics,” he says. “Even if they don’t perfectly understand the reinforcement learning algorithms that are driving the AI that we invested in, they definitely do understand the problem” and appreciate that these start-ups are able to address it.
Creative’s Fund III is expected to be about four times the size of its second fund, which closed in 2018. One reason is that the firm wants to contribute more per round – 50 percent or more of the total instead of less than 10 percent – as it did with Fund II, says Wang.
“A lot of sectors of deep tech have accelerated significantly, so they’re raising larger rounds earlier just because the technology is able to get to market faster,” he explains. Timelines to an exit are being shortened from eight to nine years to three to five years in some cases, depending on the technology.
SineWave is also looking to more than double the size of its second fund, which closed on $50 million in 2019. The firm expects to hold a final close for Fund III, whose target is said to be up to $150 million, by the end of the second quarter.
Suissa acknowledges that the overall fundraising environment has gotten tougher, but he says SineWave hasn’t had trouble getting commitments from LPs because of its track record of healthy cash distributions. “For every dollar LPs have given us, we’ve given LPs back on average five dollars,” he says. “We’ve already made those returns. We’re not just claiming we made you five times your money on paper. If you don’t have DPI right now, you can’t raise for the life of you.”
The 19th and 20th paragraphs have been updated to reflect Benjamin Joffe’s correct title at SOSV and the correct target amount of Fund V and when its final close is expected.