How Next-Gen Chipmakers Are Raising Money, Taking On Tech Giants

The advent of big fundraising rounds for startup chip innovators—a class shunned by most venture capital firms only a few years ago—seems to mark a turnaround in recent VC attitudes about semiconductor investments.

But it turns out VCs weren’t the only driving force behind the change in fundraising prospects for the young companies now inventing chips suited for artificial intelligence tasks, and daring to compete with giant chipmakers such as Intel (NASDAQ: INTC) and Nvidia (NASDAQ: NVDA), says Silicon Valley Bank’s Matt Trotter.

Trotter says the financing logjam was largely broken by companies that were eager to be customers of such next-generation chipmakers, because their sensor-studded products, such as cars and drones, were collecting masses of data that needed AI-powered analysis.

“The demand from customers to have innovation was so strong that they invested alongside VCs,” says Trotter, Silicon Valley Bank’s managing director for hardware and frontier technologies.

In some investment deals, there might have been only one VC firm, says Trotter, who has been observing the changes in chip startup fundraising for some time. Silicon Valley Bank serves as a commercial bank, networking resource, and sometimes a lender to most of these young chip companies, he says.

Five years ago, venture capital firms were gun-shy about semiconductor investing because the risks were very high, Trotter says. It could cost as much as $100 million to bring a new chip to market, and then it might not turn out to be a competitive product, he says.

“We saw a lot of VCs get burned by that,” Trotter says.

But recent years have seen the rise both of AI and of devices capable of scooping up unprecedented amounts of data that could, with sophisticated analysis, make it possible for a car to drive itself, say, or a virtual-reality headset to pan over digital scenery in time with the movement of the user’s head. Existing chipsets couldn’t meet those needs for computing power within a device, Trotter says. The established practice of exporting reams of data to a high-powered Web-based computing platform wouldn’t work for a driverless car.

“You can’t send that data up to the cloud and have it come back before it decides whether to put on the brakes,” he says.

Hence the recent run of AI chip investments. Redwood City, CA-based Mythic, which developed a chip designed for use in devices at the edge of a network, raised its $9 million Series A round from venture firms DFJ, Lux Capital, Data Collective, and AME Cloud Ventures last year. In March, it raised $40 million more, but the lead investor was Japanese telecom giant SoftBank—which owns the leading chip design firm ARM. SoftBank was joined by another strategic investor, Lockheed Martin Ventures, along with other backers.

Though “edge computing” is a hot area, other chip innovators such as U.K.-based Graphcore are trying to meet the needs of high-performance data centers.

Graphcore’s early investors included venture capital firms such as Foundation Capital and Amadeus Capital, but also the venture investing arms of industrial companies, including Robert Bosch Venture Capital, Dell Technologies Capital, and Samsung Catalyst Fund. After raising a total of $60 million in its Series A and B rounds, Graphcore scored $50 million from Sequoia Capital in a Series C round in November.

Trotter wouldn’t say that VCs in general are lining up to get into fundraising rounds for chip startups. However, the resistance among VCs is easing, he says. “There were hardly any before; now, there are some.”

Silicon Valley Bank has introduced semiconductor startups to potential customers that might also become investors, Trotter says. The bank has also provided venture debt deals for young companies that have already raised outside funding, and want to fluff up their capital cushions to have more time to meet further milestones before they raise their next rounds, he says.

The challenges facing chip startups can seem daunting, because their competitors include well-heeled semiconductor giants such as Intel and Nvidia, which are revamping their own chips to better handle the demands of artificial intelligence—and to defend their market dominance.

However, Trotter says such big rivals may not be as formidable as they seem—and they can turn out to be customers or acquirers of startups. An established public chip company has to present quarterly results that satisfy shareholders—a deterrent to spending too much money on a novel chip design that may not pan out, he says. Or, even if the in-house invention is successful, it could cannibalize revenue that might have gone to the company’s existing product lines.

To augment their own R&D, big companies such as Intel, Qualcomm (NASDAQ: QCOM), and Broadcom (NASDAQ: AVGO) are on the hunt to acquire chip startups that have demonstrated some success, Trotter says. Intel began a run of acquisitions in 2016, buying chip startups Nervana Systems and Movidius, as detailed by The Verge.

Such acquisitions can encourage more investment from VCs and other backers who want to know that there are exit possibilities other than an initial public offering.

Meanwhile, Intel and Nvidia, which now share the market for silicon in high-performance data centers, are facing competition from other big tech companies getting into the game with their own chip development schemes: Google, Microsoft, and Apple.

It seems that startups, rather than being trampled underfoot in this forest of roving giants, can find niches among them. Trotter, who observes the startups from the perspective of their fundraising needs, describes a typical young chipmaker’s progress in four stages:

1. Tape-out: When entrepreneurs feel their engineering is ready for the real world, they contract with a semiconductor fabrication factory, or “fab,” to make a chip for testing. To get to this goal, young companies might raise $10 to $50 million, depending on the kind of chip, Trotter says.

That chip test is a fateful moment that highlights the big risks of investing.

“If it doesn’t work—that’s it,” Trotter says.

However, there may be hope of resurrection for reconfigurable chip designs with software components built in, because their inventors can tweak them after an unfavorable test to try to make them work better, Trotter says. “That takes some of the risk out of it,” he says.

2. Early customer traction: After a successful tape-out, chip startups go on the road to try to land contracts with selected customers who will build their chips into the next version of the customer’s product, Trotter says. This phase doesn’t require as big a fundraising round, because a massive sales and marketing drive isn’t necessary. About $10 million to $20 million might do it—and scoring just a few customers counts for a win, he says.

“If you land one, you’re in a pretty good place,” Trotter says. Although it may be years before the product goes to market containing its chip, the startup may begin to earn non-recurring revenue for engineering work to customize the chip for the partner’s design, he says.

3. Manufacturing scale-up: When the customer’s device is ready to be made, the chipmaker gets a fab set up to crank out its chips, and as the device hits the market, the real revenue flows in for the chip developer. “Once they get going, it’s a good margin product,” Trotter says. The novel chip is not a long-used component sold at commodity prices, he says. The young company can then finance its operations based on the expected revenue.

4. The next generation of the next-gen chip: The startup can now work on a newer version of its innovative chip, Trotter says.

Which is the better investment—chips designed for edge computing, or for big data centers? Trotter says he often gets this question, and doesn’t have a definitive answer. “At least from what I see, there’s going to be a home for both,” he says. “The customers they’re talking to are quite different.”

According to an estimate by CB Insights last year, cited by CNBC, investing in chip startups was on pace to reach $1.6 billion in 2017. Trotter says we may be seeing a fraction of the potential right now, because the uses of AI-powered technology—in robotics, aerospace, transportation, virtual and augmented reality, and other applications—are barely beginning to hit the market.

“A lot of those companies are in the early stages of their development, like autonomous cars,” Trotter says. “As they become much more prevalent, the returns to chipmakers should grow in proportion.”

Bernadette Tansey is Xconomy's San Francisco Editor. You can reach her at btansey@xconomy.com. Follow @Tansey_Xconomy

Trending on Xconomy