If you read the press today, you could get the impression that the US and China are in some form of Technology Arms Race. (Headlines like “America and China Are in a Technology Arms Race” offer a pretty clear clue.) According to this, there is a global race to build the most advanced semiconductors to control the infrastructure of all future technology and thus everything else. Will China win the “Race to 5G?”. “Who is Best Positioned in the AI War?”, and presumably the ensuing Robot Apocalypse? Clearly there are some people who see the war as already underway, surely Silicon Valley is rising to the task.
Most of the coverage of this ‘Arms Race’ is about semiconductors. In part, this is because semis are the fundamental building block of everything else. The Cloud? Yeah, that needs a lot of chips to power servers and networks. 5G? That too is all about basebands and radios. AI? Data is important here, but so are chips to process all that data. IOT? Yes again, actuators, sensors, microcontrollers and connectivity. And for years, China’s government has very vocally and actively built an industrial policy around semiconductors. Chairman Xi famously gave a speech a few years ago calling out the fact that China spends more on chips than on oil.
Sounds pretty serious. Except the situation on the ground in the US tech industry bears no resemblance to this. Yes, everyone is very worried about the Trade War, but some big War for the Future of Technology? Not so much. It’s almost as if someone declared war, but one side has not bothered to show up.
For the past several months, we have been very busy working with a half dozen clients who build hardware and semis, helping them raise money. Except, there is almost no one to raise money from. US Venture Capitalists do not invest in hardware, let alone semis anymore. From our latest study, there are only about 20 US-based VC firms who have made a semis investment in the past three years, out of roughly 1,000 VC firms. And of those who have made semis investment, only a handful have made more than one investment. A few years ago, when AI and Google’s TPU (their AI chip) were fresh on the scene, the biggest firms all made a bit on one AI chip start-up. Aside from that, and a few (less than 5, probably less than 3) firms have ventured a little further afield.
By contrast, there is pretty much an infinite amount of venture money in China for chip investments. Of course, US firms cannot tap that for fear or running afoul of the US’s CFIUS committee which has to approve all foreign investments in the US.
Does this mean the US VC community is lagging in their patriotic vigor? Is the US doomed to lose this war? There are really two ways to explain this disconnect, and as always, the reality is much less stark than the headlines paint.
First, we are not at war with China. Let’s hope it does not come to that. Yes, the two countries have multiple areas of conflict and disagreement. And count us in the camp who have experienced that country’s unfavorable practices towards US intellectual property (IP). But there is still immense overlap in our economies especially around technology. China naturally wants to develop its economy and move into higher value-added sectors, including semiconductors. But to call that a ‘War’ is to sensationalize the situation greatly. So it is unfair to question the patriotic spirit on Sand Hill Road.
That being said, the VC community has a serious herd mentality problem. Over the past decade, the rise of mobile, cloud-based apps has made VC investors immense fortunes, and that has crowded out thinking about anything else.
Put simply, American VC’s have a spreadsheet problem. Investing in cloud software is too easy to quantify. Take a software company’s user growth, calculate monthly recurring revenue, subtract customer acquisition cost and you have your investment decision. We have had serious conversations, with serious VCs who ask for these kinds of metrics when evaluating semis investment.
Admittedly, investing in semis is hard, by which we mean capital intensive. Two people in a proverbial garage or broom closet can start a software company, and grow it to meaningful size on tiny amounts of capital. We work with those companies too. By contrast, getting from inspiration to first revenue for a chip company costs $20 million. Scaling that is likely to take another $20 million (and up). And bugs can be fatal. If your app has a bug, you can patch overnight. If a chip has a bug, that can be the end of the company. (More prosaically, it means that chip companies have to spend A LOT of money testing before going to manufacture, see $20 million above).
Does that mean VCs are making the right decision by shunning chips? No.
Chips are capital intensive, but that just means the VCs have to work harder at vetting the management teams they choose. In the middle of the decade, when the big VCs were investing in AI chips, they threw huge checks at companies with little more than PowerPoint and some good ideas. Little thought was put into the execution side of things. Now that many of those companies are sliding downwards, the VCs have adopted a once-bitten-twice-shy attitude, rather than admit they should have done more diligence.
The other thing to keep in mind is that building a chip has gotten much less expensive. Last year we wrote about all the IP licenses and various forms of taxation that chip companies have to pay, as well as the need to support a large back-end operations team to bring chips to production. Those problems still exist, but the market has adopted many solutions to lessen this burden, largely in the form of 3rd Party ‘ASIC Shops’ who can amortize many of those upfront costs across multiple customers. A good example of this is SiFive who can eliminate the need for an expensive ARM license and produce a hardened chip.
At the same time, the market for semis is MUCH better than it was a decade ago. In the 2000’s, chip companies needed $200 million to get to break-even and faced low-growth, saturating end-markets. Since then, a wave of consolidation has shaken the industry which has greatly improved competitive dynamics and thus pricing. In fact, the consolidation has led the smaller number of larger chips to leave big gaps in their product roadmaps. At the same time, we are poised for major growth in demand for semis in a wide swathe of markets. As above – AI, robots, data centers, cars – are all either major markets already or poised to become so in coming years.
By avoiding semis entirely, VCs are leaving a lot of money on the table. This kind of investment drought has hit semis several times over the last 40 years. The 1990’s saw a similar shunning of chip investments for high flying dot.com companies. Today, there are only two or three companies left in the US that can actually manufacture its own chips. So there is still significant risk that bad decisions by VCs today will not only lose out on major gains but also create new, major gaps in the US’s technological capabilities.