Start-up Semis Live

Last week, CRN’s Dylan Martin wrote a great list of 10 ‘hot’ semis start-ups, showing that semis companies can still raise capital. Here we want to take a look at what these companies do and how much they have raised, to give a sense of what can get funded in semis land these days. We have written a lot (and here) about how hard it is to raise venture money for chip companies. And while we would like to believe that list proves us wrong, this list marks a good start, that hopefully gets other investors interested.

The list explicitly does not include AI chip start-ups, promised for a future article. Our only quibble with the article is that it is presented in the form of a ‘listicle’ with each company listed on a separate web page. So here we have provided the list showing their latest funding round and amount.

CompanySectorLatest RoundLatest AmountURL
AmpereServer CPUNot disclosedA Lot
AsteraInterconnectNot disclosedNot disclosed
AyarOptical InterconnectB35
Kameleon SecurityNot disclosedNot disclosed
NuviaServer CPUB240
Source: CRN “The 10 Hottest Semiconductor Startups of 2020” by Dylan Marin, November 23, 2020

A few things stand out to us on this list.

First, the amounts raised are big. Some of these companies are huge. A few like, Nuvia, have immense job listings pages. But we also know headcount is not the only expense for chip companies. They need to pay significant amounts to actually manufacture chips. Only about half of the names on the list have raised enough to reach production, let alone revenue.

A second note is that none of these companies are targeting consumer end markets. (SiFive is a partial exception, we will come back to them in a moment.) These companies are all squarely going after data center and enterprise customers. Only one company, EdgeQ, has anything telecom related as a major use case, and we think they are likely to focus on enterprise markets as much as anything. Just over half the names on the list are going after ‘data center’ problems – largely around network and storage. All of this is sensible. The consumer markets are saturated and not growing, while data center demand is booming.

Encouragingly, two companies are building server CPUs. This not an easy market to enter, but it is large and immensely lucrative. The hard part about building a CPU for data center servers is that the chips need to make sure they support the software used in data centers. Intel has built a huge competitive moat for its data center CPUs by investing heavily in software for 20 years. True, any software can quickly be compiled to run on these new, ARM servers, but optimizing them for performance takes a lot of work and time, and performance matters when we are talking about data centers. A chip that runs core software 5% slower than Intel probably won’t be competitive. As a result, the two CPU companies on the list – Ampere and Nuvia – have more job listings for software roles than hardware. We know that Ampere has crossed this particular chasm and has built up its ecosystem, and we suspect Nuvia has made credible progress. That being said, Nuvia’s last round raised $240 million, for a total of $293 million raised (according to Crunchbase). Ampere, which is owned by private equity shop Carlyle does not disclose the amount it has raised, but we believe it is a very big number, larger than anyone else on the list by a considerable margin.

The last thing we want to note is that most of these companies are all targeting relatively small segments. Storage and networking are great markets but they are an order of magnitude smaller than the CPU market. So carving off the proverbial 10% market share nets a much smaller opportunity. There is nothing wrong with that from an investor’s perspective. These companies all have the potential to achieve respectable exits, but they will all have to exit at some point.

Of the ten names on the list, only 3 really have the potential to survive as stand-alone companies – the two CPU companies and SiFive. The CPU makers are attacking a sizable market, and with Intel’s current troubles, it is a very vulnerable market. SiFive is a bit less clear. They are the flagship commercial entity for the RISC-V open source processor. We are believers in RISC-V (especially if Nvidia succeeds in buying Arm), but commercializing open source projects is hard. SiFive has made a lot of progress, and could build a sizable platform, but faces a lot of questions about its business model.

Overall, we are heartened by the number of companies, their quality and the amounts they have raised. We hope that this encourages more investors to look at funding semis companies.

One response to “Start-up Semis Live

  1. Pingback: Did Qualcomm Just Buy a Server CPU Company? – No | DIGITS to DOLLARS·

Leave a Reply