Re: Showdown: Part 1 Download
Jahed Stetter
Next up, I wanted to understand the VCs who compete in the Chinese AI/ML ecosystem, as China has some massive funds like the Yale-influenced Hillhouse Capital, which are for the most part overlooked in the West. The top VCs (across all stages) by deal count are:
Going forward, I wonder how western VCs with China-specific funds will choose to underwrite investments when there is significant geopolitical risk, especially if such risk can result in total portfolio wipe-outs.
In aggregate, I analyzed a total of 3,159 fundings and M&A transactions that fall under the general bucket of AI/ML companies, of which 2,421 were applicable to the MAD Index (there are companies that claim to be MAD companies, but did not have significant elements of AI/ML embedded in their products).
As an early-stage investor, I spend a lot of time looking at all things machine learning, AI, and data infrastructure. One of the projects that I recently worked on was Firstmark Capital\u2019s annual MAD Landscape. In analyzing over 3,000 companies (that have raised an aggregate of $153B in funding across different rounds!), one of the trends that stood out was the scale and investment direction of venture capital in China. This is especially interesting as the western world and China go through a period of \u201Cconscious uncoupling\u201D (Gwyneth please don\u2019t sue me) in their technology development, especially within the realm of AI.
Part 1 of this post explores some of the macro funding trends in China, the funders who back AI/ML/data startups, and how China and the United States have broadly chosen to allocate capital in very different ways. Zooming out, Part 2 (coming in a week or two) highlights the importance of semiconductor sovereignty in the progress of AI research, product development, and investing. I also look at some of the broader, second-order effects as a result of our current semiconductor tensions. Let\u2019s dive in.
For the rest of this blog post, MAD companies broadly encapsulates companies in the various MAD categories/sub-categories (Infrastructure, ML & AI, Analytics, Applications, etc.) that received funding between August 2021 and December 2022. For detailed methodology, please refer to the \u201CMethodology\u201D section of the post at the end.
At first glance \u2014 it would seem that China is far behind on ML/AI investments. The reality is much more nuanced, however. First off, the incremental venture dollar invested in China likely goes a lot further due to the significantly lower cost of labor. Using software engineer salaries as a rough proxy, the median Chinese software engineer makes just $71K, while the median US software engineer makes $170K (according to Levels FYI at the time of writing).
A reopening China that will undoubtedly be very focused on LLM products after the success after ChatGPT (and the announcement of GPT-4, Anthropic\u2019s Claude, and Google\u2019s PaLM APIs all on the same day!)
Here, we see that for the top 10 US investment categories by dollar amounts, the US leads by a wide margin \u2013 except for Transportation (generally autonomous driving companies). Flipping things around, we do the same for China
Here, we see that Chinese investment \u201Cspikes\u201D in Transportation and AI hardware (where >50% of all Chinese venture dollars are allocated). Given Biden\u2019s recent restrictions, China is likely looking to cultivate its own national champions and develop a parallel/independent autonomous driving and AI hardware stack. What\u2019s also notable is that the remaining categories tend to focus on the deeper tech (industrial, life sciences, computer vision, NLP) or the core \u201Cutility\u201D (Automation & Operation, Customer Experience / Service) side of things, while the US can afford to allocate heavy investment dollars into categories like Commerce and Finance & Insurance in addition to \u201Ccore\u201D categories.
The US VC ecosystem has generally taken the laissez-faire, \u201Cletting a thousand flowers bloom\u201D (thanks Matt / Mao) approach to investing. This has traditionally worked because the amount of LP dollars that flow into the American ecosystem dwarfs every other nation in the world
China, on the other hand, seems to have some invisible hand that has directed investment dollars into the more \u201Cdeep tech\u201D parts of the tech ecosystem (and concentrated in fewer companies)\u2013perhaps with the understanding that having these technologies available onshore will be accretive in future (potential) great power conflicts
As we seemingly enter a period of capital scarcity, venture firms in the US have to be much more deliberate in how they allocate their dry powder to drive returns (especially as we see overinvestment in many of the \u201Clow-hanging\u201D MAD categories). This could mean taking on more technology risk or betting on new business models. The meta question is then \u2013 will capital be concentrated in a select few investment categories (given how many Web3 VCs are seemingly now gen AI experts, that seems to be the case), and if so, does that present openings in investment categories with relatively saner competitive dynamics for investors willing to look harder?
My previous Petit Verdot blind tasting was a 2017 vintage-only event, so I switch it up in order to avoid being repetitive. This time, I allowed Petit Verdots from any vintage, from any area around the world.
The addition of a pair of California Petit Verdots added an interesting dimension. I did a little bit of research and realized that there are only 800 acres of Petit Verdot in California (compared to around 23,000 acres of Cab Sauv).
Veritas and Glen Manor have great reputations. The wine from True Heritage was made by Emily of Veritas, although using fruit from a farm near Keswick. Glen Manor needs no introduction; their 2017 Petit Verdot was one of my favorites from the previous event.
Wine 1: 2018 True Heritage: 4.5 votes (winner). Well balanced. Not a lot of any one particular note but it seemed to be a well-rounded crowd pleaser. I later saw that this was 25% Merlot, which made it considerably easier to drink as a stand-alone wine.
Wine 4: 2019 DuCard: 2 votes. Younger and very expressive wine. Ripe fruit on the nose, some fruit notes. Spice notes and higher alcohol on the palate. I thought the color was a lighter shade than the rest.
Wine 5: 2018 Vint Hill: 2 votes. Smooth. Little spicy but not overwhelming. Also a lighter shade than the rest. I personally voted for bolder, in-your-face wines but the guests who prefer easier drinking wines liked this a lot.
Wine 1: 2018 True Heritage: 2.5 votes. Our notes were all over the place. Several noted it was very tart and improved. I thought it actually lost something from an earlier round. Others felt it was very consistent, well balanced, and drinkable.
In the decades past, chip fabrication was seen as a commodity to be outsourced, and the arrival of pure-play fabs like TSMC enabled chip design companies to decouple design of hardware and the manufacturing of that hardware. As production processes became more advanced, however, the CapEx required for each successful generation of chips (node shrinks) became significantly more expensive (a leading edge fab might cost 10+ billion).
In the US, Intel stands alone as the only domestic company attempting the leading edge (though they are behind Samsung and TSMC). Global Foundries, one of the other leading edge players until relatively recently, essentially gave up at 7nm and have since then disbanded its advanced node exploration teams.
Given that access to semiconductors (GPUs in our case) is so critical in the advancement of AI (likely one of the core economic drivers of this decade), it would seemingly make sense for the supply chain of semiconductors be onshored or friendshored so we have steady access to compute. Unfortunately, because of the consolidation dynamics I mentioned above, the world is completely reliant on one company in TSMC, on one island (if the US loses access to leading-edge semiconductors, the accelerators we have will have to tide us over for the next few years until onshore fabs come online). As a result, TSMC has become a flashpoint in the AI wars.
I think there are several second order effects which make our current semiconductor showdown so consequential should tensions between China and the US escalate to a more kinetic domain (beyond any balloon misadventures):
China invades before the US onshores its own semiconductor manufacturing (pre-2026): likely heavier US military involvement given that a strategic resource would essentially be cut off
China invades after US semiconductor onshoring: the hope here is that we have a more negotiated cool down in tensions given that there would potentially be less of a reason for the US to put American lives at risk.
There will be a premium placed on startups that can leverage smaller models/less data/less training epochs for specific use cases. ASIC companies that can use older nodes in clever ways for inference/training will also benefit
In Part 1 of this two-part blog post, I highlighted the macro funding trends as the \u201CAI Wars\u201D between the US and China reach new levels of intensity. In Part 2 of this series, I explore why owning the entire semiconductor stack is crucial to maintaining the (seemingly) accelerating pace of innovations in AI, and how, until recently, investment in onshore semiconductor production hasn\u2019t been given as much focus given the steep curve required to catch up to the production economics in Asia. I contend that the western world\u2019s uncertain access to an unimpeded supply of semiconductors against the backdrop of broader geopolitical tensions in the Taiwan Strait creates a dangerous reflexive loop, with downstream ramifications in the broader commercialization of AI (which I deeply care about as an early-stage investor in the space). On the commercial side, losing access to semiconductors would mean a material slowdown in AI innovation. More broadly (in a geopolitical lens), this may force the major powers to take sub-optimal actions to protect their technological interests. This post is a bit more technical but was something I really enjoyed researching/thinking about. Let\u2019s dive in.
7fc3f7cf58