Happy New Year and welcome to our Q4 2022 update. We hope you're as excited about the new year as we are. Before we dive into the numbers, we want to assure you that this newsletter is hand-crafted with the utmost care. While we may have used some GPT-based tools and Grammarly to help us write more concisely, rest assured that the heart and soul of this newsletter are all our own.
In short, Q4 2022 was a relatively slow quarter for venture capital. Many investors (us included) spent more time tweeting about ChatGPT than actively investing. As a result, the number of new deals was at its lowest. Many startups postponed their rounds until spring and reduced the round sizes to smaller bridge rounds. Despite this, the macroeconomy shows signs of a possible future upswing, so the valuations might start rebuilding in the year's second half. Terms for early-stage investments will stay VC-friendly, so it's a good time to build a portfolio. We see huge potential for early-stage startups that effectively increase human labor productivity.
At DVC, we invested in just five new startups during Q4, but have a number of exciting deals in the pipeline and are planning to make some follow-on investments as four more companies from our portfolio are planning to raise. The market's slow pace allows us to thread more carefully while still avoiding the risk of losing deals to competition. We've accumulated over $2M in rolled-over investable funds, ready to move quickly when we see strong signals in the pipeline (such as Sequoia investing).
Our portfolio is now 41-strong and boasts an average of 18.5 months runway - more than enough to withstand the period of uncertainty.
We see a lot of opportunity with the rising tide of GenAI startups for DVC - we are not only uniquely positioned to see awesome startups early but are also actively improving our odds by recruiting Deal Captains and Scouts within OpenAI itself.
Our community continues to grow and prosper, adding new members. We're now back to active fundraising, aiming to reach a $2M per quarter fund size by Q3 so we can effectively follow-on on some of the best deals in our portfolio.
We also want to make sure we're having fun along the way. We just had Silicon Valley's best ugly sweater murder mystery party with some of our investors and portfolio founders, and we're already planning more events for this year. We will continue our quarterly party tradition and add a few more offline events, such as a dinner for GenAI founders and investors in early February.
Everyone and their uncle probably have already heard about the rise of Generative AI by now. You may want to know how well we are prepared for this technological revolution.
The rise of AI startups has been widely discussed in recent years, and it was clear that this technology was rapidly advancing. We’ve been keeping a close eye on the development sin this field since 2013 (after AlexNet, ImageNet, CUDA, and other acronyms that the AI founders of today might call prehistoric). In 2016 we launched Gagarin Capital I, LP, a $20M fund dedicated to investing in AI startups in the US, Europe, and Israel.
Our early investments have already seen significant success, with one of them, Prisma, generating $30M-$50M in revenues in just two months with its genAI product, Lensa. We have also seen exits from some of our other investments to companies such as Facebook, Google, Headspace, and Mitek, which have helped us to expand our professional network and expertise in the field.
As the President and co-founder of Cherry Labs, I have had the opportunity to work with a talented team, experimenting with cutting-edge technologies such as autoencoders, generative networks, few-shot learning, and early transformers. I even received one of the first API Beta invites for GPT-3, which we continue to use to develop DVC's internal tools (yes, we are one of the first, if not the first, VC firms powered by GPT). Through these experiences, we have built strong relationships with AI professionals, researchers, and engineers, many of whom are now investors in DVC.
All of this gives us the confidence that we have the knowledge and experience to identify and invest in some of the most promising AI startups. We are excited to continue positioning DVC to benefit from the coming generative AI revolution.
For a long time, Large Language Models based on a transformer architecture seemed to be a very promising but impractical technology: GPT-3 wasn’t very useful without a lot of fine-tuning, and too much fine-tuning seemed to work on it as a frontal lobotomy, making it noticeably dumber. Hence it was mostly used for narrow, usually non-conversational tasks, such as DVC’s “AutoVC” tool, which our team uses to write personalized rejection emails to startup founders.