How the Cost of Code is Disrupting the Venture Industry.
Thoughts on the impact of a 1,000x reduction in the cost of validating and launching software products.
Hi all! If you’re already a subscriber feel free to dive right into this week’s edition.
If you’re new, a warm welcome and thank you for subscribing! This newsletter shares the best VC wisdom and knowledge curated from seasoned investors, researchers, and technologists.
You’re part of 200+ industry insiders who’ve joined along on a journey to mastery & impact.🚀
When Mark Suster was a founder in 1999, it took $5m to build and launch his internet startup. That’s $9m in 2023 if you adjust for inflation. His team had to spend money on UNIX servers, enterprise-grade disk drives, various database and software licenses, onerous long-term contracts with infrastructure providers, and expensive engineering talent.
![](https://substackcdn.com/image/fetch/w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa682ceb6-faab-4867-aea3-f2f0d4164a11_1280x836.jpeg)
By 2011, the cost to launch a similar software product had shrunk 1000x from $5m to just $5k. Free open-source technology (e.g. Linux, Apache, MySQL) and pay-as-you-go cloud services (e.g. AWS) eliminated the need for significant upfront infrastructure costs. This made it easier and faster to launch web products. Instead of hiring dozens of engineers to code and maintain a web app, a small team could achieve the same result with open-source frameworks such as Ruby on Rails, which abstracted away the complexity of building interactive apps.
Today, one engineer can develop a basic version of a platform like Netflix in just three hours. Similarly, a solo developer can create a functioning prototype of a service akin to Uber within the same time frame. Even a teenager, given a few days, can build an operational e-commerce platform resembling early Amazon. This feat probably consumed most of Jeff Bezos’s $1 million seed funding some 30 years ago, illustrating the remarkable advancements in software development.
AI is now igniting a further shift in the speed and cost of building software. GitHub co-pilot – a GPT-powered auto-complete plug-in for writing code – makes software developers 55% faster. Large language models also make it easier to build AI applications. It used to take the AI veteran Andrew Ng and his team around 6-12 months to develop and deploy commercial-grade machine learning systems. Thanks to LLMs, engineers can do the same work in days.
![](https://substackcdn.com/image/fetch/w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff085fa82-f5b7-4653-aee7-06ad428e6925_1282x730.jpeg)
I’ve personally experienced some of this momentum. It took me several evenings and hours to build a custom GPT bot six months ago. But earlier this month, the same work took minutes. I’ve used GPT to write simple apps (e.g. Midcafe) and code that has automated away hours of professional work.
We are seeing unprecedented changes in how quickly software ideas can be validated and deployed. Moreover, the opportunity for software to improve our world is larger than ever. Vast areas of the economy are yet to be digitised, and 2.7 billion people are still offline (see the charts below for the opportunities ahead.)
![](https://substackcdn.com/image/fetch/w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F67e97d4f-a145-4791-b371-e007ebad3e01_1282x1606.jpeg)
![](https://substackcdn.com/image/fetch/w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe93194ed-893e-4643-bd9e-c12408d7f0d2_1600x900.png)
![](https://substackcdn.com/image/fetch/w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F2a1e91e2-5978-4654-93ee-f7fed874f284_1284x974.jpeg)
What does ‘cheaper’ & ‘faster’ code mean for startups?
Tech founders can go from idea to prototype faster than ever. This benefit is available to all entrepreneurs, so expect (a) more competition and (b) smaller early-stage funding rounds. The exceptions to this are deep tech startups or ventures that require long gestation periods of R&D (e.g. OpenAI was founded in 2015, yet its flagship product ChatGPT launched seven years later.) For everything else – particularly in areas where the entry barrier is mostly commonplace code – expect more early-stage competition and modest funding rounds.
![](https://substackcdn.com/image/fetch/w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fea9d2746-9dcf-4e83-ae58-73a260ca233b_1192x660.png)
That said, building and scaling a company will remain formidable. Once you graduate from a proof-of-concept, you must recruit, manage, and retain the best talent. Teams make companies, not individuals. You also have to nurture relationships and partnerships to drive growth and distribution. Founders who have deep knowledge and networks in their markets will have a greater advantage here. In addition, you have to establish business operations and playbooks that scale effectively. All this work isn’t going away. Maturing startups will still need significant capital, networks, and scaling experience to be competitive – areas that VCs can support.
What will happen to venture capitalists?
I see two implications for VCs as the cost of code plummets. First, early-stage funds might bifurcate between those who adopt a high-volume/small-ticket strategy and a low-volume/large-ticket approach. Both are responses to an environment where technical entry barriers are shrinking and more software entrepreneurs are coming to market.
A high-volume/small-ticket strategy would back lots of small teams that can build and validate products fast and get to market quickly. (Chamath Palihapitiy recently highlighted this possibility on an episode of the All-In-Podcast.) An example of what winning looks like here is Instagram. It had just 13 employees (and 30 million users) when Facebook acquired it for $1 billion.
Instagram is, of course, an outlier. But venture capital is an outlier business. So, if you can fund more startups by writing smaller cheques without diminishing what products a company can build, it makes sense from a returns perspective to have more shots at goal in the pursuit of outliers.
![](https://substackcdn.com/image/fetch/w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fc2f6ebac-2f40-49ab-8888-af1ecb5b5ea5_1416x708.jpeg)
The low-volume/large-ticket approach would do the opposite: Back fewer teams with larger cheque sizes and focus on areas with significant technical or operational entry barriers. One venture fund that’s succeeding at this already is Lux Capital. It has a 9-person investment team managing $5bn+ of venture capital assets. They invest primarily in science-driven companies founded by entrepreneurs who they consider “equal parts mad scientist and business mastermind.”
The low-volume/large-ticket approach can generate significant returns but requires a large fund, comfort with exceptionally technical areas, and possibly longer investment horizons.
Aside from the bifurcation of investment strategies, there’s one more bit of disruption in venture that I’m still wrapping my head around. The industry has always been small in terms of headcount. A typical fund in the UK has six investors and three support staff. The US numbers are similar: the median number of employees is six and the average is 14.5. These figures may shrink. Not just because of the usual cyclicality of booms and busts (we’re currently coming down from a peak), but also because of automation and technology.
I already shared an example of how software developers can write code 55% faster thanks to GPT. The impact here is that startups can build more with smaller teams. But even strategy consultants – a different segment of knowledge workers – are seeing gains. An experiment that involved 700+ employees at the Boston Consulting Group found that workers who used GPT completed a portion of their work 25% faster and with a 40% boost in quality relative to workers who did not use AI. I expect that VCs who leverage these new tools will also do their jobs more productively. This will undoubtedly change the nature of work in venture capital.
What are your thoughts on what comes next for VC, and what should we do to prepare for these changes?
Let me know what you think in the comments below!
I forgot to add: Later stage VCs will be automated faster than early-stage VCs. There's more data at later stages and more data = more automation. This is something Aydin Senkut (https://twitter.com/asenkut) points out on this podcast https://youtu.be/iMBc0DIwc24?feature=shared&t=3425