The End of Early Stage Venture

Early stage venture is doomed.

If your inner monologue just exclaimed “wow, that’s a bold claim” — you’re right! However, after talking with hundreds of fellow founders, operators, and investors during our weekly GenAI Collective community events, I have yet to hear a compelling counter to this claim in light of its supporting arguments.

To get a better sense of where I’m coming from, let me walk you through the seismic shifts I’m seeing playing out in Silicon Valley right now and what happens after the death of early stage venture as we know it.

Early Stage Venture 1.0

Since the inception of software venture capital in the 1980s, investors have been taking the same fundamental gamble: invest in the 1% of companies with the potential to generate 100x+ returns by reaching a $100M+ run rate and $1B+ valuation, paving the way to an exit via initial public offering (IPO) or large scale M&A. Decades later, startups seeking venture funding face the same implicit requirement to tackle a $1B+ total addressable market (TAM) with a T2D3 (“triple, triple, double, double, double”) growth trajectory.

In 2024, this standard venture economic model no longer applies. Why? Underlying every capital allocation is a cost-benefit analysis — and while investors are chasing the same benefit as they were in the 80s, the cost to build software has been dropping precipitously. As generative AI promises to intensify this flame, even Silicon Valley’s most successful (and stubborn) stalwarts won’t be able to neglect this new reality.

A New Reality

Software companies in the Bronze Age of Silicon Valley faced significant headwinds scaling their businesses. Even as personal computers began to reach commercialization in the 1980s, software distribution was entirely physical as the internet didn’t achieve considerable adoption until the late 1990s. It wasn’t until the Gold Rush of the Dot-com bubble that low-level programming languages such as Fortran, C, and C++ finally gave way to high-level, object-oriented languages including Java, Javascript, and eventually Python.

The Golden Age of the past ~20 years has ushered in a series of revolutions in software development and distribution. In mere days and for under $100, founders can now stand up a fully-featured full stack web app on their own domain using a cloud computing service like Vercel and get that app in front of thousands of eyeballs using Google AdWords. Oh, and don’t forget that Microsoft, Google, and Amazon are all giving away $100k+ in free credits!

Generative AI is the final piece to this new reality. While Github Copilot promises 55% efficiency gains, dozens of ambitious startups such as Poolside, Magic.dev, Marblism, and Cognition promise a transition to fully autonomous coding systems, obviating the need for humans altogether. With over a month of hands-on experience using Cognition’s Devin coding agent, I can personally vouch for the efficacy of this new paradigm.

If we take this observation to its limit, it’s not an if anymore, but when software development becomes fully commoditized. According to my most trusted sources, this happens in the next 2–5 years, if not sooner.

We Have No Moat?

In this new reality where technical execution ceases to be a competitive moat, how will software companies build products which capture value in the market? What will still matter? I hypothesize that the following two areas will be the difference between success and failure for the next generation of software companies:

• Product-market fit (PMF): achieving alignment between the value a product offers and the demands of the market. The product must also provide differentiated value when compared with alternatives (competitors) and the customer must have a high willingness to pay. You can build all the software in the world, but if you don’t solve a customer’s problem with a high willingness to pay then you’re not going to capture any value.
• Go-to-market (GTM): designing and executing a plan for sales and distribution that ensures your product reaches your intended audience with the lowest amount of friction and communicates its value effectively. You can build the best software in the world that exactly meets your customer’s every demand, but if you fail to get it into their hands, then you’re not going to capture any value.

Now, you might be saying “sure, sounds reasonable in the short term; however, what about when we have AI built into our phones which can generate apps for us on-the-fly according to our exact preferences? Why will we still need software companies at all?” While that is an intuitive assumption, I’d like to draw an interesting parallel which demonstrates otherwise.

The Democratization of Media

When broadcast television rose into prominence in the 1950s, the “Big Three” dominated the airwaves as cost and complexity created an extraordinarily high barrier to entry. Since then, mobile phones and social media platforms have eroded the barriers to development and distribution, birthing a $250B+ long tail creator economy — threatening traditional media in the process.

At over 14 billion views, “Baby Shark Dance” is the single most viewed video of all time. While it is by no means a visual masterpiece, it found product-market fit in a way no piece of traditional media had ever done before.

Posing a similar question to the above, what happens when media becomes ostensibly free to develop and distribute? Well, that reality is already here. Announced last Wednesday, Udio can generate music indistinguishable from music produced by human artists. This democratization of music will lead to the death of the music industry, right?

Not so fast. While generative music may take a share of the pie, I’m confident that it will never swallow the whole pie for the following reasons:

• Paradox of choice: An infinite menu of options sounds great until you actually have to pick an option. Curation increases satisfaction.
• Loss aversion: People prefer certain outcomes as opposed to uncertainty, even if strictly rational thinking dictates otherwise.
• Laziness: Prompt engineering requires deep thought and expertise. People want the more convenient option, even if it’s suboptimal.
• Human connection: People crave a shared human connection with the craft which AI can never provide.

What is the antidote to all the concerns above? Having someone else create the experience or product for you. People often don’t know what they want until you show it to them. Even in a world where content is free to produce, people need tastemakers.

Early Stage Venture 2.0

As the economics for software change and the long tail opens up, Silicon Valley is experiencing a micro-SaaS revolution. Sam Altman is openly predicting the world’s first single-person unicorn ($1B+ valuation). Anecdotally, I’ve talked with a higher concentration of bootstrapped founders and indie entrepreneurs over the past half year than ever before.

The early stage venture 1.0 model that has been in operation since the 1980s will fail in this new paradigm where software development is commoditized. Much like traditional media behemoths of the past, the venture landscape is facing a generational reckoning and must adapt in order to survive or risk radical disruption.

Beyond generating returns, venture has always served the noble purpose of helping founders close the gap between technical feasibility and market adoption. With technological progress accelerating faster than ever before, how does venture continue to fulfill this mission?

Venture’s solution up until now has been a deepening focus on providing more and more tangible value to founders. Consider the rise of venture incubators which have seen a 10x increase in the preceding decade, or platform funds that provide a rolodex of operators ready to fill the gaps necessary for efficient go-to-market. However, these small changes to the venture model don’t go far enough. Are they really substantively helping startups grow beyond providing a community of experts, GTM seminars, and emotional support? Not really.

The startup economy needs a first principles rethink. Venture needs to shift to helping the tastemakers of the future solve PMF and GTM at scale — a problem set AI may be farther away from addressing effectively.

What about the longer term? What happens if or when AI can fully automate PMF and GTM, thereby unlocking infinite scalability? I guess we’ll just have to wait and find out. 😉

Pro tip: you can give up to 50 claps for an article on Medium! Just click and hold the clap icon for few seconds and watch the magic happen! 😉