A note to founders:
Right now, all the VCs want you to raise tons of money to pursue big swings in deep tech, hardware or neolabs. They want you to vertically integrate and re-imagine industries from first principles, unconstrained by time to commercial traction, technical viability or quantum of capital. They advertise the purest motivations to support your wildest aspirations. They cheer loudly as you enter the arena to do battle with the laws of the universe.
But, in exchange, they require you to craft companies that are maximally dependent on the capital markets they control. Along this path, each technical milestone achieved results in existential judgement. A thumbs up or down from the emperor and, with a thumbs up, evermore common ownership fed into the machine, ground to dust against the grindstone of asset management.
Don’t do this.
Build capital efficient software businesses instead.
Artificial intelligence automates problem solving from identification to solution. Now is the time to build applications to provide access to this intelligence.
The internet collected global knowledge and information. Software provided access to the internet and created previously unimaginable economic and societal benefits. This system of data, infrastructure and applications expanded human capacity with free and unfettered access to information.
Now, the very best founders should build the software that allows intelligence to flow where it is highest leverage, most impactful and necessary to support the next steps in advancing humanity across fundamental science, medicine, energy and the rest of the knowledge economy.
If you do this, you will
Maximize your impact on the world by evenly distributing the future you know is here
Maximize control of your own destiny and ownership in your company
Advance innovation more broadly (while re-establishing a healthy seed and pre-seed ecosystem)
If you’re ready to build this way, please reach out. If you’re not sold on software yet, more detail on my thinking below.
The math has changed
I was a Principal at First Round when Rob Hayes invested in Uber. The idea was to reimagine global transportation. The investment decision was simple, and it became one of the most epic seed rounds of all time. From what I remember, we partnered with Ryan, Travis and Garrett with $500K for 10%, invested another $1.2M when Bill Gurley led the A and owned enough at exit to return 20-30x the fund. This was only possible because of how capital efficient Uber’s growth was. The company creation arc that made this possible is no longer typical.
I think we’ve all seen the seed stage break since then. The math doesn’t make sense anymore, with hugely priced bets placed on businesses that need huge dollars upfront - often raised pre-product and pre-revenue. We’ve lost all sense of stages in general. But, if we’re going to see the benefits of artificial intelligence become accessible across society, it’s critical that a more traditional seed stage exists to fund more businesses that can make a real impact right now.
We need to stem the tide of capital consolidation and the monolithic judgement that comes with it. We need more founders building more diverse and interesting things. A healthy early stage economy supports AI distribution to impact more diverse industries and maximize the value creation of this technology.
This style of company building also works better for founders and the early team.
Seed investors with small funds focused on the craft of building early stage startups are the most founder aligned over the long run - we’ve seen that be true since the industry started. A small fund’s structure aligns the VC’s financial incentives - and views on company control - more closely with the founders.
What do you own when it matters?
Capital efficiency is the key to long-term founder ownership and seed stage returns. Entry price matters, but capital efficiency may matter more and this is where today’s market goes wrong.
The higher entry prices and larger initial round sizes support limited founder dilution (at least upfront) and larger swings with more audacious technical ambition. I love this. What I don’t love is the ongoing capital requirements of many of these businesses that end up creating punishing capital stacks for founders and early investors.
Founders who raise massive early rounds are structurally subject to the whims of future VCs. You’re placing your faith in this group to see the world the same way for a long, long time (not a strength for most VCs in my experience). Hardware and deep tech are hard. They take patience to bring to market, and over this time things can change. You could be out of luck downstream if the VCs don’t agree on the value of your progress, or if the market moves the metrics after you’ve walked through the one-way door of capital dependent tactical decisions.
Even if you do find believers in the capital markets, you may build a great company but not create meaningful value for you or your early stage partners. The COGS, forward deployed engineers, support and maintenance services, limited defensibility, low margin, complex growth dynamics of most hardware and deep tech companies create drag on the airframe of these companies.
The idiosyncrasy of valuing technical progress is a significant challenge today and these companies all require massive capital investments to get to scale. Even with dramatic technical progress against some of the world’s hardest problems, few offer a business architecture that support commensurate increases in valuation along the full journey.
Most insanity-provoking for founders in this situation, the capital efficiency and rapid valuation dislocation of hard tech emerge at scale through moats of operational complexity, technical breakthroughs/IP and economies of scale and/or scope. But you need to get there, and until the entire market converges on more consistent underwriting for frontier tech risk, the best investor entry point for this category of businesses is not seed, but late stage.
Founder choice and opportunity cost
You only get to live this life once. Let’s imagine you decided to buy into the mega-VC narrative and build a deep tech company instead of a software company. You and your team have delivered God’s work, and today the company is valued at $8B. You’re in the news and this headline looks awesome. You are supporting innovation that matters to the world.
But the capital required to get to this point is massive relative to a software application company - and the dilution much higher - maybe 65-85% post seed. If you did not get more options or financially participate in the later rounds that supported continued growth, even if you’ve increased the value of your company by 1,000x, the reality of the dilution along way results in around 20-30x for the founders and early partners.
Dance with the elephants
We’re living in a period of time with intense capital focus on infrastructure and model advancement. But, the real opportunity is in the applications that get built on top of the increasing stability in the layers below. Similar to Cursor, we’re going to see hugely valuable application layer companies that maximize the leverage available from artificial intelligence to create durable value built in the next 18-24 months. Now is the time to choose an end user you care about deeply and understand better than anyone else, create software that delivers intelligence to them and unlocks the opportunity only available to them in this future.
Do things the model providers can’t - either on value creation or cost reduction. Build companies that maximize learning per dollar spent. Focus on your specific users, within your domain expertise, who no one else will ever have more real-time data on than you. Serve your users in the most cost effective way. Build deterministic software to only consume intelligence at high leverage points as humans and agentic users traverse your systems. Capture context on this use and optimize and personalize for your customers as you scale.
The software you create will organize human activity, making it more efficient and less error prone. The integration of AI increases the level of resolution for understanding the problems your customers face and increases the scale of value you can deliver. Your moat lives in each customers’ world, encoded in digital processes that evolve to meet specific user demands as they arise. Your market of problems to be solved expands the deeper your understanding of your customer gets and the further it extends beyond what is available to the labs.
In our own portfolio we’ve partnered with founders focused on delivering cutting edge intelligence to the American soldiers serving at the front lines of global conflict, enabling individuals in the US to navigate the healthcare system with confidence, protecting critical data and individual right to privacy, reimagining advanced manufacturing and internet service provider operations, leveraging intelligence to reduce interest rates for consumers, advancing how videos are produced and the definition of interactive entertainment more broadly.
All of these founders have deep understanding and passion for their customers and are building systems that will compound over long periods of time as the scale of the problems they can solve expands with each new solution they deliver.
The cost structure of artificial intelligence forced software companies to re-think unit economics and account for the marginal cost of software delivery. But the cost of intelligence is not static. Open weight models and the software to allow easy access to advanced intelligence are racing to dominant positions for increasing percentage of workloads. Model routing and local inference will align user requirements, software workloads and intelligence infrastructure with minimal marginal cost. Success here could take us back to the holy grail of SaaS business models (arguably the most healthy and sensical models of all time) - predictable revenue, costs and margins for the software buyer and seller.
I am excited to invest in the next generation of founders pursuing “traditional software companies” because that is where the largest wave of intelligence distribution and value creation will be found.
If this is you, and you got this far, shoot me an email at phin@thegp.com. I’d love to hear about the customer you love and the software you’re building that they won’t be able to live without.