Generative AI Companies Have Moats (Eventually)
A few weeks ago, I had the pleasure of moderating the generative AI panel at CloudNY, an annual conference for cloud CEOs co-hosted by Battery and Firstmark.1
In conversation with Aparna Dhinakaran from Arize* and Victor Riparbelli from Synthesia, our panel explored how to get started with large language models (LLMs), how to address common challenges in deploying these models and more.
All very timely issues — but we had the most fun chatting about the obsession that VC has with moats.
As Victor put it:
“Moat is a VC concept… for founders, yes competitive differentiation matters, but focus on building something people love, it’s a very big world out there”
I agree with Victor! The focus that most VCs have on moats for generative AI businesses seems misplaced.
In business parlance, a “moat” refers to a competitive advantage that empowers a company to outperform its rivals and protect its market share over time (just like how moats of water protected medieval castles).
In his book Zero to One, Peter Thiel identified four types of moats:
Proprietary Technology: A unique technology that no other company possesses. To offer a significant advantage, a product should be at least 10x better than its closest substitute.
Network Effects: The phenomenon in which a product's value increases as more people use it.
Economies of Scale: A cost advantage that businesses obtain due to their size, output or scale of operation.
Branding: A strong brand, built on the perception of value in the customer's mind, can also serve as a moat. It extends beyond the physical attributes of a product.
Now, if you're a budding startup with fewer than 20 employees, how could these four types of moats apply to your business? Is the product you started building less than a year ago a unique technology that outclasses its substitute by a factor of ten? Do you have network effects from a limited user base, economies of scale without the scale, or a brand moat when you’re virtually unknown?
What about mature cloud companies?
Network effects can occur in social media companies and marketplaces, where additional users are more impactful to any given user’s experience, but they are rare in the cloud: another company using the same CRM as me doesn’t impact my experience that much. True proprietary technology is rare in application software, and many software companies eschew patents because innovative features are easy to replicate. Brand is a common moat, but a weak one. “Nobody gets fired for buying IBM” was true twenty years ago, but you don’t hear that phrase as much these days.
Yet, venture capital continues to pour into cloud businesses. Why? Because the cloud business model reliably produces one important moat: economies of scale.
Unlike manufacturers of tangible goods such as furniture or shoes, cloud companies don't incur proportionally significant costs as their user base grows. When Nike sells a shoe, a substantial amount of the revenue needs to be channeled back into producing more shoes. For cloud companies, the cost associated with allowing a new customer to use their software is minuscule — a mere uptick in their cloud-hosting bill.
As a result, more revenue can be directed towards product development and R&D. And the company’s entire customer base can benefit from this increased investment due to the distribution model of the cloud.
Accumulated R&D investment is not a moat in itself — a company with a substantial balance sheet can pour the same amount of capital into a competitive offering — but the learnings from users experiencing a product cannot be copied. Very few products are built flawlessly from the start. Instead, they evolve and improve based on user feedback, and this accumulated insight on how to best serve users becomes a powerful advantage.
Aggregate R&D investment and usage data are accumulating advantages for cloud businesses. Cloud companies benefit from more straightforward economies of scale — e.g. the spreading of infrastructure and development costs across users and other operational efficiencies — but the advantages that scale offers for thoughtful, user-informed product building creates a powerful moat for cloud businesses over time.
Let's take a journey with a hypothetical generative AI company, LobbyGPT, to illustrate this idea.
Lobbyists are often overwhelmed by the sheer volume of proposed legislation and struggle to efficiently parse, understand and draft responses to new bills. LobbyGPT uses a generative AI model to scan and summarize legislation, suggest potential impacts for particular companies and provide a foundation for drafting responses.
In its early stages, LobbyGPT identified a key group of power users who provided invaluable feedback. For instance, they highlighted a desire for customizable parameters that could reflect their latest areas of interest and their company’s evolving stances on key issues. This feedback was critical in shaping the early product development roadmap.
As the LobbyGPT user base grew, the economies of scale inherent in the cloud-based business model became more apparent. With its cloud infrastructure already in place, the incremental cost to the company of adding an additional user was relatively low. This allowed the LobbyGPT leadership team to reinvest a significant portion of their revenue back into the product in the form of R&D.
LobbyGPT refined and enhanced its product based on the accumulation of user feedback, introducing advanced features such as real-time legislative alerts, automated stakeholder analysis and a platform for collaboration and secure sharing of drafted responses.
It's important to note that LobbyGPT doesn’t experience traditional network effects: A lobbyist using LobbyGPT at another company doesn’t necessarily improve someone else’s personal experience. And, while the company’s generative AI is sophisticated, the technology itself is not unique or especially challenging to replicate. LobbyGPT is cultivating its brand, but that’s not the primary reason lobbyists choose them.
What sets LobbyGPT apart is the iterative improvements the company has made based on user feedback and the expertise it’s acquired serving this niche market. Given the wide range of thoughtfully built features developed by the LobbyGPT team, and the experience the team acquired serving this persona, it would be difficult for a new challenger to replicate the company’s success.
When the company went out to raise its seed round, LobbyGPT was viewed merely as "a thin wrapper on ChatGPT.” But over time, the company cultivated a unique value proposition through extensive R&D and customer-driven improvements. The depth and breadth of its expertise in this niche market created an accumulated advantage that now serves as a powerful moat for LobbyGPT against potential challengers.
Obviously, not every generative AI company is destined to succeed. But if a company fails, the reason is not likely to be because “OpenAI released this” or “(insert existing cloud company) copied them.” It’s likely because the company never found product/market fit in the first place, or if so, the team didn’t execute on incorporating market feedback to build the product that customers wanted.
The best product/market spaces are competitive. You may start a company with a proprietary insight, but over time, it is very rare to maintain proprietary technology. That was true in the cloud wave of the past fifteen years, and it is OK that this continues to be true for generative AI companies.
While differentiation may be minimal at the outset, this should not deter companies from taking the plunge. In fact, it emphasizes the importance of early entry into the market. The sooner you can start accumulating an advantage, the harder it will be for others to catch up.
The information contained herein is based solely on the opinions of Brandon Gleklen and nothing should be construed as investment advice. This material is provided for informational purposes, and it is not, and may not be relied on in any manner as, legal, tax or investment advice or as an offer to sell or a solicitation of an offer to buy an interest in any fund or investment vehicle managed by Battery Ventures or any other Battery entity.
This information covers investment and market activity, industry or sector trends, or other broad-based economic or market conditions and is for educational purposes. The anecdotal examples throughout are intended for an audience of entrepreneurs in their attempt to build their businesses and not recommendations or endorsements of any particular business.
Content obtained from third-party sources, although believed to be reliable, has not been independently verified as to its accuracy or completeness and cannot be guaranteed. Battery Ventures has no obligation to update, modify or amend the content of this post nor notify its readers in the event that any information, opinion, projection, forecast or estimate included, changes or subsequently becomes inaccurate.
*Denotes a Battery portfolio company. For a full list of all Battery investments, please click here.
You can see the panel intro video here featuring “AI Matt Turck” — though I must say, I thought his joke about me being the second best AI investor in NYC was rather unaligned!