What is business moat

Moat can be defined as the time and energy needed to copy a business.

A business is often more than just its user-facing end product, but also the brand, unique distribution channels, existing enterprise contracts, legal contracts, patents and operational processes behind it.

The time and energy needed to rebuild the whole Amazon logistics ecosystem and global distribution by building a strong brand is way bigger than rebuilding an app like Todoist.

The moat of companies is dynamic and changes over time with new and improved tools allowing us to get things done more efficiently.

For example, large language models made it way easier to produce code; therefore, the time and energy needed to replicate software is drastically lower than it used to be before LLMs.

Also the overall moat of companies is often a combination of various sub-moats (Legal Contracts, Team, IP, Brand...).

Different types of moat

These are just high-level mental models. Don't think of each type of moat as actually being something different from other types of moat - it's all connected.
If you have an interconnected ecosystem, this mostly also leads to a corporate conglomerate moat. If you have a successful marketplace, this often leads to a brand moat, and vice versa...

Interconnected Ecosystem Moat

You could think that Meta, for example, has an interconnected ecosystem moat like Apple does because they own a number of social platforms like Instagram, WhatsApp, Threads, and Facebook.

But unlike Apple, Meta does not connect these different puzzle pieces in a way that makes the end-user experience better.
The user experience of WhatsApp or Instagram didn't become better after Meta's acquisition.
But the user experience of Shazam got better after Apple's acquisition because of its native integration into the operating system.

The interconnected ecosystem of Apple is insane:
First of all, operating systems like iOS or macOS themselves are interconnected ecosystems:
The work Apple did in building a great file manager, application launcher, system settings, system APIs, native development environments, native apps, etc., made the overall user experience better.

Additionally, Apple has an ecosystem of hardware products. If you own an iPhone, Apple Watch, Mac, and Apple TV, the overall experience is better than if you were to use fragmented hardware products from different vendors:
The iPhone leads to Watch, on Watch you benefit from a native Apple Wallet integration, shared notifications, Apple Health, Apple Homekit (there are hundreds of useful interconnections like this in Apple's ecosystem).

The important part here is that most of the subparts of Apple's ecosystem make the end user product experience better, you are not experiencing one single product but a whole ecosystem.

Tesla
Tesla’s moat builds up on a similar effect, as a Tesla customer you don't just experience solely the car but the whole ecosystem of supercharger availability, the Tesla wall connector for your home, the mobile app, and self-driving.

Amazon
Amazon's investment in logistics centers, their own driver network, and Amazon Lockers across cities has improved the overall shopping experience. Additionally, using Amazon as my primary online shopping platform simplifies order management within a single app.

Microsoft
Microsoft bundles VSCode, GitHub, and GitHub Copilot more and more into one ecosystem.
Windows bundles Copilot with Office.

Teams, Outlook, and Loop. The MS Loop vs. Notion distribution chart will look the same as the Teams vs. Slack chart in a few years, even with Notion and Slack being way better products than the MS counterparts.

Adobe
Adobe bundles the whole visual creation process into one ecosystem: vector illustrations, pixel editing, photo editing, video editing, UX/UI editing, video creation, photo creation - creating both interconnected ecosystem and habit moat.

Interconnected ecosystem moat is one of the strongest moats a business can have - because to offer an experience as good as the one of the current ecosystem you would need to rebuild the whole ecosystem.

Corporate conglomerate moat

I already mentioned that there are also a lot of businesses which bundle products in one organization that are not truly connected.

This can still make sense from an operational efficiency perspective or from a product distribution perspective.

If you're Meta and already running multiple social networks, acquiring another one could be strategically smart because you can use similar tech stacks, operational procedures and distribution strategies for all of them.

Same applies for companies like Unilever, Procter & Gamble, Nestle, LVHM, Walt Disney, Volkswagen which bundle multiple similar sub-products like in one organization because the operational and distribution processes of each of the single products overlap a lot.

Corporate conglomerates often lack strong interconnections between each of the products in the end-user experience.
Utilizing multiple products from the Unilever conglomerate simultaneously in my household does not enhance the overall end-user experience of a single one. Purchasing each one from various companies would result in the same user experience.

Therefore conglomerate moat is not nearly as strong as interconnected ecosystem moat because to copy one of the sub-products, you don't need to replicate the whole interconnected ecosystem to give the user the same experience. For Tesla to outperform the Volkswagen conglomerate, they didn't need to rebuild Porsche, Audi, Seat, Skoda, etc. They just needed to be better than the best carmaker of the VW conglomerate.

Proprietary dataflywheel moat

A lot of products, especially in the age of AI, get better the more data you can gather. Google's search engine improves every second by Google measuring what the world currently searches, and what the world searched over the last years.

The ability alone to gather a lot of data does not necessarily give you a data flywheel.

You need a to use the data continuously to improve the product.

Also the timelessness of the data you gather matters. If your product is great because it utilizes a lot of historical data + current data, the moat is stronger.
If the historical data matters less and the latest gathered data is the key ingredient.

Marketplace moat

Similar to the data flywheel, a marketplace improves with every user joining it. The moat of marketplaces is the users and creators who actively use it.

There are different types of marketplaces:

If you want to create new marketplaces, you either need to enable a completely new kind of supply or compete with existing marketplaces by offering higher quality in lower quantity.

Physical Logistics Moat

R&D / IP Moat

Engineering talent moat often leads to IP moat and vice versa.

If you are building a product in a regulated space or one that requires strong partnerships for it to function, securing these partnerships (ideally with a winner-takes-all arrangement like sports broadcast licenses that often go to a single service per country, or telecommunications partnerships like with the 5G network) can strengthen the moat of your business.

Headstart moat through right timing

It is evident that timing plays a crucial role in building an empire. To be more specific, timing matters significantly in the following ways:

Interconnected ecosystem moats are almost always built over time. By being the first company to create or distribute a product, you can then leverage your head start to continually expand your ecosystem.

This principle applies to various types of moats: interconnected ecosystem moats, IP moats, team moats, physical logistics moats, interface-as-habit moats, and marketplace moats. A lot of them grow stronger over time.

If you utilize your head start effectively and do not rest on your laurels, you gain more time than other companies. For instance, convincing local politicians to approve a second supercharger network is challenging if Tesla has already established one. Similarly, attracting talented engineers to join your company over OpenAI is difficult if OpenAI has a significant head start in building foundational models, acquiring customers, partnerships, and funding.

This dynamic underscores the value of startups. If your product bet or hypothesis proves correct, the risk of innovation can result in a significant competitive moat.

Your headstart as a moat can fade away when the technologies you built on become outdated.

We saw this with Tesla replacing the foundational principles German car manufacturers relied on. They built around internal combustion engines and formed the best manufacturing and engineering teams in the world. However, Tesla created better technology to solve the same problem, thereby rendering the German car manufacturers' foundational principles obsolete.

Creating your own distribution rules - by owning a platform

If you are the creator of a platform, you can create your own rules.

Even though the internet was conceived as a federated standard not owned by anyone, Google is creating the core products that most people use to use core parts of the internet.

Chrome Browser, Google Search Engine, Google Sub search engines (maps, images, shopping, flights, finance, books...) etc.

Google owns the search engine market. They can create their own rules and, for example, distribute their own products at the top of their search results or even integrate them natively as a widget.
Same with Amazon listing Amazon basic products at the top of search results etc.

Brand Moat

Elite universities are prime examples of brand moat.
Every great person enrolling at your university each year strengthens the brand, as students often consider the success of previous graduates when choosing a university.

Capital moat

Having access to large amounts of money can be a big advantage, especially in fields where big investments can take a business to the next level.

We saw this in the early days of the cloud industry. Companies like Microsoft, Google, and Amazon used their access to money by investing huge amounts into cloud infrastructure. No other company could do this on the same scale, which made their ecosystems even stronger.

The same thing is happening with Tesla. The company can invest more and more in building gigafactories around the world, increasing their production in ways that companies with less access to money simply can’t.

Switching cost moat

If the users have high rates of switching away from your product, you have a problem. Switching to a different social media platform when you already have a great audience on one of them does not make much sense. Switching from Adobe Photoshop to a competitor does not make much sense when you and your team have already learned Photoshop over the years and are not very familiar with it.

Language models in their current form have a small moat because of low switching costs:

Language models today don't have nearly the same moat as operating systems because they don't rely on proprietary standards. They function as single endpoints with low switching costs to switch to another model (taking about 5 minutes) and use a universally recognized standard: natural language.

This demonstrates that the time and energy needed to build a technology does not directly correlate with the strength of its moat. Other AI labs can train and distribute similar models; thus, the moat of a company like OpenAI is not in people building on top of them, but in having a head start in assembling the most talented AI research team in the world, raising significant funds, and building infrastructure for model training—which is harder to replace. Therefore, always having the best model in the world at the best price point will be the moat, rather than the fact that developers are building on top of proprietary standards.

If OpenAI uses their advanced models to improve subsequent model developments, this moat could become extremely strong. Otherwise, it remains a challenging moat to defend, especially compared to the robustness of an interconnected ecosystem moat like Apple's.

Habit as moat

Creating a habit moat is one way to increase switching costs.

Being an "industry-standard" to solve a specific task often means nothing more than being the solution that most people have trained their brains on.

Whether it’s the programming language you learned, the operating system interface you’re familiar with, Excel macros ingrained in your workflow, the medical equipment you trained on, the CAD program from university...

In a world of human knowledge workers, this moat is incredibly strong. However, as AGI starts replacing more human knowledge work, this type of moat may become less critical because AGI can quickly learn any novel solution in seconds and does not need months and years to learn it.

Also one might argue that companies could simply copy a competitor’s core interactions to minimize changes in users’ muscle memory. However, for highly complex products like SolidWorks, replicating every single interaction isn’t feasible. If you were to build a new SolidWorks, you’d likely start from first principles and current technology, resulting in different interaction paradigms.

Collaboration standard moat

Even as interface habits become less important due to the increasing role of generative AI, information standards will remain crucial for reliable communication and data transfer between multiple AIs. AIs will still produce prototypes in standardized tools and share them with other AIs.
Providers will continue to facilitate code creation based on these prototypes. Additionally, there will be more data for AIs to create Figma-to-code workflows than for any other tool, leading to better performance. We already see this with AI generating superior code in JavaScript and Python compared to less widely used programming languages and frameworks.

Products that leverage habit as a moat often also rely on being a collaboration standard. For example, when two people want to exchange money, both using PayPal simplifies the process. The same applies to using Figma, Google Sheets/Excel, and other collaborative tools. The world collectively builds habits around these standards, making it challenging for new products to replace established ones.

Examples:

Operating systems: The ultimate "featureizers" of personal computing

Many technical-founders solely focus on building a product but not on building an empire around the product.

That's fine if you:

What Does It Mean for a product to get “featureized”?

Being "featureized" means that your whole product becomes just one of many features of another meta-product.

For example:

While I was never really bottlenecked in providing features for Fixkey on macOS due to deep system APIs, building Fixkey for iOS would have been impossible the way Apple did it.

Apple's iOS does not offer many ways to integrate features deeply into their operating system, so you will be limited in your capability as a developer to provide the best experience.

For apps like Salesforce, SAP, or Figma, this is not the case because the data they handle is not, or only slightly, connected with the rest of my personal operating system data.

Given the importance of operating systems for personal computing but the increasing restrictions on customization for app developers, the Unix idea of a community-owned operating system might become more appealing in the future. Especially when going towards cloud computers which are not running on proprietary hardware anymore (personal cloud computing mostly applies to desktops though, which is a decreasing market).

Nevertheless, with LLMs and future AI systems making it easy to copy software, this is not only beneficial for proprietary vendors but also open source community-driven projects which could theoretically recreate proprietary operating systems.

Personal standalone Apps - a common victim of operating system integration

I cannot think of any business-empire which was built by creating a “personal standalone app”.

Let's define them:
For personal standalone apps, the core product is the app itself, not a physical service or proprietary information accessed through the app.

So you are not using the app to:

Excluding all of these what's left are Note-taking, Calendar, Contacts, Personal Memory Management (Rewind), Meeting Recording, personal CRM's, Browsers, Grammar Correction / Text Replacement, Bookmark Manager, File manager, and Mail apps.

These apps are solely handling personal data like your todos, notes, calendar entries, contacts, mails, bookmarks - and if you wouldn't mind syncing between devices, most of them could run fully locally.

If you are a startup building one of these kinds of apps where the main value you provide is the user experience itself, not something behind the app.

If your app is useful enough for a general audience, you can be sure it will eventually be rebuilt or integrated natively by the operating system. This is because:

Examples:

I think of the Operating System as the ultimate bundler of useful startups resulting in the closest you can be to an “everything company,” a “roman empire of technology", especially because I cannot think of a future where personal computers will not be one of the core products shaping the human experience.

Future scenarios where the operating systems companies might lose their dominance are:

Special case: standalone prosumer apps

We looked at a bunch of companies having strong moats. Now, let's examine companies with weak moats (to be clear it's a spectrum, not binary).

I also want to clarify that many of these are products I love using, so this has nothing to do with the product itself being bad, but is more about evaluating if the business has a chance in becoming an empire.

There are several powerful standalone apps for power users providing more powerful alternatives to more popular generalist apps, such as:

These apps are unlikely to be fully replicated by operating system vendors because they would be too complex for the general population to use. However, since they target a specialized, professional-consumer audience, the market size is often very limiting.

These startups make a bet that we will see the 10x superhuman developer equivalents in a number of verticals, so that single people creating the productive output of 10 people: 1) need power-user software, and 2) would pay 10 times more for it.

Empires to build:

Big fragmented markets fragmented markets - which don't have a clear standard (yet)

It's clear that the most powerful companies in the world can often be seen as ubiquitous standards for human needs:

So the obvious questions are: what empires are left to build?

Fragmented:

Potential zero billion dollar markets:

Obviously there are also markets where fragmentation is a feature, not a bug. Take restaurants and bars, for example: they should be fragmented because each bar or restaurant you visit should feel slightly different and unique every time. You don't want to build habits for entertainment experiences, as that would make them boring.

Deeptech

An obvious bet to make with AGI being right in front of us is that products which are currently hard to build will be easy to build tomorrow; products which are impossible today will be hard tomorrow.

Aside from hardware projects being harder to build, another dynamic is that there is generally less platform risk involved with hardware than with software. This is because, with hardware-first products, every startup operates within the same fundamental physical constraints and battles the laws of physics, rather than navigating artificial software systems governed by potential competitors.

B2B software

I think that advanced B2B software will continue to have strong moats at least for the next 4-7 years. It's unclear how this will play out with autocoders lowering the cost of software drastically and making it possible to copy any software out there just by observing it (including every update made to the software).

Unlike personal standalone apps, most B2B software at least would not profit from deep operating system integrations. SAP, HubSpot, Salesforce, or Celonis couldn't be used in a vastly more efficient way if they were integratively integrated into macOS, Windows, or the Chrome Browser.

Also, at least as of right now, they would be too complex to engineer (Figma, Solidworks, Adobe Premiere...) and run operationally (sales, legal, support) for operating system creators like Google and Apple. Even though there are good counterexamples with Apple's Logic as an Ableton competitor, or Apple Final Cut Pro as an Adobe Premiere competitor, or Windows Power Automate as a UiPath competitor.

Random thoughts