In the AI era, how to capture the next "pattern smasher"?

God Translation Bureau is a compilation team under 36Kr, focusing on fields such as technology, business, workplace, and life, and highlighting new technologies, new ideas, and new trends from abroad.

Editor's note: Don't let financial models limit your imagination. True winners dare to shatter old paradigms and reshape the growth curve of trillion - level markets in the AI era. This article is from a compilation.

As we all know, every major platform leap gives rise to a group of new winners. These new winners leverage brand - new technological infrastructures and business models to outperform financial models and general market expectations. We call these winners "modelbusters".

We are discussing "modelbusters" now because in the AI era, such companies will emerge in large numbers. AI is the biggest platform - level change we've seen in our lifetime. It will reshape consumers' lifestyles, create brand - new work models and business models, which really excites us. The more radical the changes in user interfaces, data acquisition volume, and business models, the more favorable this technological shift is for new entrants rather than existing giants.

Before exploring the specific opportunities in the AI era, we want to deeply analyze the concept of "modelbusters" in this article - including their definition, importance, forms of existence, and signs of success - to provide a blueprint for entrepreneurs and offer several intuitive tests to help you determine if you're building a "modelbuster" company.

Two Forms of "Modelbusters"

"Modelbusters" usually take two forms: one is a company with a Total Addressable Market (TAM) far exceeding expectations, and the other is a company that opens up more opportunities for its business by expanding its product line. These two forms are not mutually exclusive; some companies start with one and end up with both. Although not all "modelbusters" are category creators, almost all successful major category creators are "modelbusters". While we won't delve into the specific business details of each company here, we'll conduct a high - level analysis through several typical cases to give you an intuitive understanding of how and why they stand out.

Total Addressable Market (TAM) Far Exceeding Expectations

When a platform changes and a product with a 10 - fold performance improvement emerges, people tend to unanimously underestimate its speed and scope of popularization. Take Apple and Roblox as examples.

iPhone

When the iPhone was released in June 2007, Steve Jobs said it combined three separate devices into one and redefined the mobile phone. Nevertheless, many analysts thought it was just a niche product for high - end executives and tech enthusiasts, and even considered Apple's goal of selling 10 million units by the end of 2008 too ambitious. However, 15 months later, Apple completely shattered this expectation and became the world's third - largest mobile phone supplier, surpassing Sony Ericsson and LG.

Apple was perhaps the world's largest company at that time, and almost every analyst was tracking it, yet everyone still underestimated this opportunity by nearly three times. The financial models at that time overlooked three key points when predicting the market space: 1) The reimagined mobile phone experience expanded the audience beyond tech enthusiasts (it turned out that everyone wanted such a pocket - sized phone); 2) Apple built a powerful app store ecosystem on iOS, creating strong network effects and huge monetization potential; 3) The user experience of the iPhone was so immersive that it increased the threshold for users to switch to other phones and also motivated users to upgrade their devices annually for new features.

Roblox

When we first invested in Roblox, it was widely regarded as a rudimentary gaming platform for kids, obviously a niche area with limited lifetime value. But Dave Baszucki didn't see Roblox just as a children's toy; he saw it as a cloud - native developer platform and a virtual economy that could "grow" with its users.

He spent over 15 years building rich creative tools and a self - sufficient closed - loop market, enabling any user to become a creator and profit from their own experiences, with a cost of less than $0.10 per user - hour (or even free if they wish!). This brought about a revolutionary change in consumers' entertainment methods because its cost was only a tiny fraction of other entertainment options in the market, ultimately motivating more users to join the platform.

The early investment led to a powerful User - Generated Content (UGC) flywheel. As the number of players increased, developers had the motivation and tools to create more complex games for older age groups. What initially seemed like a children's audience turned out to be a springboard to the "over 13" group, whose daily active users have soared to 61 million, and the current growth rate is almost three times that of the under - 13 group (36% year - on - year compared to 13%). Now, they're even developing products for audiences over 21.

In short, this platform - based flywheel effect has brought better games, more engaged users, stronger cross - group expansion ability, and ultimately financial performance far exceeding anyone's model predictions.

New Products Open Up New Opportunities

Another type of "modelbuster" serves adjacent markets or captures a larger share of the budget from existing customers by launching new products. Let's look at CrowdStrike and Anduril.

CrowdStrike

As early as 2019, most sell - side analysts positioned CrowdStrike as an integrator in the cloud - native endpoint security market - essentially a replacement for Symantec or McAfee, with its market ceiling limited to traditional antivirus software (AV) users. But founder George Kurtz foresaw a grander future: he firmly believed that CrowdStrike could grow into a cross - category cloud - based security platform.

A few years before founding CrowdStrike, George had founded an antivirus software company and sold it to McAfee. He knew that traditional antivirus software was insufficient to defend against endpoint attacks. At that time, cloud computing was just emerging, and he bet that enterprises would need cloud - based agents for monitoring. So, he invested early in building the architecture to support these agents, sacrificing short - term profits for a unified data pipeline. Different from traditional antivirus tools that update signature databases regularly, CrowdStrike's agents continuously stream rich telemetry data to the cloud for real - time detection, response, and analysis, far surpassing the capabilities of local scanners.

This early bet not only helped CrowdStrike win the antivirus software market but also laid the architectural foundation for every security module it later launched on the Falconer platform (from threat hunting, identity protection to IT system cleanup). George was also obsessed with internal innovation to squeeze out competitors in adjacent markets; he would give engineers vacation time to conceive new products and award stocks to the winners. While competitors had to piece together their product lines through mergers and acquisitions, CrowdStrike organically developed 29 modules on its own. Since these modules share the same architecture, CrowdStrike only needs to click a button in the console to enable them, without additional installation or integration, making cross - selling extremely easy. Ultimately, this allowed the company to redefine the originally narrow antivirus category as a more powerful Endpoint Detection and Response (EDR) market.

The rewards are reflected in the data: Currently, 48% of customers are running 6 or more CrowdStrike modules, contributing about $500 million in net annual recurring revenue (ARR) growth each year without adding any new customers. CrowdStrike's expansion into other products has significantly expanded its potential market and extended its ability to maintain high compound growth, resulting in revenues far exceeding sell - side expectations - growing at more than twice the rate predicted by the models.

Anduril

Some "modelbusters"' innovation engines don't just open up adjacent markets or new customers; they unlock huge opportunities that others can't even capture. Take Anduril, for example. They used Silicon Valley - style product innovation, a software - native business model, and in - depth understanding of the procurement process to outperform other startups and force their way into the extremely conservative U.S. Department of Defense (DoD), unlocking a $3 trillion budget in the process.

Before Anduril, the burden of defense innovation fell on the U.S. Department of Defense: the DoD outlined the products it wanted, invited contractors to bid, and paid them on a "cost - plus" basis. Unsurprisingly, this led to a lot of budget inflation, inefficiency, and project delays. Anduril completely overturned this model. They developed products on their own without a DoD contract and then actively promoted them to the DoD.

Although this "product - first" thinking exposed Anduril to significant upfront risks (the DoD wouldn't pay for their R & D), it closely linked the company's economic benefits with rapid innovation and iteration, ultimately providing the DoD with far more innovative products than before. It also allowed the founder to follow the cloud trend and build their autonomous system architecture, Lattice, into a cloud - native software platform compatible with various sensors, spanning data centers and "tactical edge" sites. By streaming rich sensor information from different products into Lattice, Anduril achieved large - scale real - time AI model training, mission planning, and software updates, which were impossible under the old "one - time deal, cost - plus" model.

When selling products to the DoD, Anduril's founders used their experience in the defense procurement process accumulated at Palantir to secure "Program of Record" status for multiple product lines. This allowed the company to sell new products through streamlined procurement channels, obtaining distribution leverage that other startups could hardly match. Moreover, since all of Anduril's products run on the same cloud - native Lattice operating system, they can cross - sell upgrade packages, intelligence services, and new autonomous functions across projects, transforming fragmented contracts into a unified revenue stream across multiple DoD procurement lines.

To this day, Anduril has successfully cross - sold more than 35 of its products into artillery, intelligence, surveillance, and reconnaissance (ISR), and electronic warfare budgets, with many more spending categories waiting to be tapped.

How to Determine if You're Building a "Modelbuster" Company?

Although the above companies are in completely different fields, they reveal the commonalities of all "modelbusters": products that focus on "what" rather than "how"; a "pull" rather than a "push" market; network effects in the best - case scenarios; and strong compound growth.

"What" Over "How" Products

Although spreadsheets and financial data can indicate whether you're on the path to becoming a "modelbuster", they alone can't confirm it. You also need to have insights into the market, be immersed in the product, and be proficient in consumer behavior.

Typically, "modelbuster" products are "what" products rather than "how" products. "What" products fundamentally reimagine existing products or offer completely new capabilities. Think about the iPhone, Anduril's creation of new - type defense products without any modern competitors, or Roblox's cloud - based developer platform and innovative gaming economy. "How" products mainly innovate in the delivery method. For example, being the first company to sell a certain brand of shoes online. You may have a slight competitive advantage now, but it will disappear as soon as others follow. Focus on "what" products: it's extremely difficult for others to compete with underlying innovation.

Pull, Not Push the Market

Customer love is one of the best indicators to predict whether a company can grow into a "modelbuster". While it's important for founders to have strong opinions on how the product can please customers, we prefer to hear from customers that this vision truly resonates with them. In fact, David George had a sticky note on his computer that said, "Is the market craving more of the product?" If so, you may be a company that "pulls" customers rather than "pushes" products onto them. If customers are attracted to and love your product, they're likely to stay and buy more.

This customer love will ultimately be reflected in top - notch gross retention and net retention rates. CrowdStrike's data is a good reminder: on a base ARR of over $4 billion, the company has a 97% gross retention rate and a 112% net retention rate.

Network Effects

Network effects are the most powerful form of competitive advantage: they expand your market leadership by building a competitive moat. Every new user, data point, or integration makes the product better, reducing customer acquisition costs and accelerating popularization at a rate far exceeding linear model predictions. (Just look at Roblox.)

Network effects are not a prerequisite for "modelbusters" as they are usually associated with transaction platforms. If a company doesn't show network effects, a strong market leadership position, brand influence, and excellent go - to - market (GTM) capabilities are the next - best competitive advantages.

Strong Compound Growth

The magic of "modelbusters" lies in the power of compound growth: they grow faster and for longer, exceeding any financial model's expectations. In the long run, driving faster revenue growth is far more effective in increasing a company's current valuation than pursuing higher profit margins.

If we use the traditional "Rule of 40" framework, assuming a company can trade one percentage point of profit margin for one percentage point of growth and standardize the market's premium for growth, then each percentage point of growth is worth about twice as much as a percentage point of profit margin.

This doesn't mean founders should only focus on growth. But today's high valuations ultimately come from the expectation of high - speed future growth.

The Next Generation of "Modelbusters"

As we mentioned at the beginning of the article, we believe the opportunities in AI are huge. We've seen that AI infrastructure construction is moving towards an investment goal of over $1 trillion in 5 years, which is larger than the U.S. shale oil industry in the early 2010s, the broadband construction in the late 1990s, or the Apollo moon - landing program in the 1960s. This generation of AI companies is the fastest - growing in the world, and we're starting to see some business model shifts, which makes us firmly believe that more creative models will soon emerge.

Translator: boxi.