You can see the appeal. In 2025, Spotify said 200 million users (+19% year-over-year) shared Wrapped over 500 million times (+41% year-over-year) within 24 hours1.

Naturally, the copycats follow. YouTube Recap, Strava Year in Sport, Duolingo Year in Review, Granola Crunched, Grammarly Writing Wrapped, Apple Music, Amazon Music, LinkedIn, Uber, Tiktok, Google Search, Twitch, Discord, Reddit, the list goes on.

But why does Spotify’s version generate significant anticipation and engagement, while the others struggle to gain traction? For comparison, YouTube has nearly 4x the number of users (YouTube 2.7B2, Spotify 700M3) yet Spotify Wrapped generated ~10x more interest.

Product Or Campaign?

Adam Fishman, who writes the newsletter FishmanAF, spoke with former Spotify’s Head of Growth Marketing Patrick Moran about what separates Spotify from the rest4.

“To do well, this takes months to plan out because it is just as much work for product and engineering as it is for marketing.”

“In reality the Wrapped campaign is really seen as a “product” at Spotify and gets way more resources, support, budget, etc. than a one-off campaign would.”

If you want results like Spotify, you need to invest in it like Spotify does. Treating your year-end recap as a holiday marketing campaign will not produce the same outcome.

Before You Ship

Adam’s piece includes five questions product leaders and CEOs should evaluate before committing to a year-recap feature. Building on those, I recommend three overarching questions.

1. What is the Goal?

What do you hope to accomplish with this feature is probably the most important question.

The goal could be brand awareness, or customer acquisition and retention. It could be feature discovery to remind users about underutilized capabilities like Spotify’s playlists or podcasts. Or you might want to reinforce the value or ROI your product generates, as Cursor does by showing users how many lines of code they generated. Sometimes, the goal could simply be fun and vibes, which creates goodwill in harder-to-measure ways. Each goal leads to different design choices and strategies. For example, NYT Games makes their year-end recap accessible to free users, but shows ads since the high engagement makes the impressions valuable. Others like Strava locks theirs behind the subscription paywall as a customer acquisition play.

2. Will users be happy to see such personalized usage data?

Some products generate data users genuinely want to reflect on. Good candidates include music listening, fitness achievements, learning streaks. These activities showcase effort, taste, and identity. Bad candidates include job applications submitted, fast food orders, or hours spent scrolling. These activities carry shame or anxiety. Dating apps would not want to remind users of how many failed matches they accumulated or food delivery apps would not want to remind users they ordered from McDonald’s 300 times last year. In fact, Saturday Night Live lampooned exactly this dynamic in a hilarious sketch. Products serving multiple user types must also segment their recaps. Spotify wisely gives artists different data than listeners while it seems like LinkedIn’s 2025 Year-In-Review did not differentiate between creators, recruiters, and regular users.

3. Will users get social validation from broadcasting this?

Spotify’s Wrapped succeeds in part because it creates social capital. Being in the “top 1% of Swifties” gives users something to brag about and bond over belonging to the tribe. Compare that to LinkedIn’s Year in Review, which tells you that you used “Applicant Insights” frequently. As one user in Lenny Rachitsky’s Slack community5 put it: “I actually have no idea what the applicant insights feature is that apparently I used or the connection between views and searches and premium.” Just because a feature name is well known inside your company does not mean it has similar recognition in customers’ minds. If users don’t even understand it, they won’t get any social validation from broadcasting it.

Unintended Consequences

Some products have more surface area for pitfalls than others. Personalized data that works for some users can become problematic for others. For example, a fitness app highlighting workout streaks works great until a user goes through a difficult period, such as illness, depression, or major life changes, and the data reminds them of a decline they would prefer to forget. A year-end recap that highlights low usage creates the same problem, effectively calling out customers for not exercising enough. Some pitfalls cannot be predicted in advance. When they surface, teams need to monitor social channels and respond quickly. As , a former Facebook executive, noted6:

For products with high risk of surfacing unwelcome personal data, pivoting to aggregate information provides a safer alternative. Google, Netflix and TikTok year-end recaps lean toward showing users what was popular across the platform rather than potentially embarrassing personal statistics.

Does year-recap make sense for B2B or work-related products? Generally, no. Most people would not want to know they spent 4,000 hours in Jira, especially if it reminds them they did so because Jira’s errors forced endless troubleshooting. Yet Granola, an AI meetings tool, received notably positive reception for its “Crunched” year-end feature despite being work-related. In the same Lenny’s Slack community discussion from earlier, users reacted to Granola’s ‘Crunched’ feature positively, a stark contrast to their reaction to LinkedIn’s recap. “Just in the Lenny community we have so many people talking positively about Granola. In a crowded AI meetings tools market, that’s massive.” versus “The LinkedIn one was awful. Graphics were bad, wasn’t fun, wasn’t useful.” The difference may be that Granola focused on entertaining AI-generated summaries with simple/minimal raw statistics.

The Missed Opportunities

Some products that seem like natural fits for a year-end recap feature are yet to ship one. Kindle and Goodreads are the most obvious candidates to me. Amazon owns both properties and has rich data about reading habits. Yet users have had to resort to creating their own year-end Kindle and Goodreads recaps because the official product does not exist. This seems like a missed opportunity because reading, like music listening, reflects identity and taste.

LinkedIn’s daily games have built an engaged audience, but the 2025 LinkedIn Year-In-Review omitted gaming statistics entirely. For players, like me, who solve Queens and Tango daily, games data would have provided more shareable content because competitive streaks and leaderboards already drive engagement7.

The Copycat Tax

The year-in-review format has a long history. News Publications pioneered it with their annual “Year In Review” issues, recapping the stories and events that shaped the previous twelve months. In tech, it was actually Facebook, not Spotify, which launched the first personalized Year in Review in 2014. Spotify followed two years later. Facebook eventually killed its feature after bad press over traumatic memories8. Spotify, on the other hand, went from strength to strength. After moving from email recaps to in-app experiences and rebranding as Wrapped in 20199, the feature became a cultural phenomenon.

Spotify now owns the word “Wrapped” and sets the standard in the consumer mind. The association is so strong that users describe every recap as “Wrapped” regardless of which company ships it or what the company names their feature. Every competitor pays a copycat tax unless they do something genuinely different.

2026, in my view, will likely be the year we will see truly spectacular AI-native wrapped experiences. Granola showed how AI can be used to generate witty summaries of users’ meeting patterns. In the next evolution, one could imagine your favorite artist calling you by name, or previewing a personalized song they’ve created just for you with your name worked into the lyrics. Memory from AI products could draw magical insights about you that other apps likely cannot. Your Year With ChatGPT is an early glimpse of this. These would be orthogonal to what Spotify does.

For the others, late November will continue to bring the same annual ritual. Somewhere, a tech CEO or VP of Product will watch Spotify Wrapped go viral yet again and send their product teams a request.

“We should do this.”

If you enjoyed this post, please consider sharing it on your socials or with someone who might also find it interesting. You can follow me on Twitter/X or LinkedIn for more frequent business and tech discussions.

Share

Appendix: Examples from 2025

Spotify (the OG): 2025 Wrapped

Strava: Year in Sport 2025 personalized (and aggregate).

Granola.ai: Crunched 2025

Google Photos: 2025 Recap

Google: Year in Search 2025

ChatGPT: Your Year With ChatGPT

Duolingo: 2025 Year in Review

Uber: YOUBER 2025

Cursor: Your Year in Code

NYT: Year in Games

Amazon Music: 2025 Delivered

Apple Music: Replay 2025 (separate review for Artists)

LinkedIn: Year in Review 2025

Discord: Checkpoint 2025

The year prior, Apple had opened the iPhone to third-party developers, and Amazon had launched two apps on the App Store: a shopping app that let customers browse and buy physical products from Amazon’s e-commerce store, and a Kindle app where users could read ebooks but could not buy them. Apple’s 30% commission on in-app digital purchases meant Amazon couldn’t afford to sell ebooks or digital music/video through its iOS apps (see footnote for more context on Amazon’s digital business economics1).

The traffic trend exposed a risk. Mobile would become the primary shopping channel, and if Apple decided to expand its take rate to physical goods (say 1-2% or so like credit card interchange fees), Amazon’s entire retail business would face the App Store tax, not just the digital business. In response2, Amazon started a new personal device project in 2010. From Fast Company:

The project, code-named “Tyto” for a genus of owl, got rolling in 2010. Customers often come to Amazon via iPhones or Android devices. Not controlling the hardware can create problems. For instance, you can’t buy e-books through the Kindle app on your iPhone because Apple takes 30% of app-driven sales—a cut that would hurt Amazon’s already razor-thin margin.

This device (Fire Phone) flopped, but the strategic concern that motivated it was valid. Mobile Operating Systems (OS) had inserted themselves as the new layer between users and applications, dictated what apps could and could not do3, and imposed significant transaction fees.

Today, AI is emerging as another new layer.

The AI Agent Layer

Players from both up and down the stack want to own this new layer.

Subscribe now

Application AI Agent. ChatGPT, Google Gemini, Perplexity etc. are building browsing and shopping AI agents into their apps that sit between users and other apps. OpenAI announced its Instant Checkout feature in September 2025, initially with Etsy and Shopify, followed quickly by Walmart and Target4. Amazon has notably stayed out of this partnership. But when Perplexity launched its AI browser Comet that lets users browse shopping websites using Perplexity’s AI agent, Amazon blocked Comet’s access and sent a cease-and-desist letter. I am skeptical of current AI chatbot shopping experiences5, but the public dispute between the two shows that neither wants to cede control of the user interaction.

Operating System AI Agent. Apple Intelligence and Google’s Gemini Android integration represent attempts to make AI a native OS capability. For example, Apple’s App Intents framework is trying to create structured ways for the OS AI to interact with applications.

The App Intents framework provides functionality to deeply integrate your app’s actions and content with system experiences. With Apple Intelligence, Siri will suggest your app’s actions to help people discover your app’s features and gains the ability to take actions in and across apps.

The vision is that when you ask Siri to “order dinner,”6 the OS AI compares across multiple apps (such as DoorDash or Uber Eats), shows you the options (including which ones have discounts), handles the back-and-forth, and completes the transaction. The user may never open some apps directly7. Eventually, the OS AI layer, in addition to coordination and orchestration, might even generate a custom UI across different apps if required.

Applications from foundation models have the technological advantages.

1. Their models are more capable.

2. They have built dedicated AI infrastructure, from physical resources like compute and power access to operational systems that serve models with low latency.

3. They have accumulated deep AI expertise and routinely publish world-class research, creating a flywheel that attracts other top AI talent.

Apple Intelligence, by contrast, has struggled. John Gruber, an influential writer and podcaster, captured the sentiment in his critical piece “Something Is Rotten in the State of Cupertino.” Google has competitive models with Gemini and can even outline a compelling vision for the OS AI layer, but organizational dysfunction has prevented Android from shipping a cohesive one. This dysfunction reminds me of that old joke about Google promising to make messaging simple again with yet another chat application (Google Talk, Google Hangouts, Android Messages, Google+ Messenger, Huddle, Google Allo, Duo, Google Voice, Google Chat, Google Meet; did I miss any?).

Structurally, though, Operating Systems are better positioned8.

1. They have system-level access to all apps on the device, so they can operate across multiple apps to finish a task, something a ChatGPT cannot do.

2. They have access to personal data across messages, photos, calendar, and location that any individual application cannot match.

3. They have a distribution advantage since they can ship to current devices via software updates and can come pre-installed on future ones.

So Amazon may be able to block Perplexity from intermediating users, but will they block Siri when Apple does the same?

I suppose, in a way, this is the “Everything App/Super App” that Chinese companies pioneered finally reaching Western markets, realized through the operating system rather than a single application.

The Chinese Dark Horse

Speaking of China, a new entrant from Beijing suggests that the pursuit of OS AI will not be an American-only affair.

Earlier this week, ByteDance unveiled a demo of their Doubao Phone Assistant, a Graphical User Interface (GUI) based OS AI that takes a different approach. Instead of relying on system-level hooks, it uses multimodal understanding of screen content to enable cross-app device control via simulated tapping, swiping, and typing. This direct interpretation means it can theoretically operate inside any app, including those the device has never seen before and with or without explicit developer cooperation. Apple Intelligence lacks this capability because it, as of now, requires apps to implement in advance the App Intents framework mentioned earlier. The demo showed price comparison across multiple retailers, car control integration with Tesla, and other agentic tasks.

From a market perspective, the parallels to the automotive industry are striking. Legacy US and European automakers once dismissed Chinese electric vehicles (EV) as cheap knockoffs9. But now Chinese EVs are capturing market share across Europe by competing both on price and quality/performance/features10. Tariffs and regulatory barriers have kept Chinese EVs out of the US market for now11.

Similarly, Chinese personal device manufacturers are currently dismissed as copycats12. But with their novel and ambitious OS AI efforts, proven open-source AI models such as DeepSeek, and lower cost structure, they could soon compete both on capability and cost. If that happens, the US will likely have to impose restrictions on these Chinese AI devices just as it has on Chinese cars and telecom equipment. But can the US influence the rest of the world to adopt American OS AI if ByteDance/Doubao or something like it proves superior at lower costs? Your guess is as good as mine.

Application Layer’s Strategic Choice

Every product that lives on the application layer faces a similar question to the one Amazon’s leadership faced fifteen years ago as they see a similar chart in their business reviews.

The Verge, a tech news website, notes that CEOs of companies in the application layer such as Uber, DoorDash, Airbnb, and Lyft all believe their accumulated advantages (supply networks, operational know-how, brand loyalty etc.) will protect them from this new AI intermediary (quotes slightly edited for better readability/context):

Brian Chesky (Airbnb CEO): There’s this AI maximalist view that there’s going to be like one or two AI models and one or two applications that rule them all and you use this one app and this one model for everything in the world. If you take that to its logical conclusion, you also start to go to this place where almost one company rules everything, and I think there’s numerous problems with the AI maximalist view that it’s one company to rule them all. I don’t think companies just want to be data layers, and so these platforms or these new interfaces are only as good as the companies that participate, and the companies will only participate if they can have a relationship with their own customer.

Dara Khosrowshahi (Uber CEO): As it relates to AI and these agents, we want to work with these players. We come from a place of strength because of the unique inventory and the fragmentation of these markets that we’re organizing. People spend so much time trying to figure out what the economics might be when the first thing is to try it out. Figure out the experience first and then the economics. Once you optimize the experience, we can measure. Are you an incremental consumer for Uber or are you totally cannibalistic? If it’s cannibalistic, then I’m going to charge a lot of money. If it is incremental, then I would pay some take rate. Is it a 5, 10, or 20 percent take rate? It depends on the incrementality.

Ania Smith (Taskrabbit CEO): Siri, in this case, wants to be able to provide these types of services, and they can’t really do it without Taskrabbit because only Taskrabbit actually has a network of thousands of Taskers. We cultivated that network. We know who they are and we understand their skills. They’ve all been background checked. Siri, Apple, or whoever is not going to do that.

Amazon’s lawsuit against Perplexity suggests they don’t share this view. Fifteen years ago, facing similar concerns, Amazon tried to build its own device. That device failed, but Amazon’s e-commerce business continued to thrive by accepting its position on the application layer and optimizing within those constraints. They refused to sell digital content through mobile apps and continued to invest in the customer experiences that App Store didn’t tax.

Just as Amazon’s leadership once saw personal devices as a way to escape the App Store tax, some AI companies (notably OpenAI) believe they too must make a personal computing device to own the user relationship. Besides the obvious belief that AI-native companies like themselves can design AI personal devices better than today’s device makers can, they don’t want to live at the mercy of Operating Systems. They don’t want to ask Cupertino for permission13. They want to own their own destiny.

Disclosure: I worked at Amazon from 2014-2024, but was not involved in personal device discussions.

If you enjoyed this post, please consider sharing it on your socials or with someone who might also find it interesting. You can follow me on Twitter/X or LinkedIn for more frequent business and tech discussions.

Share

Two lines in Cursor’s press release stood out to me. The first:

We’ve also crossed $1B in annualized revenue, counting millions of developers.

This disclosure means Anysphere Inc. (Cursor’s parent company) is the fastest company in history to reach $1B in annual recurring revenue (ARR). Yes, faster than OpenAI, and faster than Anthropic4.

Engineers are trying every new AI coding tool. As a result, the AI-coding tool market is growing exponentially (+5x in just over a year)5. But it’s still early. As I wrote in Why Some AI Wrappers Build Billion-dollar Businesses, companies spend several hundred billion dollars a year on software engineering, and AI has the potential to unlock productivity gains across that entire spend.

Software developers represent roughly 30% of the workforce at the world’s five largest market cap companies, all of which are technology firms as of October 2025. Development tools that boost productivity by even modest percentages unlock billions in value.

In my view, this nascent market is splitting based on three types of users.

On one end is Handcrafted Coding. These are engineers who actively decline to use LLMs, either because of skepticism about quality or insistence on full control of every code. They argue that accepting AI suggestions creates technical debt you cannot see until it breaks in production. This segment continues to decline as the quality of AI coding models improves.

The opposite end is Vibe Coding. These are typically non-engineers, who use AI to build concepts and prototypes. They prompt the model hoping for an end-to-end solution, accept the output with minimal review, and trust that it works. The user describes what they want and lets the model figure out the implementation details of how to build it.

In the middle sits Architect + AI Coding. The engineer uses the AI/LLM as a pair programmer exploring system designs, analyzing data models, and reviewing API details. When the work is something entirely new or something that needs careful handling, the human programmer still codes those pieces by hand. But for boilerplate code, package installations, generic User Interface (UI) components, and any kind of code that is typically found on the internet, they assign it to the model6. The engineer stays in command of what is important to them and delegates what is not.

The Market Split

Based on the user types, I think, the AI coding market splits into two.

1. Hands-off: Non-engineers (product managers, designers, marketers, other internal employees) use these tools to vibe code early product concepts. They look to AI as the lead engineer to spin-up concepts/prototypes of apps, websites, and tools by simply prompting the AI to make something for them. Lovable, Vercel, Bolt, Figma Make, and Replit fit here7. Code from these users, as of now, are not typically pushed to prod.

2. Hands-on: Professional software engineers use these tools in their existing workflow to ship production code. They use AI as an assistant to write boilerplate code, refactor existing services, wire new features or UI screens, and triage bugs in codebases. Cursor, Claude Code, OpenAI Codex, Github Copilot, Cline, AWS Kiro play here. These products live where the work is done, and integrate into the engineer’s workflow. This is, at least as of now, the bigger market segment.

To see an evaluation of all the major AI coding tools currently in the market, checkout this breakdown by Peter Yang, who runs the newsletter Behind The Craft.

That brings me to the second thing in Cursor’s press release that stood out to me:

Our in-house models now generate more code than almost any other LLMs in the world.

While I am not convinced about that claim8, what I am convinced about is that Cursor is still growing despite its previous reliance on foundation models. From Why Some AI Wrappers Build Billion-dollar Businesses again:

But Cursor and other such tools depend almost entirely on accessing Anthropic, OpenAI and Gemini models, until open-source open-weight and in-house models match or exceed frontier models in quality. Developer forums are filled with complaints about rate limits from paying subscribers. In my own projects, I exhausted my Claude credits in Cursor mid-project and despite preferring Cursor’s user interface and design, I migrated to Claude Code (and pay ten times more to avoid rate limits). The interface may be better, but model access proved decisive.

Cursor’s new in-house model Composer-2, which just launched last month, is a good example of how this model versus application competition is evolving. Cursor claims (without any external benchmarks, I must say) that Composer-2 is almost as good as frontier models but 4x faster. It’s still early to say how true that is. Open-source models have not yet come close to the top spots in SWE-bench verified or in private evals9.

To me, model quality is the most decisive factor in these AI coding wars. And in my view, that’s why Claude Code has already overtaken Cursor, and OpenAI’s Codex is close behind, despite both having launched a year or so later.

Even though the newcomers Cursor, Claude Code, and OpenAI Codex are the talk of the (developer) town, incumbents such as Microsoft with Github Copilot, AWS with Kiro, and Google with Antigravity, can utilize their existing customer relationships, bundle their offerings with their existing suites, and/or provide their option as the default in their tech stack to compete. As an example, Cursor charges $20–$40 monthly per user for productive usage, while Google Antigravity launched free with generous limits for individual users. Github Copilot still leads this market, proving once again that enterprise bundling and distribution has structural advantages. This is the classic Microsoft Teams vs. Slack Dynamic10.

One way for startups to compete is by winning individual users who may use a coding tool with or without formal approval, and then be the tool’s evangelist inside the organization. That organic interest and adoption eventually forces IT and security teams to officially review the tool and then eventually sanction its usage.

Yet, even as these newer tools capture developer mindshare, the underlying developer tools market is changing. Both the IDEs developers choose and the resources they we consult have changed dramatically. StackOverflow, once the default for programmers stuck on a programming issue, has seen its traffic and number of questions decline dramatically since ChatGPT’s launch, suggesting that AI is already replacing some traditional developer resources.

Just as compilers freed programmers from writing assembly code, AI tools are freeing software engineers from the grunt work of writing boilerplate and routine code, and letting them us focus on higher-order thinking. Eventually, one day, AI may get so good that it will generate applications on demand and create entire software ecosystems autonomously. Both hands-off and hands-on AI coding tools, as well as incumbents and newcomers, see themselves as the path to that fully autonomous software generation, even if they are taking different approaches. The ones who get there will be those who deliver the best model quality that ships code reliably, go deep enough to ship features that foundation models can’t care enough to replicate, and become sticky enough that users will not leave even when they can11.

If you enjoyed this post, please consider liking it on Twitter/X or LinkedIn, or sharing it with someone who might also find it interesting

Share

Your computer modem screeches as it tries to connect to something called the internet. Maybe it works. Maybe you try again.

For the first time in history, you can exchange letters with someone across the world in seconds. Only 2000-something websites exist1, so you could theoretically visit them all over a weekend. Most websites are just text on gray backgrounds with the occasional pixelated image2. Loading times are brutal. A single image could take a minute, a 1-minute video could take hours. Most people do not trust putting their credit cards online. The advice everyone gives: don’t trust strangers on the internet.

People split into two camps very soon.

Optimists predict grand transformations. Some believe digital commerce will overtake physical retail within years. Others insist we’ll wander around in virtual reality worlds.

“I expect that within the next five years more than one in ten people will wear head-mounted computer displays while traveling in buses, trains, and planes.” - Nicholas Negroponte, MIT Professor, 1993

Pessimists call the internet a fad and a bubble.

If you told the average person in 1995 that within 25 years, we’d consume news from strangers on social media over newspapers, watch shows on-demand in place of cable TV, find romantic partners through apps more than through friends, and flip “don’t trust strangers on the internet” so completely that we’d let internet strangers pick us up in their personal vehicles and sleep in their spare bedrooms, most people would find that hard to believe.

We’re in 1995 again. This time with Artificial Intelligence.

And both sides of today’s debate are making similar mistakes.

One side warns that AI will eliminate entire professions and cause mass unemployment within a couple of years. The other claims that AI will create more jobs than it destroys. One camp dismisses AI as overhyped vaporware destined for a bubble burst, while the other predicts it will automate every knowledge task and reshape civilization within the decade.

Both are part right and part wrong.

The Employment Paradox: Why Automation’s Impact Depends On The Industry

Geoffrey Hinton, who some call the Father of AI, warned in 2016 that AI would trigger mass unemployment. “People should stop training radiologists now,” he declared, certain that AI would replace them within years.

Yet as Deena Mousa, a researcher, shows in “The Algorithm Will See You Now,” AI hasn’t replaced radiologists, despite predictions. It is thriving.

In 2025, American diagnostic radiology residency programs offered a record 1,208 positions across all radiology specialties, a four percent increase from 2024, and the field’s vacancy rates are at all-time highs. In 2025, radiology was the second-highest-paid medical specialty in the country, with an average income of $520,000, over 48 percent higher than the average salary in 2015.

Mousa identifies a few factors for why the prediction failed - real-world complexity, the job involves more than image recognition, and regulatory/insurance hurdles. Most critical she points is Jevons Paradox, which is the economic principle that a technological improvement in resource efficiency leads to an increase in the total consumption of that resource, rather than a decrease. Her argument is that as AI makes radiologists more productive, better diagnostics and faster turnaround at lower costs mean more people get scans. So employment doesn’t decrease. It increases.

This is also the Tech world’s consensus. Microsoft CEO Satya Nadella agrees, as does Box CEO Aaron Levie, who suggests:

“The least understood yet most important concept in the world is Jevons Paradox. When we make a technology more efficient, demand goes well beyond the original level. AI is the perfect example of this—almost anything that AI is applied to will see more demand, not less.”

They’re only half right.

First, as Andrej Karpathy, the computer scientist who coined the term vibe coding, points out, radiology is not the right job to look for initial job displacements.

“Radiology is too multi-faceted, too high risk, too regulated. When looking for jobs that will change a lot due to AI on shorter time scales, I’d look in other places - jobs that look like repetition of one rote task, each task being relatively independent, closed (not requiring too much context), short (in time), forgiving (the cost of mistake is low), and of course automatable giving current (and digital) capability. Even then, I’d expect to see AI adopted as a tool at first, where jobs change and refactor (e.g. more monitoring or supervising than manual doing, etc).”

Second, the tech consensus that we will see increased employment actually depends on the industry. Specifically, how much unfulfilled demand can be unlocked in that industry, and whether this unfulfilled demand growth outpaces continued automation and productivity improvement.

To understand this better, look at what actually happened in three industries over a 200-year period from 1800 to 2000. In the paper Automation and jobs: when technology boosts employment, James Bessen, an economist, shows the employment, productivity, and demand data for textile, iron & steel, and motor vehicle industries.

After automation, both textile and iron/steel workers saw employment increase for nearly a century before experiencing a steep decline. Vehicle manufacturing, by contrast, holds steady and hasn’t seen the same steep decline yet.

To answer why those two industries saw sharp declines but motor vehicle manufacturing did not, first look at the productivity of workers in all three industries:

Then look at the demand across those three industries:

What the graphs show is a consistent pattern (note: the productivity and demand graphs are logarithmic, meaning productivity and demand grew exponentially). Early on, a service or product is expensive because many workers are needed to produce it. Most people can’t afford it or use them sparingly. For example, in the early 1800s, most people could only afford a pair of pants or shirt. Then automation makes workers dramatically more productive. A textile worker in 1900 could produce fifty times more than one in 1800. This productivity explosion crashes prices, which creates massive new demand. Suddenly everyone can afford multiple outfits instead of just one or two. Employment and productivity both surge (note: employment growth masks internal segment displacement and wage changes. See footnote3)

Once demand saturates, employment doesn’t further increase but holds steady at peak demand. But as automation continues and workers keep getting more productive, employment starts to decline. In textiles, mechanization enabled massive output growth but ultimately displaced workers once consumption plateaued while automation and productivity continued climbing. We probably don’t need infinite clothing. Similarly, a patient will likely never need a million radiology reports, no matter how cheap they become and so radiologists will eventually hit a ceiling. We don’t need infinite food, clothing, tax returns, and so on.

Motor vehicles, in Bessen’s graphs, tell a different story because demand remains far from saturated. Most people globally still don’t own cars. Automation hasn’t completely conquered manufacturing either (Tesla’s retreat from full manufacturing automation proves the current technical limits). When both demand and automation potential remain high, employment can sustain or even grow despite productivity gains.

Software presents an even more interesting question. How many apps do you need? What about software that generates applications on demand, that creates entire software ecosystems autonomously? Until now, handcrafted software was the constraint. Expensive software engineers and their our labor costs limited what companies could afford to build. Automation changes this equation by making those engineers far more productive. Both consumer and enterprise software markets suggest significant unmet demand because businesses have consistently left projects unbuilt4. They couldn’t justify the development costs or had to allocate limited resources to their top priority projects. I saw this firsthand at Amazon. Thousands of ideas went unfunded not because they lacked business value, but because of the lack of engineering resources to build them. If AI can produce software at a fraction of the cost, that unleashes enormous latent demand. The key question then is if and when that demand will saturate.

So to generalize, for each industry, employment hinges on a race between two forces:

1. The magnitude and growth of unmet market demand unlocked by lower prices, and

2. Whether that demand growth outpaces productivity improvements from automation.

Different industries will experience different outcomes depending on who’s winning that demand and productivity race.

The Bubble: Irrational Exuberance Builds the Future

The second debate centers on whether this AI boom is a bubble waiting to burst.

The dotcom boom of the 1990s saw a wave of companies adding “.com” to their name to ride the mania and watch their valuations soar. Companies poured billions into fiber optics and undersea cables - expensive projects only possible because people believed the hype5. All of this eventually burst in spectacular fashion in the dotcom crash in 2000-2001. Infrastructure companies like Cisco briefly became the most valuable in the world only to come tumbling down6. Pets.com served as the poster child of this exuberance raising $82.5 million in its IPO, spending millions on a Super Bowl ad only to collapse nine months later7.

But the dotcom bubble also got several things right. More importantly, it eventually bought us the physical infrastructure that made YouTube, Netflix, and Facebook possible. Sure, companies like Worldcom, NorthPoint, and Global Crossing making these investments went bankrupt, but they also laid the foundation for the future. Although the crash proved the skeptics right in the short term, it proved the optimists were directionally correct in the long term.

Today’s AI boom shows similar exuberance. Consider the AI startup founded by former OpenAI executive Mira Murati, which raised $2 billion at a $10 billion valuation, the largest seed round in history8. This despite having no product and declining to reveal what it’s building or how it will generate returns. Several AI wrappers have raised millions in seed funding with little to no moat.

Yet some investments will outlast the hype and will likely help future AI companies even if this is a bubble. For example, the annual capital expenditures of Hyperscalers9 that have more than doubled since ChatGPT’s release - Microsoft, Google, Meta, and Amazon are collectively spending almost half a trillion dollars on data centers, chips, and compute infrastructure. Regardless of which specific companies survive, this infrastructure being built now will create the foundation for our AI future - from inference capacity to the power generation needed to support it.

The infrastructure investments may have long-term value, but are we already in bubble territory? Azeem Azhar, a tech analyst and investor, provides an excellent practical framework to answer the AI bubble question. He benchmarks today’s AI boom using five gauges: economic strain (investment as a share of GDP), industry strain (capex to revenue ratios), revenue growth trajectories (doubling time), valuation heat (price-to-earnings multiples), and funding quality (the resilience of capital sources). His analysis shows that AI remains in a demand-led boom rather than a bubble, but if two of the five gauges head into red, we will be in bubble territory.

The demand is real. After all OpenAI is one of the fastest-growing companies in history10. But that alone doesn’t prevent bubbles. OpenAI will likely be fine given its product-market fit, but many other AI companies face the same unit economics questions that plagued dotcom companies in the 1990s. Pets.com had millions of users too (a then large portion of internet users), but as the tech axiom goes, you can acquire infinite customers and generate infinite revenue if you sell dollars for 85 cents11. So despite the demand, the pattern may be similar to the 1990s. Expect overbuilding. Expect some spectacular failures. But also expect the infrastructure to outlast the hype cycle and enable things we can’t yet imagine.

The Predictably Unpredictable Future

So where does this leave us?

We’re early in the AI revolution. We’re at that metaphorical screeching modem phase of the internet era. Just as infrastructure companies poured billions into fiber optics, hyperscalers now pour billions into data centers. Startups add “.ai” to their names like how companies once added “.com” as they seek higher valuations. The hype will cycle through both euphoria and despair. Some predictions will look laughably wrong. Some that seem crazy will prove conservative.

Different industries will experience different outcomes. Unlike what the Jevons optimists suggest, demand for many things plateaus once human needs are met. Employment outcomes in any industry depend on the magnitude and growth of unmet market demand and whether that demand growth outpaces productivity improvements from automation.

Cost reduction will unlock market segments. Aswath Damodaran, a finance professor, (in)famously undervalued Uber assuming it would only capture a portion of the existing taxi market12. He missed that making rides dramatically cheaper would expand the market itself as people took Ubers to destinations they’d never have paid taxi prices to reach. AI will similarly enable products and services currently too expensive to build with human intelligence. A restaurant owner might use AI to create custom supply chain software that with human developers, say at $100,000, would never have been built. A non-profit might deploy AI to contest a legal battle that was previously unaffordable.

We can predict change, but we can’t predict the details. No one in 1995 predicted we’d date strangers from the internet, ride in their ubers, or sleep in their airbnbs. Or that a job called influencers would become the most sought-after career among young people. Human creativity generates outcomes we can’t forecast with our current mental models. Expect new domains and industries to emerge. AI has already helped us decode more animal communication in the last five years than in the last fifty. Can we predict what jobs a technology that allows us to have full-blown conversations with them will unlock? A job that doesn’t exist today will likely be the most sought-after job in 2050. We can’t name it because it hasn’t been invented yet.

Job categories will transform. Even as the internet made some jobs obsolete, it also transformed others and created new categories. Expect the same with AI. Karpathy ends with a question:

“About 6 months ago, I was also asked to vote if we will have less or more software engineers in 5 years. Exercise left for the reader.”

To answer this question, go back to 1995 and ask the same question but with journalists. You might have predicted more journalists because the internet would create more demand by enabling you to reach the whole world. You’d be right for 10 or so years as employment in journalism grew until the early 2000s. But 30 years later, the number of newspapers and the number of journalists both have declined, even though more “journalism” happens than ever. Just not by people we call journalists. Bloggers, influencers, YouTubers, and newsletter writers do the work that traditional journalists used to do13.

The same pattern will play out with software engineers. We’ll see more people doing software engineering work and in a decade or so, what “software engineer” means will have transformed. Consider the restaurant owner from earlier who uses AI to create custom inventory software that is useful only for them. They won’t call themselves a software engineer.

So just like in 1995, if the AI optimists today predict that within 25 years, we’d prefer news from AI over social media influencers, watch AI-generated characters in place of human actors, find romantic partners through AI matchmakers more than through dating apps (or perhaps use AI romantic partners itself), and flip “don’t trust AI” so completely that we’d rely on AI for life-or-death decisions and trust it to raise our children, most people would find that hard to believe. Even with all the intelligence, both natural and artificial, no one can predict with certainty what our AI future will look like. Not the tech CEOs, not the AI researchers, and certainly not some random guy pontificating on the internet. But whether we get the details right or not, our AI future is loading.

If you enjoyed this post, please consider sharing it with someone who might also find it interesting. Follow me on Twitter/X or LinkedIn for more insights like this.

Share

3

Historical data from the Industrial Revolution shows that even as aggregate textile employment grew, workers shifted between job types within the industry as some roles became redundant. And correspondingly, some jobs saw wages collapse while others saw increases. For example, domestic hand-loom weavers were displaced by power looms and saw their wages collapse, while self-acting mule spinners (a newly created role) and factory workers saw stable employment and steady compensation growth. Additionally, Britain’s deflationary period (1815-1850) saw food prices fall by half, meaning real purchasing power often rose even when nominal wages declined. Despite all this, the psychological reality was harsh. Even with falling prices, watching your paycheck shrink while neighbors lost jobs and debts grew harder to pay (lower wages but fixed obligations) created real instability regardless of aggregate statistics improvements. Also see Acemoglu & Johnson, 2024.

The put‑down feels familiar to people developing something new using Artificial Intelligence.

The push-back feels just as familiar.

“Everything is a wrapper. OpenAI is a wrapper around Nvidia and Azure. Netflix is a wrapper around AWS. Salesforce is an Oracle database wrapper valued at $320 billion,” says Perplexity CEO Aravind Srinivas1.

For those not familiar with the term “AI Wrapper,” here’s a good definition2.

It is a dismissive term that refers to a lightweight application or service that uses existing AI models or APIs to provide specific functionality, typically with minimal effort or complexity involved in its creation. A popular example of an AI wrapper are apps that enable users to “chat” with a PDF. This type of AI application allows users to upload a PDF document, such as a research paper, and interact with an AI model to quickly analyze and obtain answers about the specific content. In the early days of ChatGPT, uploading documents as part of the prompt or creating a custom GPT was not possible, so these apps became very popular, very fast.

In my view, this AI wrapper debate misses a larger point. Wrappers are not all the same. Some enjoy a brief run and last only until big platforms bundle them into their suites. But products that can live where work is done, builds proprietary data and learns from usage, and/or withstand incumbent distribution advantages can compete. The wrapper label is a distraction from what I think actually matters: (1) Is it a feature or a product, and (2) How big is the market segment.

Feature Or Product

Let’s first look at that earlier example of a wrapper that lets you chat with a PDF. Such a tool solves one narrow problem of answering questions about a document. It does not create new documents or edit existing ones. It typically does not capture any unique data, or learn from user behavior. So to me, it is a capability rather than an end-to-end solution. A means to an end, if I may. As a result, this kind of feature belongs inside a document viewer or editor, or in the flagship applications of model providers. So when the foundation models themselves (OpenAI/ChatGPT, Anthropic/Claude, Google/Gemini) bundle this feature natively, the standalone tool becomes redundant. This is classic feature behavior - easy to copy, no end-to-end job, no moat or long-term defensibility.

One caveat though; even those that are features can be an interesting indie businesses that make money until the platforms build it into their apps3.

• PDF.ai $500K MRR, PhotoAI $77K MRR, Chatbase $70K MRR, InteriorAI $53K MRR4.

• Jenni AI went from $2,000 to over $333,000 MRR in just 18 months5.

Too Big To Ignore

Some wrappers are genuine products but live in market segments so large that model builders and big tech platforms cannot ignore them. Two vectors of competition come into play: (1) model access, and (2) distribution.

Model Access

Coding assistants illustrate both. Tools such as Cursor have turned a wrapper into an AI integrated development environment (IDE) that reads the repo, edits files, generates code, reverts changes, runs coding agents, and reimagines the developer experience for the AI era. The market justifies the attention. Software developers represent roughly 30% of the workforce at the world’s five largest market cap companies, all of which are technology firms as of October 20256. Development tools that boost productivity by even modest percentages unlock billions in value. That makes this segment a prime target for both model builders and incumbents that already own distribution channels.

But Cursor and other such tools depend almost entirely on accessing Anthropic, OpenAI and Gemini models, until open-source open-weight and in-house models match or exceed frontier models in quality. Developer forums are filled with complaints about rate limits from paying subscribers. In my own projects, I exhausted my Claude credits in Cursor mid-project and despite preferring Cursor’s user interface and design, I migrated to Claude Code (and pay ten times more to avoid rate limits). The interface may be better, but model access proved decisive.

The model access dependency carries strategic implications beyond rate limits. OpenAI CEO Sam Altman argues the right strategy assumes continuous model improvement7:

“There are two strategies to build on AI right now. There’s one strategy which is assume the model is not gonna get better. There’s another strategy which is assume that the models are gonna keep getting better at the same pace. It would seem to me that 95% of the world should be betting on the latter category but a lot of the startups have been built in the former category. When we just do our fundamental job, because we have a mission, we’re going to steamroll you.”

The foundation model competition extends to every category that OpenAI Applications CEO Fidji Simo flagged as strategic (Knowledge/Tutoring, Health, Creative Expression, and Shopping) as well as other large market segments such as Writing Assistants, Legal Assistants, etc.

Distribution

Distribution poses the second threat. Even where model builders stay out, startups face a different competition question - can they build a user base faster than incumbents with existing products and distribution can add AI features? This is the classic Microsoft Teams vs. Slack Dynamic8. The challenge is in establishing a loyal customer base before Microsoft embeds Copilot in Excel/PowerPoint, or Google weaves Gemini into Workspace, or Adobe integrates AI across its creative suite. A standalone AI wrapper for spreadsheets or presentations must overcome not just feature parity but also bundling/distribution advantages and switching costs.

This distribution competition from incumbents also holds in other large markets such as healthcare and law. In these markets, regulatory friction and control of systems of record9 favor established players such as Epic Systems in healthcare. For e.g. A clinical note generator that cannot write to the Electronic Health Record (EHR) will likely come up against Epic’s distribution advantages sooner or later.

Three caveats here: (1) First, speed to market can create exit options even without long-term defensibility; tools like Cursor may lack control over its core dependency (model access), but rapid growth make them attractive targets for model builders seeking instant market presence. (2) Second, superior execution occasionally beats structural advantage; Midjourney’s product quality convinced Meta to use it despite Meta’s substantially larger budget and distribution power. (3) Third, foundation models may avoid certain markets despite their size; regulatory burden in healthcare and legal, or reputational damage from AI companions or pornographic adult content may provide opportunities for operators willing to face extreme regulatory scrutiny or controversy.

The opportunity remains large10, but competition (and/or acquisition) can come knocking.

• Cursor went from zero to $100 million in recurring revenue in 18 months, and became the subject of recurring OpenAI acquisition rumors.

• Windsurf, another coding assistant, received a $2.4B acquisition licensing deal from Google.

• Gamma reached $50 million in revenue in about a year.

• Lovable hit $50 million in revenue in just six months.

• Galileo AI acquired by Google for an undisclosed amount.

The Entrepreneur’s Opportunity

Not every market gap attracts model builders or big tech. A long tail of jobs exists that are too small for venture scale but large enough to support multimillion-dollar businesses. These niches suit frugal founders with disciplined scope and lean operations.

Consider those Astrology or Manifestation or Dream Interpreter AI apps. A dream interpreter that lets users record dreams each morning, generates AI videos based on them, maintains some kind of dream journal, and surfaces patterns over time solves a complete job. Yes, users could describe dreams to ChatGPT and it even stores history/memory, but a dedicated app can structure the dream capture with specific fields (recurring people, places, things, themes etc.) and integrate with sleep tracking data in ways a general chatbot likely cannot. Such a niche is small enough to avoid model attention but seems large enough to sustain a profitable indie business.

Models Versus Incumbents

While the previous categories frame opportunities for new ventures, incumbents face their own strategic choices in the wrapper debate when model builders arrive. Those that navigate model builder competition, in my view, will share two characteristics.

1. First, they will own the outcome even when they don’t own the model. Applications already embedded in user workflows (Gmail/Calendar, Sheets, EHR/EMR, Figma) require no new habit formation, and building these platforms from scratch is much harder than adding AI capability to existing ones. When these applications ship actions directly into a proprietary system of record (controlling the calendar event, filing the claim, creating the purchase order, and so on), “done” happens inside the incumbent’s environment. AI then will act as another input to an existing workflow rather than a replacement for it.

2. Second, successful incumbents will build proprietary data from customer usage. Corrections, edge cases, approvals, and human feedback generate training data that refines the product over time and that a frontier model will not have access to. Cursor, though not an incumbent and despite its dependence on external models, plans to compete by capturing developer behavior patterns as CEO Michael Truell notes in his Stratechery interview:

   Ben: Is that a real sustainable advantage for you going forward, where you can really dominate the space because you have the usage data, it’s not just calling out to an LLM, that got you started, but now you’re training your own models based on people using Cursor. You started out by having the whole context of the code, which is the first thing you need to do to even accomplish this, but now you have your own data to train on.

   Michael: Yeah, I think it’s a big advantage, and I think these dynamics of high ceiling, you can kind of pick between products and then this kind of third dynamic of distribution then gets your data, which then helps you make the product better. I think all three of those things were shared by search at the end of the 90s and early 2000s, and so in many ways I think that actually, the competitive dynamics of our market mirror search more than normal enterprise software markets.

Unwrapping the Wrappers

Both critics and defenders of AI wrappers have a point, and both miss something. The critics are right that some wrappers lack defensibility and will disappear when platforms absorb their features. The defenders are right that every successful software company wraps something.

But I think the insight lies between these positions. Even if a new application starts as a wrapper, it can compete if it lives where work is done, writes to proprietary systems of record, builds proprietary data and learns from usage, and/or captures distribution before incumbents bundle the feature. More importantly, wrappers that continue to swiftly ship features that solve users’ needs even as competition arrives are difficult to compete with. These are the same traits that separate lasting products from fleeting features.

If you enjoyed this post, please consider liking it on Twitter/X or LinkedIn, or sharing it with someone who might also find it interesting.

Share

Building on that, I classify platforms and marketplaces by their business model into four kinds. Their pricing power, i.e. what they can charge and what they must give up, depends on where they land in this classification2.

Paid Cross-side Platforms – These platforms create value primarily by enabling direct exchanges between its consumers and producers, and benefit from cross-side network effects, i.e. the volume and nature of merchants attracts users, and more users attract more merchants. Interaction is cross-directional, from sellers to customers and customers to sellers. Customers do not typically interact with other customers (and merchants do not necessarily interact with other merchants). Examples are Airbnb, eBay, Amazon Marketplace, Alibaba’s Tmall, Uber, Lyft etc. I call them paid platforms because they charge merchants a take rate when they sell to customers. Paid platforms are typically take rate-elastic (I define “Take Rate Elasticity” as the sensitivity of seller participation to changes in a platform’s take rate, with pass-through to end prices; see footnote3). If fees rise too high, sellers will look for cheaper alternatives. With no hardware or infrastructure lock-in, competitors can lure sellers away by offering lower fees. Managed platforms, however, can sustain higher take rates if customers or merchants see value in the oversight and service they provide.

Infra/Hardware Cross-side Platforms – Cross-side platforms or marketplaces built on top of an infrastructure or hardware layer are unique because these platforms do not typically attract users by the volume and nature of the merchants on their platforms; rather the hardware attracts users, and the users then attract merchants. In some cases, unique infrastructure (warehouses, delivery mechanisms, curated storefronts) pulls in merchants on its own. In either case, because demand is captive to a locked-in installed base and/or the platform offers infrastructure that sellers can’t get elsewhere, these platforms are more take rate-inelastic, and can charge higher take rates without losing their best suppliers. Examples include Apple Appstore, Xbox Gamestore, Google Playstore. Theoretically, even a physical marketplace in a unique location that allows merchants to sell their wares, and charges them on a revenue share model falls in this category. Fulfillment by Amazon fits here too. Its vast warehouses and logistics network are costly to copy, giving Amazon structural bargaining power to set storage/fulfillment fees.

Free Cross-side Platforms – These platforms charge neither consumers nor producers at the point of exchange. Exchanges between their consumers and producers are almost always free. Instead, the platforms monetize attention; more users bring more engagement, which advertisers then pay to access. They benefit from cross-side network effects. Google and Yelp are examples of this kind of platform. Pricing power comes from audience quality, auction dynamics, and measurable return on investment (ROI). Platforms with affluent, high-intent users sell higher-value attention, and if ads convert and targeting is superior, bids rise.

Same-side Platforms – These platforms create value by enabling direct exchange among its users, and are monetized either by charging the users directly (rare), or by charging advertisers to sell services to users (most often). Messaging apps, social networks, and communities are examples, and these platforms benefit from same-side effects (more users -> more users) and once there is a valuable installed base, one-way cross-side effects (more users -> more advertisers). But note that more advertisers do not directly mean more customers. In fact, more ads -> (arguably) poorer customer experience -> fewer customers. WhatsApp, Facebook, Instagram, WeChat, Snapchat are examples.

Some companies sit across multiple models or blend different characteristics but products typically pertain to one model. For example, Google Search is a Free Cross-side Platform for search, but Gmail is a Same-side Platform. YouTube is a Paid Cross-side Platform despite being free to users, as it takes a portion of advertising revenue before paying creators.

Platform Type Determines Take Rate Elasticity

Based on the above classification and given a platform business model, how much take rate can the marketplace bear? The most important factor is the actual margin of each supplier on the platform. If suppliers themselves only have 20% margin, they cannot bear 20% in platform fees. But beyond that, the platform’s structural position determines what it can extract without losing suppliers.

Paid Cross-side Platforms are more take rate-elastic because a competing platform with lower take rate can first attract sellers on fees, and then eventually win customers as sellers pass the savings on as lower prices. That’s why I think Airbnb must keep its take rate around 15%. Because attempting to increase it would invite competitors to undercut on fees, attract its hosts, and ultimately capture its customers with lower prices. Infra/Hardware Platforms get away with higher take rate because of their hardware differentiation. That’s why Apple’s Appstore can get away (and they have for 15+ years) with charging developers a high take rate (the famous 30% Apple Tax) because even if a competing platform like Android offered a lower take rate (say 5%), users are unlikely to switch as long as Apple’s hardware remains attractive. Here’s a thought exercise - if Apple increased its take rate to 50%, would the quality and quantity of apps materially suffer? Would customers switch platforms to Android? In my view, change is more likely to come from regulators or technological shifts (e.g. AI devices) than from competitors or customers.

In Free Platforms and Same-side Platforms, prices charged to advertisers don’t (directly) affect the prices on customers, so theoretically new platforms cannot undercut the price (charged to advertisers) of the dominant platform and capture more users. The way to capture marketshare is to offer differentiated service (Facebook vs. Myspace, TikTok vs. Facebook).

From a cost point of view, there are three costs at play in a platform business. Fixed operating costs that do not vary with each unit of output (i.e. expenses that are largely constant over a range and only step changes as the platform grows such as office rent, most employee salaries), variable costs that vary with each unit of output (i.e. costs that vary with the number of sellers and customers serviced such as payment processing)4, and an upfront capital expense incurred in developing the platform (primarily in physical assets and hardware costs)5.

In my view, a balanced paid cross-side platform pricing strategy recognizes four factors - the platform’s position on the pricing-power spectrum above, the underlying cost structure, end-customer price elasticity, and access to patient long-term investors.

In the early stages, a platform may choose their take rate based on access to such long-term investment capital. Those with access to capital can charge a take rate low enough to subsidize liquidity (attract sellers, and thus customers), and those without may charge high enough to account for capital expenses, and fixed and variable costs as a percentage of Gross Merchandise Value (GMV). This ensures sufficient cash flow to keep the business viable.

Eventually, as upfront investments are depreciated and fixed costs are spread over more customers and orders, these expenses fall as a share of GMV. At this point, with improved unit economics and a captive base, platforms in winner-takes-most markets might be tempted to raise the take rate to increase profitability. But I think they should resist this temptation and instead pursue the opposite strategy. One that sets the take rate just a touch above variable costs and fixed costs. Such a low but sustainable take rate attracts sellers and is prohibitively hard for new entrants to match/undercut. Importantly, it keeps prices low for end-customers. Sellers typically pass through fee increases to customers as higher prices, which, if the product is price elastic, suppresses transaction volume. If the profit loss from reduced sales exceeds the profit gain from the higher take rate, the rate increase destroys enterprise value.

Complementary Services, Concentration, and Contribution

Beyond the base take rate, platforms can raise their effective take rates by offering value-added services such as payments processing, quality control, insurance, and advertising. These services let suppliers who want more distribution or trust signals pay extra while keeping base prices accessible for others. Consider Amazon’s advertising business, which generated $56B in 2024. Amazon Marketplace’s base take rate sits at around 15-20%, but suppliers who pay for prominent search placement raise Amazon’s effective take rate to an estimated 40-50% if you include advertising and other value added services.

Tanay Jaipuria, a tech investor, explores a few more factors that influence take rates, two of which I find particularly compelling:

• Supplier (and buyer) concentration. Fragmented suppliers have little bargaining power and accept higher take rates; concentrated suppliers can negotiate lower rates or threaten to withhold supply. The same logic applies to buyer concentration, though most consumer marketplaces naturally have fragmented buyers.

  The best example to highlight this is by digging into Expedia, which sells both flights and hotels. On the flight side, there are typically only ~4 airlines that matter for a given geography, so the supply is very concentrated. Meanwhile, on the hotel side, the supply is much more fragmented. While they don’t breakdown the take rate by segment, most estimates suggest that on the hotel side, their take rate is in the 15-20% range, while on the airlines side it is in the 3-5% range. The concentration is a key driver in this. Think about if Expedia didn’t have United when you performed a search. If you wanted the best price, you could search on Expedia, and then separately search on United before making your purchase.

• Incremental sales justify higher take rates. Platforms that generate sales that suppliers wouldn’t have captured on their own can charge more for that incremental value. And the more a platform contributes to a transaction, the higher the take rate it can justify.

  One way to see this is to consider the difference between Shopify and Amazon. Shopify provides merchants the tools to set up a store and process transactions but isn’t necessarily bringing them sales and traffic. Meanwhile, Amazon has aggregated 100s of millions of buyers, and sellers lose the ability to reach them on a given search (and potentially make an incremental sale) if they aren’t on Amazon’s third-party marketplace. Shopify’s effective take rate is ~3% while Amazon’s is 10-15% depending on the category of product.

  In fact, some marketplaces go as far as charging different take rates on transactions depending on their level of contribution to driving the transaction. Take Udemy’s pricing structure, which highlights the importance of distribution:

  • If the course is sold through the instructor’s link → instructor keeps 97%

  • If the course is sold organically on Udemy → instructor keeps 50%

  • If the course is sold through Udemy’s partners’ → instructor keeps 25%

The chart above raises an obvious question, what’s going on with Shutterstock?

Unlike most marketplaces that recognize only their take rate as revenue, Shutterstock reports the full customer payment because it controls the transaction before transferring content. So, per its 2024 financial statements, GMV was approximately reported revenue of $935M. Sales & Marketing (S&M) was $223M (24% of GMV). What this means is that just to cover S&M, Shutterstock needs 24 percentage points of their take rate. Add in other operating costs and their operating profit margin sits at just 7%. In comparison, Airbnb generated $81.8B in GMV, $11.1B in revenue, and only spends $2.1B (2.5% of GMV) in S&M for an operating profit of $2.5B, or 22.5%. I think, unlike Airbnb with its direct traffic and brand strength, Shutterstock competes in fiercely competitive paid search markets for search terms such as “stock photo for”.

This discussion raises a related question about YouTube. How does the platform sustain a 45% take rate? Alphabet/Google does not break out YouTube’s financials separately (technically compliant with legal requirements but debatable given YouTube’s scale and distinct creator business model6). This means we do not know their true cost structure. Assuming YouTube operates more like Airbnb with its 2.6% S&M spend rather than Shutterstock with its 24% paid search dependency, the take rate likely sits well above the optimal rate.

The answer, in my opinion, is Game theory dynamics.

Potential competitors must calculate whether triggering a price war justifies the investment. In YouTube’s case, if a new video platform (say Netflix) launched by offering better rates for creators, YouTube would immediately match it. At equal rates, creators would not move to the new platform since they have already uploaded their entire video portfolios, and built reputations on YouTube. The new entrant would have invested time and capital building the platform for minimal market share gain. From a potential entrant’s perspective, this market is not worth entering. The result is monopolistic/oligopolistic behavior where incumbents can charge above-equilibrium rates7.

The game theory deterrent breaks when competitors exploit technological shifts (AI-generated videos), or pursue different models (shorter video formats such as TikTok). More likely, regulators will need to step in when they see platforms earning outsized profits. Even then, meaningful competition takes years to materialize. New entrants must raise capital, build platform technology, recruit sellers, acquire customers, and critically, help sellers build the reputation and reviews that make them credible. For platforms with geographic constraints such as Uber or Airbnb, competitors must light up city by city, or region by region.

In summary, four forces determine take rates in practice. Platform type (including complementary services, concentration, and contribution) define pricing power, cost structure and supplier margins set the minimum viable take rate, end-customer price elasticity determines how take rate changes affect transaction volume and total profit, and game theory and regulatory action governs how long take rates above equilibrium last. Pricing just a bit above variables costs + fixed costs buffer is the sweet spot where the platform balances seller acquisition, low prices for end-consumers, and long-term defensibility.

If you enjoyed this post, please consider liking the post on Twitter/X or LinkedIn, or sharing it with someone who might also find it interesting.

Share

“The act of writing is the act of discovering what you believe.” This quote captures the essence of why I am starting this. Through my writing, I seek to discover and validate (or invalidate) my beliefs, and share what I discover. An experiment in thinking out loud.

You can expect to find theories, frameworks, analysis, and opinion on business, tech, and the business of tech, primarily around four areas:

1. Tech Analysis - Making sense of Artificial Intelligence, Platforms, Marketplaces, and whatever’s changing our world, and specifically what it means for business.

2. Business Strategy - Why do some win while others fail. What trade-offs shape product, pricing, and market dynamics?

3. Product Lessons - Stories from past experiences building and scaling products. Lessons from leading tech teams. Recent experiments, successes, and failures. What works, what doesn't?

4. Personal (w)reflection - Occasional essays about my life and other interests.

About Me

I am an electrical engineer by training, with a masters in business. I most recently spent 10 years at Amazon building warehouse robots, and launching new categories and products.

You can also find me on Twitter/X or LinkedIn discussing tech and business trends as they happen.

I look forward to making this worth your time.

Subscribe now