"The worst thing that Starbucks could become is a utility. Scale and ubiquity creates complexity. Complexity demands efficiency."
— Howard Schultz, 2024
Every business that scales faces two problems at once.
The first is operational: how do you take the thing the founder does well and get a thousand other people to do it without the founder in the room? Howard Schultz was probably a wonderful barista. But when you've got managers managing managers managing a bunch of college kids across 35,000 locations, you need a system. The hand-pulled espresso becomes an automatic machine. The barista's intuition becomes a training manual. The founder's judgment gets compressed into a playbook. I call this the management tax: the cost, in both dollars and degradation, of codifying one person's vision into something replicable.
The second problem is commercial: the niche market you started with doesn't exist at scale. Schultz's original Il Giornale served a tiny audience of Seattle coffee obsessives. To fill 15,000 stores, you need to sell to people who don't know what a pour-over is and don't care. You need Frappuccinos. You need caramel drizzle. You need a product that requires no education to appreciate, that doesn't challenge anyone, that converges on a flavor and aesthetic so broadly palatable it offends no one. The product waters down because the customers water down. There simply aren't enough connoisseurs to sustain a global footprint.
These two forces work in tandem. The process gets simplified so anyone can execute it. The product gets simplified so anyone can consume it. And by the time both forces have done their work, you have something fundamentally different from what you started with. Starbucks in Manhattan in 2026 is basically a public restroom with a loyalty program attached. That's not contempt. It's genuine awe at the infrastructure. But it has nothing to do with coffee.
This is the tradeoff of scale, and nobody in the MBA canon really addresses it head-on. There are frameworks for why firms exist (Coase) and how firms compete (Porter). But not much on this more structural question: which types of work can survive the process of being scaled through a firm, and which can't? At a fundamental level, this is as much an anthropology question as an economics one. How do we organize ourselves to create abundance through scale, leverage, and our own ingenuity? And where are the natural boundaries of that process?
Once you see the management tax, you start noticing a pattern in how entire sectors evolve:
Phase 1: "Highly fragmented." Thousands of small operators, each dependent on the founder's judgment. Industry reports describe the sector as "ripe for consolidation," which is analyst-speak for: someone should be able to figure out how to reduce the management tax here.
Phase 2: "Some scaled players emerging." A few operators crack parts of the value chain. They invest in technology and training, start acquiring smaller shops. If an investment banking pitch says "the company is a leading consolidator in a highly fragmented $X billion market," you're looking at Phase 2. The management tax has been partially solved.
Phase 3: "Rapidly consolidating." The playbook is proven. Private equity floods in. Roll-up strategies proliferate. The top five players go from 5% to 25% market share in a decade. The product homogenizes as the scaled players chase a broader and broader customer base.
Phase 4: "Mature oligopoly." Two or three players dominate. The original craft version of the product survives only in premium niches. This is where hotels arrived decades ago, where accounting and tax prep have largely settled, where coffee is heading nationally.
Some sectors never leave Phase 1. Fine dining. Architecture. Psychotherapy. High-end legal work. These are permanently fragmented because the product is the specific judgment of a specific person. The management tax is irreducible and the product can't survive being watered down. You can't make a great therapist's empathy "broadly accessible" without it ceasing to be therapy.
The interesting question is always: which sectors that look permanently fragmented are actually on the verge of Phase 2?
If you want to understand how sectors move through these phases, look at what happened to home services and car washes. A few PE pioneers entered car washes as early as 2014, but in 2016 only two of the top ten car wash chains were PE-backed. By 2021, all ten were. HVAC followed a similar arc. What triggered the wave?
A confluence of at least five things:
The software and automation layer arrived. For home services, platforms like ServiceTitan (founded 2012, IPO December 2024) gave HVAC, plumbing, and electrical businesses a real operating system for the first time: dispatching, scheduling, invoicing, customer management, all in one cloud-based platform. Before ServiceTitan, the knowledge of how to run an HVAC shop lived in the owner's head and a filing cabinet. That's impossible to scale. Once it lived in software, a PE-backed platform could acquire a five-person shop and plug it into a standardized tech stack within weeks. For car washes, the tech story was different but equally important: modern POS systems replaced cash-heavy operations, making revenue auditable and bankable, while automation reduced staffing needs to the point where an express tunnel wash could run with minimal labor. In both cases, the management tax on integrating acquisitions dropped overnight.
The Boomer retirement wave created a supply glut. Baby Boomers own roughly 40% of US small businesses, and about 10,000 retire every day. An estimated $10 trillion in small business assets will change hands this decade. Most owners have no succession plan; their kids don't want the business. This isn't specific to car washes or HVAC, but it created the backdrop: a generation of owner-operators looking for exits at the exact moment PE-backed platforms were looking for acquisitions. Supply of sellers and demand from acquirers converged simultaneously.
The subscription model. Car washes were transformed by the shift from pay-per-wash to unlimited monthly plans. This converted a sporadic, weather-dependent business into something with recurring revenue and predictable cash flows. Same assets, radically different financial profile. Suddenly a car wash looked like a SaaS business to a PE underwriter.
ZIRP flooded PE with capital. The post-GFC zero-rate environment meant PE firms were raising record funds and needed deployment targets. With large-cap buyouts increasingly competitive, firms moved down-market into fragmented industries where they could buy at 4-5x EBITDA, professionalize, bolt on acquisitions, and exit at 8-10x. The multiple arbitrage was enormous.
The playbook got codified. The concept of a services roll-up became well-understood within PE. Acquire a platform, centralize back-office, standardize the tech stack, bolt on, drive margins, exit. Fifteen years ago this was being invented. Today it's table stakes.
None of these factors alone would have been sufficient. Software without sellers doesn't help. Sellers without capital doesn't scale. Capital without software can't integrate. It was the confluence.
Some businesses also simply require scale to make the unit economics work. You can't have a boutique airline that isn't exclusively serving the ultra-wealthy, because aviation's fixed costs demand volume. Similar logic applies to clothing brands and grocery chains. In these sectors the question was never whether consolidation would happen, but when.
Then there are the stubborn holdouts, and the reasons for their stubbornness vary. Some sectors resist consolidation because of trust and relationships. Most American men still prefer "their guy," the barber who knows their fade, remembers their kids' names, who they can actually talk to. Same with mechanics. Same with doctors. Trust doesn't systematize.
But there's something else going on too, beyond just the relationship. It matters whether the purchase is high-consideration or low-consideration, and whether the act of purchasing itself carries any meaning. Getting your car washed at Bob's Car Wash vs. SuperLazer 3.0 with a PE owner and 40 locations doesn't say anything about you. The two outcomes are basically comparable, except the PE-backed one is probably more convenient and has a loyalty program. Same with HVAC: who cares? Get the cheapest service who can come quickly and fix the damn thing. These are low-consideration, low-identity purchases, and they consolidate easily because nobody derives joy or meaning from the choosing. The management tax is the only barrier, and once software and capital crack it, the sector rolls up fast.
Contrast that with where you get your coffee, your haircut, your wine, your clothes. These are categories where the choice is part of the product. Going to Sey instead of Starbucks is a statement, not just a purchase. The revealed preference in these categories is more fragmented than the "chains always win" narrative suggests. People will pay more, travel farther, and accept inconsistency to get the version that feels like theirs.
Different societies draw the line between scale and fragmentation in very different places, and the reasons aren't always structural.
German pharmacy law caps ownership at one main pharmacy plus three branches per pharmacist. Corporate ownership is banned. No chains exist. The result: roughly 18,000 independent pharmacies staffed by qualified pharmacists who provide real consultations and prepare custom medications. Where an American might see a massive unrealized consolidation opportunity, others see a reasonable equilibrium in which the pharmacist works not-too-many hours and enjoys professional satisfaction while the patient gets their eczema cream at a fair price.
This isn't inefficiency. It's a different equilibrium. The European utility function, in many sectors, weights free time and professional dignity more heavily than profit maximization. Germans are stubborn about their independent Laden and will actively vote with their feet away from chains. Some cultures' default state is fragmentation, not because of regulation, but because of social norms. Not every society's instinct is to maximize.
These equilibria shift. Germany's independent bakeries have been eroded by chains like Backwerk, Kamps, and Le Crobag. The German response: regulate the chains on quality, then let them scale. The Overton window on what gets consolidated moves over time, even in cultures skeptical of scale.
Here's where the story gets bigger than any individual industry.
The history of scalability is the history of how many humans you need to capture how much value. That ratio has been moving in one direction for two centuries.
In 1911, the Triangle Shirtwaist Factory in lower Manhattan employed 500 workers on a few floors, each performing repetitive garment work. The owner extracted a thin margin from each worker's labor. Profit was roughly a function of headcount. The fire that killed 146 people became a defining moment for the labor movement because it laid bare the logic: humans were the unit of scale, and they were treated accordingly.
For most of the 20th century, the biggest companies were the biggest employers. Revenue and headcount tracked closely.
Then the ratio started to decouple.
| Company | Year founded | Firm Value ($bn, 2026 dollars) | Employees | Value per Employee ($) |
|---|---|---|---|---|
| Standard Oil (in 1911) | 1870 | ~36 | ~100,000 | ~360,000 |
| U.S. Steel (at formation in 1901) | 1901 | ~29 | ~168,000 | ~172,619 |
| Dutch Royal Shell | 1907 | ~250 | ~100,000 | ~2,500,000 |
| McDonald's | 1940 | ~220 | ~150,000 | ~1,466,667 |
| Walmart | 1962 | ~950 | ~2,100,000 | ~452,381 |
| Microsoft | 1975 | ~2,840 | ~228,000 | ~12,456,140 |
| Apple | 1976 | ~3,650 | ~166,000 | ~21,987,952 |
| Alphabet | 1998 | ~3,630 | ~183,000 | ~19,836,066 |
| OpenAI | 2015 | ~730 | ~4,500 | ~162,222,222 |
| Anthropic | 2021 | ~380 | ~4,500 | ~84,444,444 |
Sources: Company information, Global Financial Data, Google, 'The Prize: The Epic Quest for Oil, Money, and Power' (Daniel Yergin, 1990)
The trajectory is clear: Standard Oil and U.S. Steel, the most dominant companies of their era, generated roughly $170K-$360K in value per employee. A century later, Walmart sits at almost exactly the same ratio. Then the curve breaks. The big tech companies hit $12-22 million per employee. And the AI labs are north of $80 million, heading toward $160 million.
But the trajectory points somewhere more radical. There's serious talk about the first $10 billion company with 10 employees. Or a $1 billion company with one. The infrastructure already exists: cloud computing, AI agents, APIs, global payment rails. The management tax on a one-person company has always been zero. What's new is that the scale ceiling has been obliterated too.
We may be witnessing two things happening simultaneously that used to seem contradictory: the atomization of the firm at the employee level and the unprecedented expansion of scalability at the service-delivery level. The company gets smaller. Its reach gets larger. The Triangle Shirtwaist Factory needed 500 people for a regional market. A future AI-native company might need 5 for a global one.
This raises the most interesting question: does AI break the management tax framework entirely? Historically, scaling required two compromises: simplify the process so anyone can execute, and simplify the product so anyone can consume. But what if AI handles the systematization while the founder's taste stays embedded in every decision? What if you could run a business with the quality and specificity of a single-location obsessive at the reach of a global chain?
There's a reason to think the demand side of this equation is shifting too. The cost of connoisseurship is falling fast. Twenty years ago, knowing about specialty coffee or natural wine or Japanese denim required physical access to niche communities, expensive travel, or the right social circle. Today, that knowledge is essentially free. YouTube, Reddit, TikTok, and a thousand niche newsletters have democratized taste at an extraordinary rate. Coffee snobbery, once the preserve of a tiny subculture, is now basically a hallmark of the upper-middle class. The same is happening in wine, food, fitness, skincare, and dozens of other categories. The audience for "the good version" of things is growing, and the barrier to joining that audience is approaching zero.
This matters because the second problem of scale, the product watering down because the niche market doesn't exist at scale, may be less binding than it used to be. If connoisseurship is being mass-produced by the internet, then the addressable market for high-specificity products is larger than it's ever been. Maybe the audience of Sey Coffee (one of my favorite coffee shops in NYC) in 2017 was a few thousand people in Brooklyn. In 2026, there are probably tens of millions of people around the world who would gladly pay top dollar for a fruity Panamanian Geisha. They just don't all live in the Bushwick.
So maybe AI solves the supply side (replicating the founder's judgment without degrading it) at exactly the moment the internet has solved the demand side (growing the audience that can appreciate what the founder built). If those two things converge, you could imagine a world where Sey Coffee doesn't lose one tiny bit of the magic and still scales. Not by opening 10,000 identical locations staffed by college kids, but by using AI to extend the founder's taste and judgment into new contexts, new cities, new formats, without the lossy compression that has historically made scaling synonymous with dilution.
I'm not sure that's possible. The product-watering-down problem might be structural, not operational. But the gap between "the good version" and "the scaled version" is narrower than it's ever been.
Paul Graham's famous advice to "do things that don't scale" is usually read as permission to stay scrappy in the early days. But I think there's a deeper reading. What he's really saying is: find a business where the market is enormous but the process of scaling is uncharted. The management tax is high today, but not irreducibly so. The opportunity is in the gap between a product that clearly works and a scaling playbook that doesn't yet exist. That's where ServiceTitan was in 2012. That's where the first HVAC roll-up operators were in 2015. That's where a lot of AI-native services businesses are right now.
The management tax has defined the boundaries of scale for as long as businesses have existed. For the first time, it's not obvious that it always will. The question is whether you're in a sector where it's falling, and whether you can see the confluence of forces that will push it lower before everyone else does. That's the bet. Everything else is just execution.