One of the biggest wars over urban food began with an experiment that fit into a three-hour workday window and cost about as much as a sandwich and coffee at a decent café.
On August 26, 2014, something happened in Santa Monica that few people noticed outside California tech blogs. Uber—the company that had by then turned hailing a cab into a swipe on a screen—launched a service called uberFRESH. The experiment was led by Jason Droege. The terms were almost comically narrow: lunch delivery Monday through Friday, from 11:30 to 14:30, a fixed price of $12 per dish, a limited menu, and the courier pulling up to the curb—curbside, as if you had summoned a car, except instead of a ride you got food. Delivery time: around ten minutes. The whole operation ran on the same infrastructure that coordinated drivers: geolocation, routing, assigning the nearest available worker.
The formula was transparent to the point of naivety. By then, Uber had already learned to move people from point A to point B, optimizing the route in real time. Lunch is also an object that needs to move from point A to point B. A courier is the same driver, only with a container instead of a passenger. The algorithm does not distinguish what exactly it is carrying: a person or pad thai. The difference lies in the expectations at either end of the route, and that is where the whole story begins—the story that would later unfold across dozens of countries and billions of dollars. But in August 2014, it was just a test: would the habit of ordering a ride work if food were offered instead?
The test worked. In 2015, uberFRESH was renamed Uber Eats and—crucially—released as a separate app. The first city to get a standalone customer and dispatch app was Toronto. This is an important detail: from that point on, food delivery stopped being a feature inside a ride-hailing app and became a separate product with its own logic. The menu was no longer limited to one fixed-price lunch—the app became a storefront for restaurants. An establishment listed its menu, a user chose, the algorithm assigned a courier, the courier picked it up and delivered it.
In 2016, the service launched in London and Paris, and the scale became European. The model essentially reproduced the logic of ride-hailing: the platform does not own restaurants, cook food, or employ couriers on staff—it connects supply and demand through an algorithm, taking a commission from every transaction. The restaurant gets orders that would not have reached it without the app. The courier gets flexible income. The user gets food without getting off the couch. The platform gets a cut from all three—or, more precisely, from the restaurant and the user, because it pays the courier rather than charging them.
What fascinates me personally is the purity of the abstraction: the same software architecture serves both the movement of people and the movement of plates. But purity of abstraction is not yet a business, and to understand why the model began to crack, you have to look at commissions.
The restaurant business is one of the lowest-margin businesses in the world. An establishment’s average profit ranges from 3–9%. Delivery-platform commissions, meanwhile, had risen to 13–40% of the order value by the end of the 2010s. The arithmetic is brutal: if your restaurant makes 7% on every dish and the platform takes 30%, then every order through the app loses money. You are literally paying extra for your food to leave for the customer. So why did restaurants agree to it? Because refusing the platform meant losing visibility: if you are not in the app, you are not in the mind of the user opening Uber Eats on their lunch break. The storefront became a trap.
In February 2019, Uber Eats cut its commission from 35% to 30%—a gesture that sounds like a discount, but for a restaurant with a 5% margin, the difference between 35% and 30% is the difference between catastrophe and merely a loss. A 2020 class-action lawsuit—Davitashvili v. GrubHub Inc.—exposed another practice: according to the plaintiffs, platforms forced restaurants to keep the same prices for dine-in and delivery. In other words, a restaurant could not offset the commission by raising its in-app price—that would violate the contract terms. The establishment found itself subsidizing the platform’s convenience out of its own profits.
At the same time, user fees were changing too. In August 2018, Uber Eats abandoned its fixed $4.99 delivery fee and switched to distance-based pricing—from $2 to $8. The logic is clear: delivery two blocks away and delivery across the city involve different costs. But for the user, this meant that a cheap order from a distant restaurant was no longer worth it, and the geography of choice narrowed to the nearest few blocks.
When Uber Eats arrived in Latin America, it found not a desert but a battlefield. PedidosYa had been operating since 2009—a Uruguayan startup that began with food orders through a website and had acquired its own call center by the end of its first year. iFood launched in Brazil in 2011 and was already firmly embedded in Brazilian users’ habits by the time Uber Eats arrived. Rappi, the Colombian “super app” that delivers everything—from food to cash from an ATM—appeared in 2015 and expanded aggressively across the region.
Latin America proved an ideal environment for platform delivery for reasons unrelated to technological progress. Dense urban development—Mexico City, São Paulo, Bogotá—meant short routes and a high concentration of both restaurants and customers per square kilometer. Growing smartphone penetration in the 2010s made the app an accessible channel for millions. Cashless payments, though slower to take hold than in Europe, were gaining ground. And then there was precarious employment: in a region where the formal labor market cannot absorb everyone looking for work, courier work on a platform became a source of income for hundreds of thousands of people willing to ride a motorcycle or pedal a bicycle for twelve hours a day.
But a global platform is not a magic key to a local market. Uber Eats ran into the fact that local players knew their cities better: they had established relationships with restaurants, adapted payment methods, an understanding of local habits. In Brazil, iFood held a dominant position by the early 2020s, and competing with it turned out to be more expensive than it was worth. In November 2020, Uber Eats left Argentina and Colombia. In January 2022, the company announced it would shut down restaurant-delivery operations in Brazil by March that same year—the region’s largest market, where iFood had proved insurmountable.
The story of Uber Eats after 2018 is, roughly speaking, the story of a company that realized it could not be everywhere. In March 2018, its operations in Southeast Asia were sold to Grab—the Singaporean super app that knew the region from the inside. In September 2019, it withdrew from South Korea, where the delivery market had been mature long before any Western platforms appeared. In January 2020, its Indian business was sold to Zomato (the full exit was completed by August 2022). In May 2020, the service shut down in seven countries at once, including the Czech Republic, Romania, and Ukraine, while in the UAE, operations were transferred to Careem.
Each of these exits was an acknowledgment that the food-delivery business is tied to a particular urban fabric, local payment habits, legal regime, and competitive environment. You cannot export an algorithm from San Francisco to Seoul and expect it to defeat Baemin, which has a ten-year head start and cultural intuition. At the same time, Uber Eats was not simply retreating—it was reallocating resources. In July 2020, the company bought Postmates for $2.65 billion, strengthening its position in its home US market. And in May 2025, it acquired 85% of Turkey’s Trendyol Go for $700 million, betting on a market where it saw potential.
Returns are a separate story. In early 2026, Uber Eats returned to Argentina, more than five years after leaving. What changed in the Argentine market over that time to make the return worthwhile is a question that has no public answer yet.
Throughout this story, there is a character who moves the food, takes risks on the roads, gets soaked in the rain, and yet is legally an employee of neither the restaurant nor the platform. The courier is an independent contractor, a “partner,” a freelancer—depending on which term the platform prefers and what legislation is in force in the country. This ambiguity is not a bug in the model but its load-bearing structure: if couriers become employees, the platform will have to pay minimum wage, sick leave, insurance, and the entire economics will fall apart.
In February 2024, couriers for Deliveroo and Uber Eats in the United Kingdom held a joint strike—Deliveroo riders demanded a minimum payment of £5 per order. The very fact that couriers had to strike for a guaranteed minimum per delivery says that without such a minimum, pay can be lower. The algorithm assigns orders, the algorithm calculates pay, and for the courier that algorithm is simultaneously boss, accountant, and HR department—one they cannot talk to.
Verification is a separate issue. Uber Eats introduced a courier identity-check system based on AI-powered facial recognition. The technology is meant to ensure that the person behind the wheel or on the bicycle is the person registered in the system. But facial-recognition algorithms are known for their errors, especially with dark-skinned people, and incidents of false verification—when a courier cannot pass the check and loses access to income—have become the subject of investigations and complaints. A system meant to protect becomes a barrier, and that barrier hits those who are already vulnerable disproportionately hard.
If you trace the line from that curb in Santa Monica to today, you can see that something larger happened than the emergence of a convenient way to order dinner. Food became part of urban logistics—a flow like passenger transport or courier mail. Every order is a route, a burden on the road network, data about preferences, a commission divided between the platform and the restaurant, and a person on a motorcycle whose labor is paid according to a formula they did not choose.
Uber did not invent food delivery—PedidosYa had been doing it in Latin America five years before uberFRESH, and Chinese and Korean platforms had been operating even earlier. But Uber did something else: it showed that the infrastructure for moving people and the infrastructure for moving food are, essentially, the same infrastructure. The app is the storefront. The algorithm is the dispatcher. The courier is outsourced logistics for a restaurant that never hired its own driver. The model proved replicable, and everyone replicated it—from iFood to Rappi, from Grab to Zomato, each in its own market, each with its own adaptations.
It feels fair to end not with a conclusion, but with an observation. In 2014, lunch in Santa Monica cost $12 and arrived at the curb in ten minutes. By the mid-2020s, food delivery had become an industry in which global and local platforms battle for control over the flow of data, routes, and commissions; restaurants calculate whether being in the app is worthwhile at all; and couriers strike for five pounds per order. The curb is still the same, but everything that happens between the kitchen and that curb has turned into an economy of its own—with winners, losers, and vast numbers of people pedaling every day somewhere between the two.