Summary

Östberg now sees food delivery as a scaled, multi-winner market in which execution drives roughly 80% of the outcome and competition 20%. Leaders can coexist profitably once they have sufficient scale; customers are more loyal and less price-sensitive than assumed, while temporary vouchers rarely change their default platform. “It’s not vice versa”: being number two limits absolute scale, but does not preclude attractive economics.
Delivery Hero’s defining contrarian bets were entering logistics in 2015 and selling Germany in 2017. Its first logistics initiative lost €6 million, yet the company rebuilt it and invested tens—and eventually hundreds—of millions because the customer experience was clearly better. Selling the home market removed an emotionally consuming distraction: “Once we sold Germany, there was no excuse for not delivering” internationally.
The nearly $200 million lost on Gorillas taught Östberg to trust the long-run signal in cohort and user-acquisition models, which behave “almost like gravity.” Repeated upside surprises persuaded him that his fundamental model might be wrong; he ultimately concluded that Gorillas’ model might have worked, but the market changed, capital became unavailable and the business burned too much. Vouchers and discounts can temporarily boost growth, but he warns against abandoning core investment principles for momentum.
Delivery Hero evaluated Glovo through 20-plus years of cohort evidence across roughly 70 markets, not through public-market sentiment. Östberg says healthy cohorts rarely deteriorate unless the operator “screws up” through poor service, missed product shifts or subsidy-led acquisition. That data implied a ten-year growth path and eventual 10–13% business margins, despite adding another negative €330 million going into 2022.
COVID-era confidence turned an avoidable dilution into a balance-sheet lesson. Believing Delivery Hero had a clear path from a roughly $30 billion valuation to $100 billion, while feeling undervalued at $35 billion, management preferred debt and resisted issuing equity. After a 5% decline seemed “too late,” another 10–15% followed, then more than 50%. Östberg’s conclusion is blunt: “It’s better to dilute and not think too much about it.”
Organizational speed compounds, but irreversible capital decisions require a different tempo. A company producing 25% more each year can build “a 10 times better product” over a decade, so most reversible decisions should be made, measured and corrected quickly. Large investments are different doors; simplicity, clear country-level ownership and output-based accountability preserve speed without treating every decision alike.
Subsidized volume is counterfeit scale unless every order has a credible path to sustainable economics. Delivery Hero’s Thailand operation jumped from a few thousand daily orders to 400,000 in under a year, then fell to roughly 25% of that level as cheap pricing ended. Competitor balance sheets therefore matter less than returns: “You can grow to 400,000 daily orders…but it’s not worth anything unless you’re being self-sustainable.”
Östberg’s next bets are quick commerce, AI adoption and delivery automation—not an investor-facing AI narrative. He expects groceries and other verticals eventually to exceed 50% of Delivery Hero’s business, believes average companies deploying AI may benefit more than the Magnificent Seven building it, and estimates drones might handle about 25% of deliveries in roughly ten years. Ground robots should account for a larger share sooner because they can serve city centers despite moving more slowly.

Deep dive

1. Booms make operators look better than they are

Östberg traces his endurance back to daily cross-country ski training in Sweden’s dark forests, while purpose and focusing only on what he can influence helped him stop optimizing for approval.
Östberg’s recurring humility lesson began with making—and quickly losing—money in the late-1990s stock market, then resurfaced when COVID demand sent the business “through the roof.” His rule across cycles: “You’re not as good as you think you are in the good times, and probably not as bad as you think in the bad times.”
The COVID hindsight is specific: Delivery Hero should have reduced risk, saved more cash and raised equity rather than debt. Management believed it could become a $100 billion company and resisted dilution at roughly $30 billion, while feeling undervalued at $35 billion; after successive 5% and 10–15% declines, the stock fell more than 50%. “We got greedy” was Östberg’s diagnosis.
Delivery Hero did react before the wider collapse, shutting Germany and Japan and removing $100–200 million from the budget at the end of 2021. Morale initially held because KPIs remained strong and employees expected a temporary drawdown; the harder damage came when valuation failed to recover for three years. Östberg thinks the company was somewhat overvalued then and is undervalued now.

2. Contrarian leadership created the logistics platform

Östberg’s leadership test is consistency between belief and action: strong leaders occasionally go “against the stream,” and successful judgment compounds followership and trust. Leaders who suppress their own convictions instead generate skepticism because the organization can see that decisions lack a coherent internal logic.
Logistics was the defining specimen. In 2015, when few believed platforms should control delivery, Delivery Hero’s first initiative lost €6 million and was shut down. Instead of abandoning the thesis, it rebuilt and invested tens, then hundreds, of millions because the service was unambiguously better for customers; the unresolved task was making the economics work.
Selling Germany in 2017 was equally contrarian. Östberg saw a smaller opportunity than the population suggested, a strong incumbent and little prospect of later consolidation under antitrust rules. More importantly, every home-market issue pulled management away from international operations; selling Germany converted overseas execution from a secondary concern into the company’s only route forward.

3. Scale supports several winners when service remains strong

Östberg explicitly changed his mind on winner-take-all dynamics. He points to two profitable large players in the US and France, three in the UK—Just Eat, Deliveroo and Uber Eats—and several Delivery Hero markets where two or three operators make money. He says one can argue consolidation might add value, but antitrust may prevent combinations in markets such as the UK.
The host’s pushback is that courier density, customer density and unit economics appear to favor a single winner. Östberg’s answer is that customers become “incredibly loyal” once a platform reaches scale and delivers reliably; a €5 or €10 voucher might purchase one rival order, but users typically return to the service that remains top of mind.
Scale is necessary, yet competitors having scale does not mechanically worsen Delivery Hero’s economics. Poor execution does: Delivery Hero entered Spain, Italy, Poland and Norway as the number two, three or four player and ultimately became the largest because incumbents failed on quality. “The day you don’t deliver a good service…it can be disrupted fast.”
His revised model is “80% our own execution” and perhaps 20% competition, reversing his earlier belief. A number-two operator will make less in absolute terms because it has less scale, but can still be profitable. That also informs his view that profitable, growing Deliveroo is undervalued, even if an acquisition cannot solve the UK’s three-player structure.

4. Speed compounds only when ownership stays legible

Another changed mind is the value of subtraction. Östberg once asked organizations to absorb more initiatives than they could handle; now he favors concentrating resources on a few priorities because “complexity is a killer of speed.”
The host frames CEOs as capital allocators, but Östberg ranks culture, pace, data and direction above allocation itself. In a well-run organization, most daily choices—another product, country or operating investment—become evident from returns. The rare CEO-specific task is making a consequential bet when data is incomplete, using accumulated knowledge of the industry and organization.
Speed matters because product improvements compound: 25% more organizational output in one year looks incremental, but repeated over ten years can produce “a 10 times better product.” Most decisions are reversible, so teams should decide, collect evidence and iterate. Large investments are non-reversible doors, which is why Gorillas required more caution than an ordinary operating experiment.
At scale, Delivery Hero preserves urgency through country-level autonomy and measurable ownership. Employees need to see how their work changes an outcome; broad company goals obscure that link and reward visible effort or internal maintenance rather than customer output. Small accountable units make strong and weak performance easier to identify.

5. Gorillas showed why momentum cannot repeal unit economics

Delivery Hero’s underlying quick-commerce decision was right, in Östberg’s view: customers want groceries and other items rapidly, and those categories should eventually exceed 50% of the business. Gorillas was an attempt to gain exposure and learning in geographies Delivery Hero could not afford to build simultaneously itself. Östberg says its business model might have worked, but the market changed, these companies could no longer raise capital and they burned too much.
The company invested close to $200 million and recovered only a small portion. Month after month, Gorillas acquired more users and grew faster than Östberg’s models predicted; repeated forecasting misses persuaded him that perhaps “my models don’t work.” His eventual conclusion was the reverse: cohorts and acquisition economics are “almost like gravity. It always works.”
Temporary vouchers, discounts and abundant funding can lift reported growth above its foundation. Östberg had seen the same pattern in an Asian market that repeatedly beat plan before falling back, and the inverse in sound businesses whose expected tailwind took longer to appear. The discipline is to update on genuine facts without letting short-term momentum erase core underwriting principles.
He concedes the host’s Keynesian challenge: investors sometimes should follow momentum—“When the music is there, you dance”—but must remain careful. The $200 million also looked smaller when Delivery Hero was worth $35 billion, roughly half a percent of market capitalization; had valuation fallen toward $5–10 billion, the same loss would have been material, especially without enough cash on the balance sheet.

6. Delivery Hero bought local teams, then supplied the scaling machine

Delivery Hero completed more than 35 acquisitions, but most were small purchases of local entrepreneurs rather than finished platforms. PedidosYa grew from about 60,000 monthly orders at acquisition to nearly 20 million; Talabat began around 70,000–80,000 and later became, in Östberg’s description, a $10 billion business. Delivery Hero contributed automation, measurement, investment discipline and efficiency.
The difficult part of M&A is not buying but “mak[ing] something good out of that company.” Delivery Hero’s advantage was retaining entrepreneurial ownership while supplying a scaling system, often buying roughly a year of head start. That produced many 100-times or even 1,000-times returns, albeit on initially small amounts.
Early on, the company bought aggressively because Just Eat and Takeaway had operated for 10–15 years and appeared impossible to catch organically. Today Östberg prefers building: with Delivery Hero undervalued, acquisitions are expensive in relative currency, and adding 3–5% of business size achieves little if integration distracts the existing organization by 2–4%.
Glovo met the higher bar. Delivery Hero’s 20-plus years of data across roughly 70 markets suggested cohorts were highly predictable; the host challenged whether later customers would sit farther from the ideal profile, but Östberg said deterioration consistently traced to operator errors—late logistics, Colombia’s delayed multi-vertical response to Rappi, or voucher-dependent acquisition—not natural cohort decay.

7. Glovo’s losses obscured a measurable ten-year earnings path

For Glovo, repeat behavior, monthly acquisition and order-level margin made the trajectory calculable. Delivery Hero expected margins to reach 10–13% for the business, then could subtract predictable marketing and overhead. Public investors saw the negative €330 million contribution as profitability became the market’s priority; Delivery Hero saw “how this business is gonna grow over the next 10 years.”
The host’s price challenge was direct: was roughly $2 billion correct? Östberg answers through dilution rather than headline consideration. Delivery Hero paid in stock while valued around $30 billion, issuing roughly 5–6%—in any case less than 10%—for a business he expects to outgrow the group. The later 75% stock decline made the timing painful but limited effective dilution.
Cohort strength still depends on sustainable customer promises. Low minimum baskets create more ordering occasions, but a €5 order may not be economical today; if economics later force a €15 minimum, frequency falls. Logistics and quick commerce took years to break even per order because operations needed optimization, not because Delivery Hero charged too little.

8. Subsidies create volume, not durable enterprise value

Heavily funded competitors mostly hurt incumbents by making them nervous enough to copy irrational spending. Discounts disproportionately recruit deal hunters, invite fraud and disappear when incentives end. Hence Uber’s or DoorDash’s balance sheet is not decisive: “It’s gonna be the return” that limits spending, because even profitable platforms cannot sustainably fund bad orders.
COVID did not materially advance the industry in Östberg’s telling. Cohorts jumped above trend, management hoped the higher level would persist, and activity ultimately returned to its old trajectory. Thailand was the sharper internal example: Delivery Hero raced from a few thousand daily orders to 400,000 in under a year, then fell to about 25% of that volume when cheap pricing ended.
A new market should demonstrate credible cohorts, lifetime value and scalable acquisition cost within roughly a year. The strongest markets may then lose money for several years because high-return acquisition deserves maximum investment; weak markets can reach break-even fastest precisely because they warrant little growth capital. Bootstrap discipline, as at OnlinePizza, preserves cash but can take a very long time and fail to maximize the business.
The host argues emerging-market revenue can receive discounted multiples and depress enterprise value. Östberg distinguishes building for a near-term sale from building for cash flow and shareholder returns: if customer returns are sound, geography should not dictate investment. “That’s the difference between price and value. We try to be focused more on driving value than driving price.”

9. Europe needs a level field while technology changes delivery

Östberg accepts some Europe criticism but remains an entrepreneur’s optimist: the continent retains educated talent, strong infrastructure, welfare systems, functioning democracies and decent balance sheets. His priorities are faster talent immigration and less cumulative bureaucracy across GDPR, sustainability reporting, the Pay Transparency Directive and the Accessibility Act—each defensible alone, burdensome together.
The sharper complaint is asymmetric enforcement. Östberg believes US and Chinese competitors sometimes circumvent rules while European companies receive heavier scrutiny because regulators can reach them more easily. In discussing Oscar’s regulatory pressure, his operating principle is to be “in the trenches in the details,” share responsibility for collective decisions and avoid blaming operators for pressures outside their control.
His contrarian AI call is that “the largest beneficiary” may be the average company deploying AI, not necessarily the Magnificent Seven building it. Delivery Hero feels no need for a cosmetic AI story, only pressure to become more efficient. For a ten-year holding outside Delivery Hero, he still chooses Amazon, while noting returns must now come from growth rather than multiple expansion.
Automation arrives unevenly: Östberg estimates drones could perform roughly 25% of deliveries in perhaps ten years, with a typical flight taking about three minutes plus loading. Noise, regulation and landing constraints limit dense cities; ground robots are slower because they follow roads and signals, but should account for a larger share sooner because “they can get you almost anywhere.”

Mitchell Green, Founder @ Lead Edge Capital: Why Traditional VC is Broken Palmer Luckey on the Next Wave of Military Tech Powered by AI | EP #158