The food hall that quietly bought the delivery app

On July 16, 2026, a company called Wonder announced a $650 million Series D at a $9 billion pre-money valuation. Read that as a headline and it sounds like another late-stage food-tech round. It isn't. Here's the deal: the company doing the raising is the same company that owns Grubhub, owns Blue Apron, and owns the robot Sweetgreen built and then gave up on.

That last part is the tell. Wonder is not a delivery app with a clever algorithm. It runs roughly 140 physical food halls across 10 East Coast states — buildings, leases, kitchens, staff. Each hall stacks multiple restaurant brands under one roof, some owned outright, some licensed from chefs (Bobby Flay among them), so you can order from four different "restaurants" and get one bag. That's a real-estate-and-operations business wearing a technology jacket, and it's the opposite of the asset-light marketplace model DoorDash and Uber Eats spent a decade perfecting.

The valuation math is worth pausing on. Wonder's prior round was roughly $600 million at a $7 billion valuation in 2025. So $9 billion pre-money is about a 29% step-up in roughly a year — solid, but not the kind of vertical repricing you see in AI. And the investor list is where this gets interesting: alongside returning venture backers Accel, GV, and NEA, the new money came from ARK Invest (Cathie Wood), certain funds managed by AllianceBernstein, and funds managed by Kayne Anderson Rudnick. Those are not venture funds. Those are public-market managers.

Which is exactly the point. Founder and CEO Marc Lore told Fortune the company will be "ready and prepared to go public early next year." Important hedge up front, because this matters more than anything else in the story: that is a stated intention, not a filed event. No S-1 has been made public. No registration statement exists in the record. Lore said Wonder will be ready to go public — which is a very different sentence from "Wonder has filed to go public." Hold that distinction through everything below.

The players — a serial e-commerce founder, a fallen delivery giant, and a discarded robot

Marc Lore is the through-line. He built Quidsi (Diapers.com) and sold it to Amazon, then built Jet.com and sold it to Walmart for $3.3 billion, then ran Walmart's US e-commerce. He is, in other words, a man with a track record of building a thing large corporations want to buy — and now he's built something he says he wants to take public instead. He also disclosed he has celiac disease, which he frames as the personal origin story behind Wonder's food-personalization thesis. Fortune reports Wonder has raised roughly $3 billion since founding, though outlets differ on the founding year (some say 2018, some date the current capital stack to 2021).

Grubhub is the asset most people miss. Wonder acquired it in early 2025 for about $650 million, including roughly $500 million of assumed debt, from Just Eat Takeaway. Do the comparison: Just Eat paid $7.3 billion for Grubhub in 2021. That is roughly a 90% markdown in four years. Wonder bought a national delivery marketplace — brand, drivers, merchant relationships, order flow — for less than the amount it just raised in this single round. Whatever you think of Wonder's food, that trade is the reason a $9 billion valuation is even arguable.

Blue Apron came earlier, acquired in 2023 for about $103 million. Meal kits were a busted category by then, but Blue Apron came with recipe IP, supply relationships, and a national cold-chain fulfillment habit. Cheap optionality.

The Infinite Kitchen is the robotics piece, and here's the correction that most coverage buries: Wonder didn't build it. It bought Spyce Food Co. from Sweetgreen for $186.4 million — $100 million cash plus $86.4 million in Wonder Series C preferred stock — agreed November 5, 2025 and closed December 29, 2025. Sweetgreen kept a license to continue deploying Infinite Kitchen in its own stores. So the "AI-driven kitchen automation" narrative around Wonder is substantially an acquired capability, not a homegrown one. Wonder describes it as the only fully automated bowl-making system in commercial production.

Then there's the money-adjacent cast. Goldman Sachs & Co. LLC, Jefferies, and J.P. Morgan served as placement agents on this round — precisely the bank roster you'd expect to reappear on an IPO syndicate. And in July 2026, former Chipotle CFO Jack Hartung joined Wonder's board. Hiring the CFO who helped make Chipotle's unit economics legendary, right before saying you're ready to go public, is not a coincidence. It's a governance signal aimed directly at public-market investors.

What actually happened — the round, the numbers, and the ones that hurt

The mechanics first. The round is $650 million — though Fortune wrote "$650 million-plus," so the final size may land higher — at a $9 billion pre-money valuation. Use of proceeds per the press release is broader than a simple "expansion" line: physical expansion, marketplace growth, and investment in technology, robotics and AI. All three legs, not one.

Lore's own framing in the release: "By building the technology, robotics and infrastructure behind a new kind of food platform, we're making high-quality food more affordable, more convenient and available to more people than ever before." NEA Co-CEO Tony Florence added: "It's been exciting to watch the evolution of what Wonder is building — a fundamentally new way for people to access great food, at a level of quality and speed traditional players can't match." And ARK's Cathie Wood: "Wonder is disrupting an industry that has been slow to change with the kind of scalable, innovative model that we look for across the ARK portfolio."

Now the part that doesn't fit in a press release. Fortune, reporting on internal company materials, describes a $618 million adjusted-EBITDA loss in 2026, cumulative cash burn of nearly $2.7 billion through 2029, and positive cash flow only arriving in 2030. Every one of those figures is a projection sourced to Fortune's reporting on internal documents — not audited, not filed, not independently verified. Treat them as directional, not as accounting. On the other side of the ledger, Fortune reports same-service-area sales growing about 20% year over year, which for a physical retail footprint is genuinely strong.

Lore's defense of the burn is the most quotable thing he said: "The economics are often misunderstood. You do need to make substantial investment up front — the robotics, the ingredient library. All those suppress profitability in the short-term. But there's a big prize at the end of the day." He also claimed this round carried "the least amount of protection we've ever offered, on a relative basis" — meaning minimal liquidation preference and structure, which would be a confidence signal. Caveat: that's Lore's characterization. The actual term sheet is undisclosed, so nobody outside the cap table can verify it.

Item Detail
Announcement Series D, July 16, 2026 (PR Newswire + Fortune exclusive)
Round size $650 million (Fortune: "$650 million-plus")
Valuation $9 billion pre-money
Prior round ~$600 million at ~$7 billion (2025) — roughly a 29% step-up
Total raised Roughly $3 billion since founding (Fortune)
Returning investors Accel · GV (Google Ventures) · NEA
New investors Certain AllianceBernstein-managed funds · ARK Invest · funds managed by Kayne Anderson Rudnick
Placement agents Goldman Sachs & Co. LLC · Jefferies · J.P. Morgan
Footprint About 140 food halls (release: 140; Fortune: 135) across 10 East Coast states
Growth From 46 locations in May 2025 to ~140 in July 2026 (press release)
Owned assets Grubhub (early 2025, ~$650M incl. ~$500M assumed debt) · Blue Apron (2023, ~$103M) · Spyce/Infinite Kitchen (closed Dec 29, 2025, $186.4M)
Board Former Chipotle CFO Jack Hartung joined July 2026
Projections (Fortune, unaudited) 2026 adj. EBITDA loss ~$618M · ~$2.7B cumulative burn through 2029 · cash-flow positive 2030 · ~20% same-service-area growth
IPO Stated intention of being ready in early 2027 — no public S-1 filing

Two housekeeping notes on the numbers. Location count is inconsistent across sources — the press release says 140, Fortune says 135 — so "about 140" is the honest phrasing. And Wonder also closed an acquisition of Mighty Quinn's BBQ in the same week as the funding announcement, which tells you the M&A machine hasn't slowed down at all.

Separately, on June 30, 2026, Wonder announced a partnership with Zipline for autonomous drone delivery in Texas. Service is slated to begin January 2027 starting in Dallas, with the majority of Texas locations expected to offer drone delivery by end of 2027, and storefronts being physically designed around Zipline Dropboxes. That last detail is the interesting one — Wonder isn't bolting drones onto existing stores, it's designing buildings around the drone.

What each side gets — why public-market money showed up early

Wonder gets two things, and the second matters more. The obvious one is runway: if the Fortune-reported burn projection is anywhere near right, $650 million doesn't even cover a single year of adjusted-EBITDA loss, which is precisely why an IPO is on the table rather than optional. The less obvious one is price discovery. Getting ARK, AllianceBernstein, and Kayne Anderson Rudnick onto the cap table at $9 billion establishes a reference point that public investors can't easily argue below, because those same investors will be in the room at the IPO. That's the whole reason crossover rounds exist.

ARK Invest gets a thesis vehicle. Wood has spent a decade underwriting disruption stories in categories the market considers boring, and her quote — "an industry that has been slow to change" — is doing exactly that framing work. Restaurants are a multi-trillion-dollar global category with near-zero technology penetration in the actual kitchen. If robotic food assembly ever compounds like she thinks it will, entering pre-IPO at $9 billion is a very different cost basis than buying on day one of trading.

The banks get the best seat in the house. Goldman, Jefferies, and J.P. Morgan placing this round means they've already done the diligence, already know the financials, and already have relationships with the crossover buyers. If the IPO happens in 2027, the underwriting mandate is effectively pre-negotiated. Placement-agent work on a pre-IPO round is, functionally, a paid audition.

NEA, Accel, and GV get mark-ups and a path to liquidity. Venture funds from the 2018–2021 vintage need exits, and a company that has raised roughly $3 billion cannot realistically be acquired — the buyer list at $9 billion-plus is almost empty. An IPO isn't just Lore's ambition; for the existing cap table it's close to the only exit geometry that works.

Sweetgreen, oddly, comes out fine. It sold Spyce for $186.4 million, kept the license to keep deploying Infinite Kitchen in its own stores, and offloaded the R&D burden onto someone else's balance sheet. It gets the automation story without the automation spend. Whether that was strategic brilliance or capitulation depends entirely on whether the robot ends up mattering.

Precedents — the pizza robot that died and the pizza chain that won

The failure case is Zume Pizza, and it's not subtle. Zume raised roughly $445 million, much of it from SoftBank, to cook pizzas with robots inside moving delivery trucks. The pitch was beautiful: bake in transit, arrive at peak freshness, cut labor. It shut down in 2023. The reason was brutally simple — the robotics never beat human unit costs. Every dollar of capex needed to be amortized across a throughput the machines couldn't reliably hit, while a human doing the same task cost less and broke down less. That's the exact question sitting under Wonder's Infinite Kitchen, and no amount of funding answers it.

The adjacent cautionary tale is Just Eat Takeaway's Grubhub trade. Paid $7.3 billion in 2021, sold to Wonder for about $650 million including assumed debt in 2025. That's a roughly 90% destruction of value in four years, in a category that looked structurally unassailable during the pandemic. The lesson isn't "delivery is bad." The lesson is that food-delivery valuations can evaporate faster than almost any other consumer category when growth normalizes and unit economics get audited. Anyone underwriting Wonder at $9 billion should have that chart taped to their monitor — including because Wonder is now the owner of the very asset that produced it.

The success case is more instructive than the obvious one. People reach for Sweetgreen as proof that food robotics can be capitalized and taken public — but the twist undercuts it: Sweetgreen sold the robot business to Wonder rather than scale it. The genuinely durable precedents are Chipotle and Domino's. Chipotle proved that obsessive discipline on throughput and unit economics, not technology, is what compounds in restaurants. Domino's spent a decade rebuilding itself as a technology and logistics platform and became one of the best-performing stocks in the S&P — but note the sequencing: Domino's already had profitable stores before it built the tech layer. It automated a working business. Wonder is building the tech layer and the store base simultaneously, on borrowed money.

And Wonder has a precedent inside its own history. The original model was cook-in-van mobile kitchens — chef-quality meals finished curbside outside your house. It didn't work, and Wonder abandoned it for fixed food halls. That pivot is worth reading two ways. Charitably: this is a management team willing to kill an expensive idea when the numbers say no. Uncharitably: this is a company that has already been wrong once about its core physical thesis, and the current one is far more capital-intensive than the last.

How rivals counter — the asset-light argument writes itself

DoorDash and Uber Eats are the named competitive frame, and their counterargument is one sentence long: Wonder's real estate and robots are capital traps. Both run asset-light marketplaces that scale without signing leases, and both will happily point at a projected $2.7 billion cumulative burn as evidence. Their own automation bets stay outsourced and optional — DoorDash with its Dot delivery robot and drone pilots, Uber with Serve Robotics and Nuro tie-ups. If autonomy works, they rent it. If it doesn't, they walk away without a balance sheet full of it.

DoorDash also isn't standing still on vertical depth. Its acquisitions of Deliveroo and SevenRooms show a company buying international scale and restaurant-side software rather than kitchens. That's a real strategic fork: Wonder is buying supply (the food itself, the building, the robot), DoorDash is buying distribution and tooling. Both are consolidation plays. Only one of them requires you to be right about cooking.

Sweetgreen has the cheapest position on the board. It retains the Infinite Kitchen license, so it can keep opening automated stores and keep telling an automation story to investors, while Wonder carries the R&D cost. If the robot works, Sweetgreen benefits without having paid for the scaling. If it doesn't, Sweetgreen already banked $186.4 million and moved on.

Chipotle is the most credible fast-casual counterweight, and this just got personal. Through Chipotle Cultivate Ventures it has backed kitchen automation projects including Hyphen (automated bowl assembly) and Autocado (avocado prep). Chipotle's version of the argument is that automation should be deployed selectively into an already profitable store model rather than used to justify one. And it now faces a rival that just put its former CFO on the board — which means Wonder has, in the room, someone who knows exactly how Chipotle's unit economics were actually built.

The subtler counter is one nobody will say publicly: restaurants are the hardest consumer category to industrialize, and everyone who has tried has underestimated it. Menu variation, ingredient perishability, local labor markets, health regulation, and the simple fact that customers notice when food gets worse — these are the walls that killed Zume and stalled a dozen others. Lore's answer is a product claim: "We don't have microwaves. We don't reheat. We actually cook to order." That's the right claim to make. It's also the hardest one to scale to 140 locations and beyond.

So what actually changes

If you're a developer or operator — the interesting technical surface here isn't the robot arm, it's the integration stack. Wonder is trying to run one ordering system across owned food halls, a national marketplace (Grubhub), a meal-kit supply chain (Blue Apron), a robotic assembly line (Infinite Kitchen), and starting January 2027 an autonomous drone network (Zipline). Each of those was built by a different company with different assumptions. Getting inventory, order routing, and fulfillment SLAs to agree across all five is a genuinely hard distributed-systems problem, and it's the thing most likely to determine whether the model works. Also worth watching: the AI personalization layer. Lore has described a system that tracks blood biomarkers and body composition to drive autonomous meal ordering, and said "AI knows me better than myself. Never would have said that's what I'd pick to eat — but I love it." That's an ambitious product claim with obvious health-data privacy implications, and it's nowhere near shipped at scale.

If you're an investor — the entire question is whether you believe the 2030 cash-flow-positive date. Fortune's reported projections say Wonder loses about $618 million on an adjusted-EBITDA basis in 2026 and burns nearly $2.7 billion cumulatively through 2029. Even if the IPO happens in early 2027 as Lore intends, that means asking public markets to fund roughly three more years of losses on a physical-footprint business. The bull case is real: ~20% same-service-area growth, Grubhub acquired at a ~90% discount to its last price, and a robot bought for less than a year of burn. The bear case is equally real: the projections are unaudited internal numbers reported by a magazine, there is no public S-1, the "minimal investor protection" claim is unverifiable, and food-delivery multiples have already demonstrated they can fall 90%. Nothing about the IPO is scheduled. Treat early 2027 as an aspiration.

If you're just someone who eats — near term, essentially nothing changes unless you live near one of the roughly 140 East Coast locations. Medium term, two things are worth knowing. First, if the Zipline partnership lands on schedule, drone-delivered restaurant food becomes a normal consumer option in Dallas starting January 2027, expanding across most Texas locations through the year — that's an actual date on an actual press release, not a concept video. Second, and more consequential, the reason a company is spending this much on robotic kitchens is to make chef-quality food cheaper than the labor cost of making it. If it works, the price of decent takeout falls. If it doesn't, this becomes another expensive lesson in why restaurants are hard — and you'll have paid nothing either way.

🥄 Three Things You're Probably Wondering

— So what does this mean for me? Right now, not much unless you're on the East Coast or in Dallas next January. But this is one of the biggest bets anyone has placed on the idea that robots can make restaurant food cheaper than people can. If it works, good takeout gets less expensive. If it fails, it fails on someone else's $2.7 billion.

— Is the IPO actually happening in early 2027? That's Marc Lore's stated intention, told to Fortune — not a filed event. No S-1 has been made public, no date has been set, and "ready to go public" is deliberately different from "going public." The pre-IPO signals are all there (crossover investors, three placement-agent banks, a Chipotle CFO on the board), but signals aren't filings.

— How can a company losing $618 million be worth $9 billion? That loss figure comes from Fortune's reporting on internal projections, so treat it as directional rather than audited. The valuation case rests on assets, not current profit: Grubhub bought for roughly 9% of what Just Eat paid, a working food-assembly robot bought for $186.4 million, and same-service-area sales growing about 20%. Whether that adds to $9 billion is exactly what an IPO would settle.

Sources

Numbers and criteria are as of announcement and may change. Investment calls are yours to make!