Sun Tzu wrote that “the supreme art of war is to subdue the enemy without fighting.” Most analysts covering the grocery wars between Amazon and Walmart miss this entirely. They keep waiting for Amazon to fight Walmart on Walmart’s terms: more stores, more square footage, more checkout lanes. That fight is over before it starts. Walmart has nearly 4,700 stores in the U.S. and a store within 10 miles of roughly 90% of the population, and Amazon cannot out-build that, nor should it try.

Amazon doesn’t need to win the store war. It needs to make the store war irrelevant.

The Fortress Problem

Walmart’s stores are the modern equivalent of the Maginot Line, an enormous, capital-intensive fixed position that assumes the next war will look like the last one. Fixed fortifications only work if the enemy attacks where you’re strongest, and Amazon doesn’t have to attack Walmart’s stores at all. It can go around them, the same way Germany’s army went around the Maginot Line through the Ardennes in 1940.

Walmart is playing a density game and mistaking store count for the win condition. Store count only matters if proximity to the customer is the only variable, and it isn’t: speed, automation, and the cost to fulfill are what actually decide this war, and none of those require owning a building the size of a Walmart Supercenter.

The Size of the Board

Amazon and Walmart are now, for the first time in over a decade, nearly the same size. Amazon closed 2025 with $716.9 billion in total revenue against Walmart’s $713.2 billion for fiscal 2026, though the resemblance stops there: Amazon holds a commanding lead online, roughly 38 to 40% of the $1.2 trillion U.S. e-commerce market against Walmart’s roughly 9%, and in groceries, the picture reverses entirely.

Walmart holds roughly 20% of all U.S. grocery spend, the largest share in the category, against Amazon and Whole Foods combined at roughly 8%. PYMNTS Intelligence published an estimate this month that at Amazon’s historical pace of share gain, closing that grocery gap would take roughly 100 years.

That estimate deserves to be taken seriously, and then set aside, because it makes the same mistake analysts have made about Amazon before. Amazon wasn’t the top U.S. apparel and footwear retailer in 2012, and skeptics dismissed the idea that a logistics company could win a category built on fit, brand relationships, and physical stores. Amazon took the top spot from Walmart in 2018, and by 2024 its apparel and footwear sales exceeded $67 billion, more than double Walmart’s, on a path past $72 billion in 2025. That happened in roughly six years, not a hundred, and not by Amazon playing grocery’s version of the game one point at a time. It happened by Amazon changing what game was being played.

A hundred-year timeline is the correct answer if Amazon keeps competing in groceries the way it has for the past seven years: general-purpose fulfillment, manual picking, no dedicated cold chain, no automation partner, no purpose-built delivery network. That’s not the Amazon this article is describing. It’s the Amazon that existed before Project Mercury, before the AutoStore agreement, before the employee grocery discount. The right question isn’t how long it takes Amazon to win at Walmart’s game. It’s what happens when Amazon stops playing it.

The Counter-Strategy: Win the Density Game Without Playing It

This is where GRIDD, the customer-density framework I developed, applies directly. Walmart wins on store proximity, a location within 10 miles of most of the country, but proximity measured in “distance to a large building” is the wrong metric in an era of automated, small-footprint fulfillment. The right metric is distance to inventory that can move.

Amazon doesn’t need 4,700 stores. It needs roughly 1,000 delivery stations, each within 3 to 5 miles, not 10, of 90% of the U.S. population. Density beats footprint whenever speed is what matters: shorter last-mile legs mean each vehicle can run more delivery cycles per day, so Amazon gains capacity without adding square footage.

On September 16, 2026, PYMNTS reported that Amazon is doing close to exactly that. Internal planning documents describe an initiative called Project Mercury, aimed at growing Amazon’s same-day delivery hubs from roughly 85 today to more than 1,000 by 2031, backed by $6.8 billion in planned spending this year and next. PYMNTS’ own follow-up analysis framed it plainly: Amazon “needs more than faster warehouses to close Walmart’s delivery advantage. It may need to build and manufacture its own version of a neighborhood store network.”

None of that is a new idea to me. I first raised the concept of Amazon building a dense network of smaller, forward-deployed facilities to counter Walmart’s store footprint on podcasts as far back as 2000, and I’ve been writing about Amazon’s broader grocery strategy in print since 2013. I put the specific 1,000-facility figure in writing beginning in 2022, well before Project Mercury’s own reporting confirmed that same number this month.

Amazon’s reported coverage standard, though, 10 miles of 80% of Prime members, is looser than the 3-to-5-mile standard I’ve argued Amazon would need to truly out-density Walmart. Walmart already sits within 10 miles of 90% of the population, so a reported Amazon target of 10 miles and 80% coverage doesn’t out-density Walmart, it merely approaches it. That’s too large a gap for a company as data-driven as Amazon to have simply miscalculated.

“Appear weak when you are strong, and strong when you are weak,” Sun Tzu wrote, and I believe Amazon is doing exactly that. Publishing an unambitious 10-mile, 80% target gives Walmart and the press a benchmark that looks matchable rather than threatening, inviting the kind of complacency a slower-moving competitor needs to keep losing gradually instead of urgently. The real operating target inside Amazon, I believe, is the tighter 3-to-5-mile standard this article has argued for from the start, the one that actually surpasses Walmart’s network rather than merely equaling it.

Announcing 10 miles and 80% while building toward 3 to 5 miles and full density coverage is precisely the deception Sun Tzu describes: convincing the competitor the threat is smaller and farther off than it actually is, until the gap is too wide to close.

Inside those delivery stations, and inside strategically located Supercenter-format sites, Amazon should install AutoStore systems to industrialize fulfillment at a scale Walmart’s store-based, manually picked model cannot match economically. Grocery retailers typically lose between $7 and $25 on every online order they manually pick and deliver, and automated fulfillment is what closes that gap. This isn’t speculative: Amazon and AutoStore formalized a global strategic supply agreement in August 2026, giving Amazon the procurement framework to do exactly this at scale.

The Vehicle Layer: Turn the Fleet Into the Network

Buildings are only half the equation. The other half is what happens after the order leaves the building, and this is where Amazon has an asset Walmart doesn’t: a fleet of Rivian electric delivery vans that can become mobile fulfillment nodes rather than passive cargo boxes. Equip those vans with automation technology like the kind Cargo Robotics has built, robotic retrieval, intelligent sorting, and AI-driven route and inventory optimization inside the van itself, and Amazon can run thousands of vans, a meaningful share of them autonomous, along fixed neighborhood routes offering near-immediate fulfillment of everyday essentials. Inventory stops waiting for demand and starts moving toward it continuously, all day, every day.

Layer a meal program on top, one priced below what it costs a customer to shop for groceries and cook a comparable meal at home, and Amazon isn’t just competing on grocery price. It’s competing on the customer’s time, which is a category Walmart’s store model cannot enter.

Supplement the ground network with drones for the last, thinnest mile, and the picture is complete: stores where they make sense, delivery stations where speed matters most, vans that double as moving fulfillment centers, and drones for the edge cases. Four layers, one system, a fraction of Walmart’s fixed capital cost.

There’s a fifth layer to this that most coverage of Amazon’s smart glasses program misses entirely, because it takes Amazon’s own explanation of the technology at face value. Officially, Amazon says the glasses, unveiled publicly in October 2025 and already piloted with hundreds of drivers across more than a dozen delivery service partners, exist to help drivers navigate routes hands-free, scan packages, and capture proof of delivery. That’s the sanctioned story, and it’s not false. It’s also not the whole story.

I reported in March 2026 that the real value of that data isn’t the navigation feature Amazon markets, it’s the training set the glasses quietly generate. Every delivery captures how a driver approaches a porch, handles a gate, navigates a stairwell, positions a package, and reads an address in poor lighting, the full physical choreography of last-mile delivery that no warehouse robot has ever had to learn. Amazon is using that data to train the models that will eventually run humanoid robots riding along in, and eventually working out of, the same Rivian vans this article has already described.

I think this is the most underappreciated part of Amazon’s entire automation strategy, and here’s why. Every other company racing to build humanoid robots, Figure, Tesla, Boston Dynamics, is training them primarily on simulation data and controlled environments, a warehouse floor, a lab, a curated test course. That’s a solvable but comparatively narrow problem. Amazon is quietly assembling something none of its competitors have: millions of real-world examples of a human navigating the messiest, least standardized environment a robot could ever be asked to operate in, an actual American front porch, in every weather condition, every lighting condition, every kind of gate, dog, stairwell, and mailbox in the country. That’s not a training shortcut. It’s a data moat, and it’s one Walmart, Kroger, and every other Amazon competitor has no equivalent way to build, because none of them have hundreds of thousands of delivery associates wearing sensors on their own last mile every single day.

I want to be precise about what’s confirmed and what isn’t. Amazon has not publicly stated that the smart glasses are capturing data to train humanoid robots, and it may never say so directly, for the same competitive-secrecy reasons this article has already described using Sun Tzu’s framework. What’s confirmed is that Amazon is building humanoid robots, that it’s investing heavily in software to let those robots operate in delivery contexts, and that the glasses are capturing exactly the kind of physical, real-world movement data that kind of program would need. I believe the connection between those two facts is not a coincidence, and it’s the single clearest signal that Amazon’s ambitions here extend well past vans, drones, and delivery stations, to eventually replacing a meaningful share of the human physical labor in last-mile delivery altogether.

The Meal Disruption: Feeding America Without a Grocery Store

Every layer of this plan so far, delivery stations, AutoStore, Rivian vans, drones, still assumes the customer is buying groceries and cooking them. That assumption is the actual target: the biggest lever Amazon has isn’t a better way to sell groceries, it’s a way to make grocery shopping unnecessary for a meaningful share of a household’s meals.

The math is simple and it’s been true for years: cooking a meal at home costs a customer money, time, and planning, the price of the ingredients, the trip or the delivery fee, the time to prepare it, and the food that gets thrown away when it isn’t used. If Amazon can deliver a fully cooked, ready-to-eat meal at a price point at or below what it costs a customer to buy the ingredients and cook the same meal themselves, the decision stops being about price. It becomes about convenience with no downside, and convenience wins.

This is a different kind of competition than anything Walmart, Kroger, or Whole Foods is built to fight. Walmart competes on grocery price; Kroger competes on grocery price and loyalty. Both are optimized to sell raw ingredients as cheaply as possible and let the customer do the rest, and neither is built to compete on eliminating the need to shop and cook in the first place. That’s a different business, food-as-a-service, not grocery retail, and it’s a business Amazon is better positioned to build than either of them, because Amazon already owns the logistics network to deliver meals hot, at scale, on the same infrastructure being built for Project Mercury.

The same delivery stations equipped with AutoStore handle center-store and packaged goods. The same Rivian vans, automated with Cargo Robotics-style technology, that deliver everyday essentials can carry commissary-prepared, ready-to-heat or ready-to-eat meals on the same routes, at the same stops, on the same 30-minute cycle. Amazon doesn’t need to build restaurants or commissaries from scratch; it can contract or acquire the food-tech and ghost-kitchen capacity that already exists, the same category of company I’ve advised and written about for years, and route the output through the delivery-station and van network it’s already building for groceries.

The disruption isn’t incremental. Once a customer can get dinner delivered hot, ready to eat, for less than the cost of buying and cooking the same ingredients, the grocery basket that funds a large share of Walmart’s grocery revenue shrinks, not because Amazon took grocery market share store by store, but because the customer needed fewer groceries in the first place. That’s a fundamentally different kind of attack than a price war, and it’s one Walmart’s model has no clean answer for, because discounting ingredients doesn’t compete with not needing ingredients at all.

Whole Foods sits partly outside this pressure. As a specialty retailer built around an intentional shopping experience, organic, high-touch, destination-driven, it draws a customer who is buying the experience of shopping as much as the food itself, and that customer is less price-sensitive to a cooked-meal alternative than the mainstream Walmart or Kroger grocery shopper is. That means Amazon’s own Whole Foods asset is comparatively insulated from a strategy aimed squarely at everyone else’s grocery basket.

The Employer Channel: Three Moves That Change the Rules

I believe this is a massive opportunity for Amazon. The grocery discount Amazon rolled out to its own employees in September 2026, 10% online, 20% in-store at Whole Foods and Daily Shop, isn’t the end state. It’s the first of three escalating moves that I recommend Amazon make, each one aimed less at winning a price comparison and more at making price comparison irrelevant.

Move one: Corporate Save. Amazon extends its employee discount structure to outside companies. A corporation, Tesla, Microsoft, Google, Caterpillar, John Deere, etc., signs an agreement, and its employees get Amazon’s grocery discount as a workplace benefit, the same category as a subsidized gym membership or commuter benefit. Amazon gets guaranteed grocery volume and a foothold inside companies it has no retail relationship with today, and the employer gets a zero-cost benefit, funded by AWS and advertising margin rather than grocery margin.

Move two: every Prime member. If Corporate Save proves the mechanism, the next move drops the employer intermediary entirely: every Prime member gets the same grocery discount Amazon’s own employees get. Prime already has more than 200 million U.S. members paying for shipping and streaming benefits, so adding a grocery discount doesn’t require a new relationship, it deepens an existing one. It turns Prime from a shipping-speed subscription into a cost-of-living subscription, something Walmart has no membership program at comparable scale to match.

Move three: everyone. The final move is a flat 20% discount available to any American buying groceries through Amazon or Whole Foods, Prime member or not. At that point Amazon isn’t competing on convenience or selection, it’s competing on price, permanently, funded by a profit engine no pure-play grocer has access to.

This is why the PYMNTS extrapolation cited earlier doesn’t hold: it assumes Amazon keeps taking grocery share the way it’s taken it so far. PYMNTS is wrong.

A flat, universal discount funded by profits from unrelated markets is unlikely to meet the legal test for predatory pricing, since Amazon holds nowhere near dominant grocery market share, which undercuts any recoupment theory. The more realistic exposure is political and reputational: the same cross-market subsidization argument already central to the FTC’s broader case against Amazon, and a ready-made talking point for competitors and lawmakers, even if it doesn’t rise to an antitrust violation on its own. Stated another way, many people will complain about what Amazon is doing but they won’t be able to do anything to stop Amazon.

Combine the meal-delivery layer and the employer channel with the density and automation layers already in place, and the strategy stops being about winning grocery market share. It becomes about shrinking the size of the grocery market Walmart and Kroger are fighting over in the first place. This is why I came up with these strategies. They allow Amazon to change the game vs forcing Amazon to play on the same field.

The Math

Walmart’s U.S. segment generates 59% of its revenue from groceries, nearly $285.5 billion in fiscal 2026, making grocery the single largest driver of the company’s core American business, a concentration that is Walmart’s strength and its exposure at the same time: a large, efficient, fixed target, one Kroger and an increasingly aggressive Costco are also competing for.

Amazon does not need to dislodge Walmart from the top spot overnight. It needs to take 10% or more of that grocery revenue through superior speed and lower fulfillment cost, and the compounding math of density, automation, mobile fulfillment, meal disruption, and the employer channel gets it there without Amazon ever opening a store network anywhere close to Walmart’s size.

Executing this at scale carries real risk beyond the regulatory exposure already noted in the employer discount strategy: unresolved FAA rules on beyond-visual-line-of-sight drone delivery, labor disruption at a scale that invites policy and public backlash, and the sheer difficulty of building 1,000 automated facilities, a mobile van fleet, and a meal-delivery network simultaneously without the kind of timeline slippage that has hit large-scale retail automation projects before. None of that changes the direction of the strategy. It changes how carefully it must be sequenced.

That’s the entire strategy in Sun Tzu’s terms: subdue the enemy without fighting. Don’t attack the fortress. Make the fortress irrelevant.

A Strategy Documented in Real Time

None of the above is a new position for me. The record shows a consistent argument, built in public, over the better part of a decade, with roots going back further still:

• 2013 I wrote the research paper recommending Amazon acquire Whole Foods (posted publicly in 2016, confirmed by the actual acquisition in June 2017), and I invented GRIDD, the density framework underlying this entire strategy.
• 2018 In Forbes, I laid out the underlying thesis that Amazon would come to lead the grocery industry. That same year, I laid out the meal-delivery disruption argument directly, arguing that hot, cooked meals delivered at a price below the cost of shopping and cooking would change consumer grocery behavior, and stating I’d been developing the underlying concept since 2010.
• 2019 In The Observer, I first framed Walmart’s store network as a modern Maginot Line.
• 2020 I recommended Amazon partner with AutoStore for automated fulfillment and separately recommended Amazon acquire Alert Innovation for micro-fulfillment technology, before Walmart acquired it instead.
• 2022 I restated the Maginot Line argument on the record with Retail Technology Innovation Hub, and put the specific figure of “a thousand or more” micro-fulfillment centers in writing for the first time.
• 2023 I put a specific number on it: 500 to 1,000 micro-fulfillment centers. I also published the case for automating Rivian delivery vans, the article that led the founders of Cargo Robotics to build the system now on the road, and restated the meal-delivery thesis.
• 2025 to 2026 I restated the Maginot Line thesis as Amazon expanded same-day grocery delivery, and reported on the Amazon-AutoStore relationship intensifying months before it became public. In March 2026, I went on record stating Ocado’s pivot to store-based automation positioned it as the most valuable technology play in grocery, retail, and logistics, and broke the story that Amazon’s DSP driver smart glasses are being used to capture last-mile delivery data to train humanoid robots.
• August 2026 AutoStore and Amazon confirmed their global strategic supply agreement publicly, and I was quoted on the record by Supply Chain Dive calling it a wake-up call for Walmart, Target, and Kroger.
• September 2026 PYMNTS reported on Project Mercury, Amazon’s plan for more than 1,000 same-day delivery hubs, the same figure I put in writing in 2022. Days later, Amazon extended a grocery discount to its entire U.S. workforce, the first proof point of the employer-subsidy mechanism I outline above.

Each piece of this plan has moved from prediction to confirmed fact roughly in the order I laid it out. Project Mercury, and the widening gap between what Amazon says it’s building and what I believe it’s building toward, is the clearest evidence yet that this strategy is no longer theoretical.