Amazon Put Uber Eats Inside Prime. The Unpriced Variable Is the Delivery Subsidy: Amazon's $28.1B Quarter Absorbs It, Uber's $1.06B May Not

On Nov. 5, 2026, Amazon named Uber Eats the official US restaurant-delivery partner for Prime members and let its Grubhub agreement lapse the same day. Prime members now order dinner on Amazon.com or in the Amazon Shopping app, pay no delivery fee on orders of $25 or more, and need no Uber account — they check out with Amazon credentials. The launch sweeteners are dated and narrow: $20 off each of the first two qualifying orders through Nov. 18, and 5% back for Prime Visa cardholders through Dec. 5, 2026.

This is not a routine supplier swap. Amazon is unwinding an equity relationship while buying a distribution relationship: its performance-based Grubhub deal had lifted its stake in Grubhub's then-owner Just Eat Takeaway to as much as 18%, with additional warrants. Just Eat Takeaway has since been taken private, Grubhub was sold to Wonder Group for $650 million, and the Prime channel itself closed on Nov. 5. Members who already activated Grubhub+ keep access until Dec. 31, 2027 — the window a departing partner usually gets to migrate a base Amazon helped build.

Amazon earned $28.1 billion of operating income in Q3 2026, roughly 26 times Uber's entire delivery-segment operating income of $1.057 billion and 26 times DoorDash's $1.060 billion of adjusted EBITDA. A delivery credit priced at a few dollars a trip is a rounding error against Amazon's profit and a real cost inside either platform's. Amazon can also absorb it quietly: sales and marketing expense was $11.6 billion in Q3 2026, essentially flat against $11.7 billion a year earlier, so a funded credit would appear as a step-up in a line that currently shows none.

DoorDash has the least room to absorb it. Its adjusted EBITDA was 3.1% of Marketplace GOV in Q3 2026, up from 3.0% a year earlier and 2.8% in Q2 2026 — a thin slice earned on $33.9 billion of gross order value. For scale, a $3 credit on a $100 basket is 3.0% of order value, essentially the entire margin earned on the book. DoorDash has also guided Q4 adjusted EBITDA to $900 million–$1.1 billion, at or below Q3's $1.06 billion, while Marketplace GOV rises to $35.4–$36.4 billion. It is guiding margin thinner while a rival introduces free delivery.

Delivery is the part of Uber that is working. Q3 2026 delivery revenue rose 23% to $5.49 billion while delivery segment operating income rose 30% to $1.057 billion — profit growing faster than revenue, a margin gain of roughly a point. Delivery and Freight outperformed management's expectations, even as Mobility growth ran 1 to 2 percentage points below them on fuel prices and driver availability. The Prime bundle lands directly on the fastest-growing, margin-expanding segment in the company.

The timing is the tell. Uber's $14.8 billion takeover of Delivery Hero at €41.50 a share entered its acceptance period on Aug. 27, 2026 and was scheduled to close at 24:00 CET on Nov. 5 — the same day the Prime channel switched hands, subject to a minimum acceptance threshold of 50% plus one share. Uber is buying international scale on the day its US distribution becomes someone else's funnel, which sharpens rather than settles the question of who the platform is built to serve.

DoorDash cannot answer a subsidy with a subsidy. It has no megafunnel, no equity tie to Amazon, and no partner willing to waive a delivery fee on its behalf. What it can do is change the cost of the last mile, and it is doing both of the available things at once: cutting cost per delivery with its own aircraft, and buying a larger basket with Instacart.

A $0 delivery fee does not make the last mile free; it only changes who pays for it. Tracing the chain shows why the funding question is not academic: the courier payout, the aircraft, the merchant commission and the consumer's order value all sit underneath the same $25 threshold.

Amazon's defense of its storefront has been legal and technical. It won a temporary injunction in March 2026 barring Perplexity's Comet agent from shopping on Amazon, and in September 2026 it cut off Meta's Muse assistant from buying on the site. When outside agents sit at the checkout, the retailer loses the customer relationship. Owning the delivery transaction is the cheapest way to keep a category of commerce out of an agent's reach — and food is the hardest category for an agent to buy, because it is local, immediate and physical. A Prime member asking Alexa for dinner is still ordering through Amazon's own surface.

Fact: Amazon is buying restaurant delivery for its Prime members at a price the customer never sees, and the terms of who pays were not disclosed. Inference: the most probable funding route is Amazon, because a delivery credit is trivial against $28.1 billion of quarterly operating income and expensive against $1.06 billion of delivery-segment operating income. If that is right, the deal is worth more to Amazon than to Uber — Amazon gets daily habit and a category that shopping agents cannot take, while Uber gets volume without the customer. If it is wrong, Uber's delivery segment absorbs a cost its 19% operating margin can carry for a while, and DoorDash's 3.1% margin on gross order value cannot. The tell is a single line in Amazon's Q4 2026 filing.