Uber Advertising Surpasses $2 Billion Annualized Revenue Run Rate

Uber's ad business hit a $2 billion annualized run rate in Q4 2025, growing more than 50% year over year. The milestone matters less than what Uber's VP of advertising Balaji Krishnamurthy said alongside it: the assumption that delivery app ad revenue caps out at roughly 2% of gross bookings is wrong. Every retail media analyst uses that benchmark to model Instacart and DoorDash too. If the ceiling was arbitrary, estimate revisions flow across the whole delivery category.

The $2B figure itself is unaudited. Uber has never broken out advertising as a GAAP segment, so nobody outside the building can separate genuine third-party brand spend from subsidized restaurant placements that would otherwise be discounts. The 50% growth rate carries the same caveat. Part of that revenue is restaurants spending co-op dollars because there is nowhere else to put them inside the Uber ecosystem.

The more consequential question is what Uber does with its first-party data. Uber has something grocery and e-commerce retail media cannot replicate: where you physically go, what you are about to order, and a captive screen while you wait. If Uber keeps that identity graph walled, it becomes a must-buy direct placement, and buyers have no programmatic path in. If it opens the data through clean-room arrangements with companies like LiveRamp or InfoSum, it resets how mobile and out-of-home attribution work across the market. The incentive points toward licensing, because the data is worth more sold many times than once, but that decision is genuinely open.

The personnel move that makes this concrete for the open web: Megan Ramm, who ran Uber's ad sales operation, left in August 2026 to become chief revenue officer at PubMatic (an open-web supply-side platform). PubMatic hired her on the thesis that her advertiser relationships could be routed into programmatic supply, which would be direct pressure on Magnite and Index Exchange for premium mobile inventory. That thesis has a problem. Uber's growth ran on closed-loop first-party purchase data that made the ads work. PubMatic routes other people's inventory and cannot replicate that signal. The advertisers Ramm knew bought Uber's data, and whether those relationships transfer to an open-web SSP without it is the open question.

Who wins and who loses

SSPs and mid-market DSPs without a direct Uber private marketplace deal are already behind. The 2% ceiling breaking means budgets that category norms used to constrain are now in play, and the plumbing to capture Uber's inventory is not standardized. For SSPs, the right move is to build the direct PMP relationship regardless of whether Uber eventually opens its data, because the fully walled outcome is the one where programmatic buyers lose access entirely.

Large brand advertisers are the ones with leverage they haven't used yet. Uber's pitch scales to national brand budgets only if buyers can verify the measurement through something other than Uber's own reporting. Until that happens, Uber is another standalone buy with its own login and self-graded attribution, which is the opposite of what large advertisers want.

Our call: PubMatic's full-year 2026 revenue, reported in February 2027, will grow year over year at well below the 50% rate Uber's ad business is claiming. A single CRO hire does not transfer a captive-data demand machine to an open-web SSP. PubMatic's base business grows at open-web rates in the low double digits. The only scenario where PubMatic approaches Uber-pace growth is if Ramm's relationships were personal rather than data-driven, which is the less likely read. We'll revisit when PubMatic reports in February 2027.

Uber's ad business hit a $2 billion annualized run rate as of its Q4 2025 results in February 2026, growing more than 50% year over year, and the company says the old assumption that restaurant delivery ads would cap out at 2% of delivery gross bookings no longer holds. For ad-tech operators, the story is less about Uber's milestone and more about a fourth closed-loop retail media network arriving at scale, and about Megan Ramm, who built Uber's ad sales machine, leaving in August 2026 to run revenue at PubMatic.

What's actually being decided: not whether Uber is a real media business (it is, at $2B), but whether its first-party purchase and movement data flows out to the open programmatic ecosystem or stays walled. That answer reshapes who wins mobile and out-of-home attribution. It's easy for operators to adjust to either outcome, so this is about positioning, not a bet-the-company call. What sets the clock: 2027 budget planning and PubMatic's next few earnings calls, where Ramm either produces Uber-style demand or doesn't.

The Market Analyst The headline that moves models isn't Uber's $2B. It's Balaji Krishnamurthy saying the 2% ceiling broke. Every retail media analyst benchmarks delivery ad revenue against gross bookings, and Instacart and DoorDash get measured on the same math. If Uber proves the ceiling was arbitrary, estimate revisions flow to the whole category. In plain terms: analysts assumed delivery apps could only sell a thin slice of ads relative to order volume, and Uber just said that cap was made up. PubMatic hiring Ramm signals that open-web SSPs believe Uber's advertiser relationships can be routed into programmatic supply, which is direct pressure on Magnite and Index Exchange for premium mobile inventory.

The Skeptic Two billion annualized is a press release, not an audited line. Uber has never reported advertising as a GAAP segment, so nobody outside the building can split genuine third-party brand spend from subsidized restaurant placements that would otherwise be discounts. The 50% growth is unaudited too. And the inventory is narrow by design: you're advertising to someone mid-transaction, which is powerful for a restaurant and useless for a car brand. "Enterprise advertising is playing catch up" is the polite version of "we haven't cracked big brand budgets yet." Plain version: this is a high-margin feature bolted onto a logistics company, and part of that $2B is restaurants buying placement because there's nowhere else to spend their Uber co-op dollars.

The Operator If you run an SSP or a mid-market DSP, Uber just moved from nice-to-have to must-integrate supply. Thirty-plus countries means currency handling, consent frameworks, and bid-stream cleanup all get harder at once. The 2% ceiling breaking is the signal that matters on Tuesday morning: budgets that category norms used to cap are now in play, and any SSP without a clean Uber private marketplace deal is leaving Q3 and Q4 money on the table. Plain version: a big new pool of ad money just opened, and the plumbing to capture it isn't standardized yet. Ramm's jump proves the org is mature enough to export senior talent, which means the relationships are real enough to poach.

The Strategist Uber is the third closed-loop retail media network outside Amazon and Walmart, with a data moat neither grocery nor e-commerce can copy: where you physically go, what you're about to buy, and a captive screen while you wait. The real fork is licensing. Wall the identity graph and Uber becomes a must-buy direct, like a walled garden. Open it through clean-room deals with LiveRamp or InfoSum and Uber resets how mobile and out-of-home attribution works for everyone. Plain version: Uber can either make advertisers come to it, or rent out its data to the whole market, and those are very different businesses. The incentive points toward licensing, because the data is worth more sold many times than once.

The Customer / End User (the advertiser) A restaurant or CPG brand loves this: reach someone with their wallet open, measure the order. But the buyer running a national brand campaign asks a harder question. Can I plan Uber alongside Snap and Pinterest in one buy, or is it another standalone login with its own measurement Uber grades itself on? Right now it's the latter. Plain version: advertisers want fewer walled gardens, not another one, and Uber's pitch only scales to big brand budgets if the data leaves the building in a form buyers can verify.

Where they disagree

The Skeptic and the Strategist split on what $2B even is. The Strategist sees a data moat compounding toward a licensing business; the Skeptic sees subsidized restaurant co-op dollars with nowhere else to go, running through a business that calls itself a media company because the margin is better. Both can be true at $2B and diverge hard at $5B.

The deeper tension is wall versus open. The Operator and the Market Analyst assume Uber supply flows into programmatic, which is why PubMatic paid up for Ramm. The Strategist thinks the incentive runs the other way: data sold through clean rooms is worth more than inventory sold through an SSP. If Uber walls it, PubMatic bought relationships to a garden it can't route around.

What it hinges on

Three things. One, how much of that $2B is third-party brand spend versus subsidized placement, which nobody outside Uber can see. Two, whether Uber lets its purchase and movement data out to external DSPs and SSPs, or keeps buyers coming direct. Three, whether Ramm's advertiser relationships actually transfer to PubMatic, or whether they were relationships with Uber's data, not with Ramm.

The council leans skeptical on the headline and bullish on the category read. The 2% ceiling breaking is real and moves money across delivery retail media. The $2B is real but unverifiable, and the open-versus-walled question is genuinely open.

What to de-risk before acting: for SSPs, assume Uber stays walled and build the direct PMP relationship anyway, because the walled outcome is the one where you lose access entirely. For advertisers, don't underwrite Uber as cross-media currency until the measurement is graded by someone other than Uber.

Prediction: PubMatic's full-year 2026 revenue, reported in February 2027, will grow year over year by less than the roughly 50% rate Uber's ad business is claiming, because a single CRO hire does not transfer a captive-data demand machine to an open-web SSP.

Confidence: Medium. Ramm's relationships may be with Uber's data, not portable to PubMatic.

Why: PubMatic hired Megan Ramm in August 2026 explicitly to channel Uber-style advertiser relationships into programmatic supply, and the market take frames her as the demand engine. But Uber's 50%-plus growth ran on closed-loop first-party data that made the ads work. PubMatic routes other people's inventory and cannot replicate that captive purchase signal, so the advertisers Ramm knew bought Uber, not Ramm. A CRO changes a sales motion in quarters, not weeks, and PubMatic's base business grows at open-web rates in the low double digits, nowhere near 50%. The opposite outcome, PubMatic suddenly inheriting Uber-pace growth, would require the demand to have been personal rather than data-driven, which is the less likely read.

Revisit by 2027-02-28: We're right if PubMatic's reported full-year 2026 revenue growth comes in below 50%. We're wrong if it hits 50% or higher.

The more interesting question will be whether Uber itself ever shows up as PubMatic supply in the bid stream, or whether Uber keeps its data walled and sells direct. If Uber stays walled, PubMatic bought a Rolodex to a garden it can't enter.