Meta, Amazon, Google Now Command 58% of U.S. Ad Spend

Full analysis

Three companies now take 58 cents of every U.S. ad dollar, up from 56 a year ago, and Madison & Wall's Luke Stillman sees the line still climbing. The story an ad-tech operator has to sit with is the other number: the open internet, the whole world of independent publishers, open CTV, digital audio, and digital out-of-home, is forecast to shrink 1.4% in 2026 while total digital spend grows 12.2%. Growing pie, shrinking slice.

What's actually being decided: not "are walled gardens winning" (they are) but whether an independent ad-tech operator should keep betting on open-internet share, or reposition around what Amazon, Google, and Meta structurally can't sell. This is a Type 1 call for anyone building product roadmaps and renewal pitches right now, because the repositioning takes 18 to 24 months and the window is closing. The forcing function is 2026 planning and upfront season, when agency budgets get committed.

The Market Analyst. The independent stack, The Trade Desk, Magnite, PubMatic, trades on an open-internet growth story. A credible sell-side analyst just handed the bears a structural argument instead of a macro one: not "spend is soft this quarter" but "your category shrinks in absolute dollars." That's worse. For a generalist: these companies' share prices assume the open web grows, and someone respected just said it won't. The live question is whether Amazon Ads gets valued as retail media, a hot category, or stays lumped in with "walled garden." The vendors walled gardens can't absorb, independent measurement and identity like LiveRamp and DoubleVerify, are the cleaner long. Both carry execution risk.

The Skeptic. 58% terrifies because the denominator hides. Total spend grows double digits, so a 1.4% open-internet decline is a relative-share story, not an existential one. Retail media, CTV, and digital audio are still adding absolute dollars, just slower than the giants. And the walled-garden AI narrative skips the ugly parts: Meta's measurement keeps drawing trust questions, Google is carrying an antitrust remedy that could force structural change, and Amazon's targeting runs out of road the moment you leave purchase intent. For a generalist: the giants look unstoppable partly because they grade their own homework. Smart CMOs hedge. Priced-right open inventory with proven incrementality still clears.

The Operator. The 1.4% isn't a rounding error to a publisher yield team, it's a demand signal that lands in Q1 and Q2 as planners move scatter budget into Advantage+ and PMax before the upfronts. Fill-rate enthusiasm from buyers drops. Floor-price strategies that held in 2023 get gamed harder because there's simply less standing demand behind them. For a generalist: the automatic ad-buying tools inside the giants are pulling money out of everywhere else, and the people who sell ad space on independent sites feel it first. The second-order hit: smaller SSPs watch their addressable market shrink in dollars, not just percentage, which pulls consolidation forward.

The Customer / End User. Follow the CMO, because Stillman is right that there's "definitely not a tide" moving away. Buyers aren't choosing walls out of loyalty. They're choosing the AI optimization that spends the budget and reports a return without a services layer. The open internet's pitch, cross-platform attribution and contextual reach the walls can't match, only wins if it's provably incremental and low-effort. Today it's usually neither. For a generalist: advertisers go where the machine does the work and shows a receipt. Until independents make "buy outside the walls" as easy and as measurable, share keeps flowing in one direction.

The CFO. Concentration is a cost story for every buyer in the market. When three sellers hold 58% and climbing, they set price, and the AI tools that "optimize" also quietly remove the buyer's ability to audit what a conversion cost. For an independent ad-tech P&L, the math is brutal: your addressable spend pool is forecast to fall in absolute terms, so flat share means declining revenue. Growth has to come from taking share inside a shrinking pool or building a product the walls structurally won't, which is a multi-year R&D bet against companies with far deeper pockets.

The tensions. The Skeptic and the Strategist genuinely part ways on whether 58% compounds or plateaus. The Skeptic says the walls hit real ceilings, measurement trust, antitrust, Amazon's intent limits, and cycles turn. The compounding case says the feedback loops get stronger with scale and there's no natural stop. Second split: the Market Analyst wants to short the SSPs most exposed to long-tail CPM compression, while the Operator notes those same SSPs are consolidation targets, which can rescue a stock the fundamentals would sink. Third, the Customer and the Strategist disagree on the differentiation pitch: cross-platform attribution and contextual precision are real, but the buyer only cares if it's easy and incremental, and right now it's a slide, not a workflow.

What it hinges on. Two beliefs. First, whether the walled-garden data-and-AI advantage keeps widening or hits a real ceiling in the next two years, that decides compound versus plateau. Second, whether independents can make buying outside the walls as automated and as measurable as PMax, because the Customer lens says that, not contextual reach, is what actually moves budget. The council leans toward the concentration continuing near-term, because nothing in the story shows the giants' AI advantage narrowing, and the buyer incentive runs one way. But the "existential" framing is oversold: the open internet loses relative share of a growing market, it doesn't disappear.

Before committing a roadmap, verify the thing the Customer lens exposed: run a real incrementality test of open-web inventory against a walled-garden control with your own buyers, and measure the workflow effort, not just the ROAS. If independents can't close the ease gap, no amount of contextual precision saves the share.

Prediction: In Madison & Wall's next annual U.S. ad-spend estimate covering full-year 2025, the Amazon-Google-Meta combined share will be reported at 58% or higher, confirming the forecast rather than reversing.

Confidence: High. The trend is multi-year, mechanistic, and Stillman already sees it "edging upward."

Why: The share moved from 56% to a projected 58% on the strength of AI campaign tools that get better as they ingest more spend, a loop that strengthens with scale rather than decaying. Stillman explicitly says there's "definitely not a tide" of advertisers leaving, and the buyer incentive, automated spend with a built-in receipt, points the same direction. For the number to come in below 58% you'd need advertisers to reverse a preference no one in the data shows reversing inside a single year, which is the far less likely outcome.

Revisit by 2026-12-31: We're right if Madison & Wall's full-year 2025 figure lands at 58% or above. We're wrong if it comes in below 57%, signaling the concentration stalled.

The relative-share panic is overcooked, but the direction isn't in doubt. The operators who win the next two years won't out-argue the walls on reach. They'll make buying the open web as push-button and as auditable as PMax, or they'll compete for a pool that gets smaller in dollars every year.