Key findings
- US ad prices on Meta are up 23%. Meta's Q3 2026 deck shows US and Canada price per ad up 23% year over year while impressions grew 8%. Worldwide it rose 14%.
- Independent data agrees. Bïrch's US tracker shows Facebook CPM up 39% in September, and Growthr's own client data was up 35%.
- Google moved far less. Google Search cost per click rose 3% in Q2 2026 while paid clicks grew 13%.
- Judge on cost per purchase. With click-through unchanged, a 23% higher CPM turns last year's 3.0x ROAS into about 2.4x. Better creative, through higher click-through, can cancel most of it.
The short answer
Meta CPMs are rising because more advertiser money is chasing US ad space that is growing more slowly. Meta's Q3 2026 earnings deck shows the average price per ad in the US and Canada up 23% from a year earlier, while ad impressions there grew 8%. Independent trackers that isolate US campaigns show the same climb.
Meta gives two reasons in its Q3 prepared remarks: "ad performance gains and improvements in macro conditions relative to Q3 of last year." Meta's ranking models are converting a little better, advertisers bid more for impressions that convert, and fewer of them are holding back the way they did during the tariff months of 2025. The other causes people have suggested, the March attribution change and the midterm elections, either cannot raise prices directly or are too small to account for a move this size. Each is tested below.
If nothing else in a campaign changes, a 23% higher CPM is a 23% higher cost per purchase and a ROAS that is 19% lower. A campaign that ran at 3.0x last October runs at about 2.4x on the same creative this October.
What the data shows
Each dataset samples a different set of advertisers. Meta's numbers cover every advertiser but report price per ad rather than CPM; third-party trackers report CPM for the accounts that use them. We used both kinds.
| Source | What it measures | Period | Change vs a year earlier |
|---|---|---|---|
| Meta earnings deck | Average price per ad, US & Canada, every advertiser | Jul to Sep 2026 | +23% |
| Meta earnings deck | Average price per ad, worldwide | Jul to Sep 2026 | +14% |
| Bïrch cost tracker | Facebook CPM, US-targeted campaigns | September 2026 | +39% |
| Bïrch cost tracker | Instagram CPM, US-targeted campaigns | September 2026 | +56% |
| Triple Whale benchmarks | Median Meta CPM, 40,000+ ecommerce brands | Aug 2025 to Jul 2026 | +13% |
| Northbeam | Average weekly Meta CPM, its accounts | Late Sep 2026 | +25% |
Meta's price per ad is ad revenue divided by impressions, apportioned by where the user is. Bïrch figures pulled from its public tracker on October 21, 2026; October was month-to-date.
How we checked Sources, dates, and limits
- Meta
- Price per ad and ad impressions by region from Meta's Q2 and Q3 2026 earnings decks, and the reasons from the Q3 prepared remarks, all published by October 21, 2026.
- Bïrch
- Monthly Facebook and Instagram CPM for US-targeted campaigns, read from its public tracker on October 21, 2026. October is month-to-date.
- Triple Whale
- Medians across 40,000+ ecommerce brands, August 2025 to July 2026 against the year before, published August 17, 2026.
- Paid clicks and cost per click from Alphabet's 10-Q filings, Q1 2025 to Q2 2026. Alphabet had not reported Q3. Tinuiti's Q3 figure comes from MediaPost's summary because the report is gated.
- Our client data
- Ads Manager totals for each date range, read October 21, 2026. Percent changes only; no spend figures or client identities.
- Limits
- Meta's price per ad is not a CPM, Google's figures are worldwide and per click, and our own client data is a check on the pattern, not a sample.
Every source points the same way, and the spread between them comes from what each one samples. Meta's worldwide 14% is held down by Asia-Pacific, where prices rose 1% while impressions grew 14%. Triple Whale's 13% is a trailing twelve months, so it averages the flat autumn of 2025 with the climb of 2026. The numbers that isolate US advertisers in recent months, Meta's US and Canada figure and Bïrch's tracker, show the steepest rise: 23% and 39%.
The timing has shifted too. In Bïrch's 2025 series, CPM was flat from August through October and rose only 2% from September to November. In 2026 the climb started in late spring and steepened in September, two months before the usual Black Friday run-up. Bïrch also shows no jump right after March, which matters for the attribution theory below. US Facebook CPM was 9% above 2025 in March, 3% in April, then the gap widened every month from May to September.
CPMs in our own client data
We checked the same periods in Growthr's own Meta client data, US fintech, managed day to day by our team.
| Period, 2026 vs 2025 | CPM | Cost per link click |
|---|---|---|
| January to March | +39% | −17% |
| April to June | +18% | −24% |
| July to September | +17% | −6% |
| September | +35% | +2% |
| October 1 to 20 | +21% | −17% |
Totals from Ads Manager, read October 21, 2026. Link clicks are counted the same way in both years; the March attribution change affected which conversions are credited, not the click count.
On CPM, our data follows the market. Third-quarter CPM rose 17%, between Meta's worldwide and US figures, and September rose 35%, close to Bïrch's 39%. The first quarter rose more than the trackers show.
Cost per click went the other way. In four of the five periods we paid less per link click than a year earlier while paying more per thousand impressions, because more of the people who saw the ads clicked. From July to September, a 17% higher CPM and a 6% lower cost per click mean click-through was up about a quarter.
That is where we think the answer is. CPM is set by the auction. What an advertiser controls is whether an impression gets clicked and whether the click converts, and click-through is mostly creative: the hook, the format, how many variants are in rotation, and how often they are replaced. At Growthr we treat creative volume and testing as the main response to rising CPMs, ahead of bid and budget changes.
What is pushing the price up
Price per ad is set in an auction, so it rises when demand for impressions grows faster than the supply of them. In the US that gap has widened for three quarters.
Multiply the two and Meta's US and Canada revenue grew about a third in a year: $28.9 billion in Q3 2026 against $21.8 billion in Q3 2025, per the same deck. Advertisers bought 8% more ads and paid 23% more for each one.
Better conversion rates, so higher bids
This is Meta's own explanation and the best-supported cause. When an impression is more likely to produce a purchase, an advertiser bidding for purchases can afford to pay more for it, and automated bidding does that without being asked. Meta credits its GEM ranking model with a 1.7% lift in offsite conversions in Q3 and its Andromeda retrieval work with another 1.4% on Facebook. Those are Meta's figures and unaudited. The conversion gains Meta reports for Q3 are in single digits; the US price increase is 23%.
The March attribution change moved reporting
On March 3, 2026, Meta announced that click-through attribution for website and in-store conversions would count only link clicks, with conversions after likes, shares, saves, and comments moving into a new engage-through category. Ads Manager's in-product notice dates the rollout to March 17. Meta's announcement says "There will be no change to how advertisers are billed." So the change cannot raise a CPM directly. It can raise one indirectly: if a campaign reports fewer conversions under the new definition, a cost-cap or ROAS-goal campaign may stop delivering, and the advertiser loosens the goal, which means bidding higher. That path is plausible and we found no data measuring it. Bïrch's March and April numbers do not show a step-up.
The midterm elections are too small
November 3 is a midterm, and 2025 had no federal election. AdImpact projects a record $11.6 billion in political ad spending for the 2026 cycle, but only $1.6 billion of it on digital, and that figure covers Facebook, Google, Snapchat, and X together across the whole cycle. Meta's US and Canada revenue was $28.9 billion in one quarter. Political money can push up prices in contested states in the final weeks; it cannot move the national average by 23 points.
A soft 2025 base explains the spring
In April 2025, ahead of the end of the US de minimis exemption, Temu cut its daily US ad spend across Facebook, Instagram, TikTok, Snap, X, and YouTube by about 31%, according to Sensor Tower data reported by Reuters. Less competition then makes this year's spring comparison look steeper, and Meta's "macro conditions" line points at it. By Q3 2025, though, US and Canada price per ad was already growing 13%, so the base was not depressed when this autumn's gap opened.
Is Google getting more expensive too?
Not at anything like Meta's rate. Google sold more clicks this year at nearly the same price; Meta sold a few more US impressions at a much higher one.
Alphabet does not report a CPM for search. Its quarterly filings report the change in paid clicks and cost per click on Google Search and its other owned properties (Gmail, Maps, Play), worldwide only, with YouTube and partner sites excluded. Those are the closest public equivalent of Meta's price per ad.
| Quarter | Google paid clicks | Google cost per click | Meta price per ad, worldwide |
|---|---|---|---|
| Q1 2025 | +2% | +7% | +10% |
| Q2 2025 | +4% | +8% | +9% |
| Q3 2025 | +7% | +7% | +10% |
| Q1 2026 | +13% | +5% | +12% |
| Q2 2026 | +13% | +3% | +12% |
Year-over-year change. Google figures from Alphabet's 10-Q filings with the SEC; Alphabet does not report Q4 on its own, only the full year (cost per click +7% for 2025). Both companies' figures are worldwide and in US dollars, so currency moves are included. Alphabet had not reported Q3 2026 when this was published.
In 2026 the two platforms diverged. Google's paid clicks grew 13% in each of the first two quarters while cost per click slowed from 5% to 3%. Meta's worldwide price per ad grew 12% in both quarters, and in the US the gap between price and volume kept widening. Agency data points the same way for the third quarter: Tinuiti's Q3 2026 benchmark put Google search cost per click up 4% year over year, as summarized by MediaPost (the full report is gated).
A click and an impression are different units, so this says nothing about which channel is cheaper for you; it says Meta reach got more expensive faster than a search click did this year. If you buy both, compare cost per purchase by channel on this year's numbers before setting Q4 budgets; a split that made sense in 2025 was priced on a smaller gap.
What a higher CPM does to your ROAS
Cost per purchase is CPM divided by the purchases you get from a thousand impressions, which is a thousand times click-through rate times conversion rate. If click-through and conversion hold still, cost per purchase rises exactly as much as CPM does, and ROAS falls by the same ratio: new ROAS is old ROAS divided by one plus the CPM increase.
| ROAS last October | CPM +14% (Meta, worldwide) | CPM +23% (Meta, US & Canada) | CPM +39% (Bïrch, US, Sep) |
|---|---|---|---|
| 2.5x | 2.19x | 2.03x | 1.80x |
| 3.0x | 2.63x | 2.44x | 2.16x |
| 4.0x | 3.51x | 3.25x | 2.88x |
Same creative, same click-through and conversion rates. A business at a 40% contribution margin breaks even at 2.5x, so its 3.0x campaign from last year is now below break-even at the US rate.
That is the worst case, and the median ecommerce account did better because the other two terms moved too. Triple Whale's 40,000-brand median for August 2025 to July 2026 shows CPM up 13% but click-through rate up 16% and conversion rate down 5%, so cost per purchase rose only 3%. Higher prices for impressions that click more often is what Meta's explanation predicts. Whether your account looks like that median or like the table above depends on creative, and you can only find out from your own cost per purchase, not from CPM. Run your own margin through the break-even ROAS calculator to see where your floor sits now.
What to do this quarter
1. Judge the account on cost per purchase, not CPM
A rising CPM with a flat cost per purchase is a healthy account in a pricier auction. A flat CPM with a rising cost per purchase is a creative or landing-page problem. CPM alone cannot tell the two apart.
2. Re-set cost caps from margin, not from last year's results
If a cost-cap campaign has stopped spending, the cap was probably set against 2025 prices. Work out the highest cost per purchase your margin allows and set the cap there. Raising it blindly until delivery returns is how a campaign ends up spending below break-even.
3. Compare like-for-like attribution before blaming efficiency
Since March, click-through conversions count only link clicks. A year-over-year drop in reported ROAS can be partly the definition changing. Compare the same window and settings, or use blended revenue against total ad spend from your store, which no attribution rule changes.
4. Treat creative as the main lever
CPM is the auction's number; click-through and conversion rate are yours, and creative sets most of click-through. In Triple Whale's data a 16% rise in click-through absorbed most of a 13% CPM increase. More ad variants, sharper hooks, and creator content that looks native to the feed are the levers that offset a price you cannot control. Meta reports its partnership ads product, which runs ads through creators, grew 97% year over year in Q3.
5. Plan Q4 on 2026 prices, and expect ugly year-over-year charts
Q4 2025 was the softest of the last six quarters for Meta's US and Canada pricing, up 9%, and Bïrch's November 2025 CPM was only 2% above September. Comparisons against last November will look worse than the change you actually feel week to week. Budget from September and October 2026 CPMs, not last year's.
If you sell internationally, look at regions separately. Meta's Q3 price per ad rose 1% in Asia-Pacific and 10% in Europe against 23% in the US and Canada and 24% in Rest of World. A blended account average can hide a US problem behind cheaper inventory elsewhere, or make a fine European campaign look worse than it is.
Frequently asked questions
Why are Meta CPMs so high right now?
Demand is outgrowing supply in the US. Meta's Q3 2026 deck shows US and Canada price per ad up 23% year over year while ad impressions there grew 8%. Meta attributes the price rise to ad performance gains, meaning advertisers bid more for impressions that convert better, and to better macro conditions than in 2025.
How much have Meta CPMs gone up in 2026?
It depends on the sample. For July to September 2026, Meta reports price per ad up 14% worldwide and 23% in the US and Canada. Bïrch's US tracker shows Facebook CPM up 39% in September. Triple Whale's median across 40,000 ecommerce brands rose 13% for August 2025 to July 2026.
What is a normal Meta CPM in 2026?
Triple Whale's median across ecommerce brands was $15.06 for August 2025 to July 2026. For US-targeted Facebook campaigns, Bïrch measured $22.31 in September 2026. Your own CPM depends on audience, placement, and season, so compare it with your own history rather than a benchmark.
Are Google Ads costs rising too?
Much more slowly. Alphabet reported Google Search cost per click up 5% year over year in Q1 2026 and 3% in Q2, while paid clicks grew 13% in both quarters. Meta's worldwide price per ad rose 12% in both. Tinuiti's Q3 2026 benchmark put Google search cost per click up 4%.
Did Meta's attribution change raise CPMs?
Not directly. Meta's March 2026 change limited click-through attribution to link clicks and said billing would not change. It could raise prices indirectly if advertisers loosened cost caps or ROAS goals after reported conversions fell, but no data we found measures that, and Bïrch's US CPMs did not step up in March or April.
Will Meta CPMs go down after the holidays?
They usually fall in January. In Bïrch's US Facebook data, CPM dropped from $14.53 in December 2024 to $11.84 in January 2025, and from $16.44 in December 2025 to $14.80 in January 2026. January 2026 was still 25% above January 2025, so expect a seasonal dip, not a return to last year's prices.
Want your account checked against these numbers?
Growthr runs paid media as an embedded team: cost caps set from contribution margin, blended attribution across every platform, and a dashboard that shows profit rather than CPM. $10,000 a month plus 10% of managed spend.
Related: What is a good ROAS? · Break-even ROAS calculator · Performance marketing · Data and attribution