Programmatic advertising is a $595B market (2024), projected to reach $779B by 2028. Roughly 90% of digital display is bought automatically. At the center of publisher-side infrastructure sits an open-source library used by over 10,000 companies: Prebid.js.

This analysis is for decision-makers: what Prebid does, how it works, why it became the standard, and which levers determine business results.

300+
bid adapters
~90%
display = programmatic
$203B
US programmatic 2026
10K+
companies using Prebid

The Problem: Waterfall Leaves Money on the Table

Publishers historically sold inventory through a waterfall — ad networks called sequentially by priority. The first one that accepted the price got the impression. Others never competed, even if they would have paid far more.

Header bidding fixed this: all buyers bid simultaneously, highest price wins. Prebid.js is the open-source library that became the industry standard for browser-side header bidding.

Waterfall

Network A (priority 1)$2.00 — wins by default
Network B (priority 2)Never asked
Network C (priority 3)Would have paid $5.00 — never asked
ResultImpression sold for $2.00

Header Bidding (Prebid)

Network A$2.00
Network B$3.10
Network C$5.00 — wins
ResultAll bid in parallel → $5.00 wins (+150%)
+20–50%
waterfall → header bidding uplift
+70%
The Telegraph (case study)
+25–50%
CPM uplift, Future plc
30–40%
average portfolio uplift (AdPushup)

How It Works: 5 Steps in ~1000ms

1

Page loads → auction starts

Prebid.js identifies ad slots and simultaneously sends bid requests to all connected buyers.

~0ms
2

Buyers respond with bids

Each DSP/SSP receives slot data and returns a bid or pass. All in parallel.

200–800ms
3

Timeout cuts slow responders

Those who didn't respond are excluded. UX protection.

timeout: 1000–1500ms
4

Best bids → ad server

Winning bids sent to Google Ad Manager to compete with direct deals and AdX.

~50ms
5

Ad server picks winner → ad shown

GAM compares all sources and serves the highest-paying creative.

~1100ms total
Prebid doesn't replace the ad server. It creates competition before the decision, increasing the effective price of every impression.

Market Context

$1.14T
global ad market 2025
$595B
programmatic (global) 2024
$203B
US programmatic display 2026

The ANA Supply Chain Study (2023) found only 36 cents per advertiser dollar reached publishers. By Q3 2025 this improved to 47.1¢ (+11 points), but $26.8B/year is still lost to supply chain inefficiency.

Advertiser
$1.00
DSP + SSP fees
−29¢
transactions
Waste
−24¢
IVT, MFA
Publisher
47¢
was 36¢
$26.8B
annual supply chain losses
47.1%
publisher share (was 36%)
0.8%
MFA inventory (record low)
87.8%
PMP share (was 64.5%)

Google Antitrust: Why Prebid Matters More

In April 2025, a US federal court found Google guilty of illegally monopolizing publisher ad server and ad exchange markets. Texas settled for $1.375B in May. The EU followed in September with a €2.95B adtech antitrust fine. The DOJ seeks divestiture of AdX.

~90%
Google's ad server share
€2.95B
EU adtech fine

“Google willfully engaged in a series of anticompetitive acts to acquire and maintain monopoly power in the publisher ad server and ad exchange markets.” — Judge Leonie Brinkema, April 17, 2025

For Prebid, this is a structural tailwind. If Google must unbundle its ad server from its exchange, vendor-neutral auction solutions become critical infrastructure.

Three Architectures

Client-side

Auction in the browser
  • Buyers: 15–20 (limit)
  • Latency: grows with buyers
  • Cookie access: full (max CPM)
  • Best for: mid-size publishers

Server-side

Auction on server (Prebid Server)
  • Buyers: unlimited
  • Latency: −40%
  • Cookie access: limited (−20–40% CPM)
  • Best for: CTV, App, AMP

Hybrid (industry standard)

Both — optimal balance
  • Buyers: 5–8 client + rest server
  • Latency: controlled
  • Cookie access: full for key buyers
  • Best for: large publishers
Hybrid is the industry standard: 5–8 key buyers client-side + rest server-side. Prebid Server delivers up to 40% latency reduction.

Business Levers

Timeout: Revenue vs. UX

800ms
~55%
~55%
1000ms ←
~75%
~75%
1200ms
~88%
~88%
1500ms
~95%
~95%

Price Granularity

Type Step GAM Lines Revenue Loss
Low $0.50 ~40 up to $0.49 (high)
Medium $0.10 ~200 up to $0.09 (moderate)
High $0.01 ~2,000 $0.009 (minimal)
Custom Variable Optimized Controlled (recommended)

Buyer Selection: Diminishing Returns

3 buyers
~60%
~60%
5 buyers
~80%
~80%
8 buyers ←
~92%
~92%
15 buyers
~97%
~97%
20+ buyers
~99%
latency↑

80% of incremental revenue comes from the first 5–7 buyers. Sweet spot: ~8 client-side.

CPM by Geography

Average banner CPMs (SSP-side, 2024):

US
$1.43
$1.43
UK
$1.05
$1.05
Germany
$0.90
$0.90
France
$0.80
$0.80
Brazil
$0.50
$0.50
India
$0.25
$0.25

An 8× spread that defines monetization economics. Q1 2025: US display CPMs fell −33–42% YoY after a record political ad year. By December 2025: display +6.3% YoY, video +33.2% YoY. This volatility underscores the need for dynamic floor prices — exactly what Prebid’s Floors Module does.

Privacy: Post-Cookie Adaptation

User ID — UID2, SharedID, EUID, LiveRamp
First-party data
Topics API / Protected Audiences
GDPR / CCPA / GPP consent
40%
US marketers using 1P data as primary targeting (2025)
60–80%
CPM preserved with Prebid User ID vs. no identity

Beyond Display

Format Architecture Market
Display Client + Server Core programmatic
Video (instream) Client + Server Highest CPM
Mobile in-app SDK → Server 71% programmatic
CTV / OTT Server only 44% share (Q2 2025), >$45B
Retail Media Server $30B+ by 2026, +29% YoY
DOOH Server +400% since 2019
$110B+
US programmatic video 2026
44%
CTV share (Q2 2025, was 28%)
$30B+
retail media by 2026

Ecosystem and Competitors

2015
Prebid.js launched (AppNexus + partners)
Open-source header bidding library released.
2017
Prebid Server + Mobile SDK
Server-side auction and mobile support added.
2019–20
95% of top US publishers on header bidding
Header bidding becomes the industry norm.
2021–22
User ID, Floors, GDPR modules
Privacy-era modules expand the platform.
2024–25
v11, CTV, 300+ adapters, PAAPI
Multi-format expansion and post-cookie readiness.
2025–26
Google antitrust → neutral infra more critical
Court ruling accelerates shift to vendor-neutral solutions.
Solution Type Differentiation Lock-in
Prebid.js Open source Neutral, auditable None
Amazon TAM Proprietary Tied to Amazon DSP Medium
Google Open Bidding Server-side in GAM Conflict of interest High
Index Exchange Proprietary Quality, but vendor-dependent Medium

Prebid runs in parallel with TAM and Open Bidding. Per Roxot data, client-side and server-side header bidding generate ~22% of programmatic revenue each, while AdX accounts for ~56%.

What Prebid Doesn’t Do

Doesn't optimize placement

Ad slot position and size are publisher decisions — Prebid handles demand, not supply-side layout.

Doesn't replace the ad server

GAM is still needed for direct deals, frequency capping, and final decisioning.

Doesn't fix traffic quality

Bots, low viewability, and invalid traffic are root-level problems outside Prebid's scope.

Not plug-and-play

An engineering tool requiring ongoing configuration optimization and monitoring.

Bottom Line

Working without header bidding in 2026 means leaving 20–40% of programmatic revenue on the table. The strategic question isn’t “do we need Prebid” but which configuration is optimal. With Google antitrust and the post-cookie transition, vendor-neutral solutions are more critical than ever.

+20–50%
waterfall → header bidding
+5–15%
+ server-side
+3–10%
config optimization

Sources

  1. Prebid.org 2. Prebid.js GitHub 3. ANA Q2 2025 4. ANA Q3 2025 5. ANA 2023 Study 6. dentsu 7. Basis 8. eMarketer 9. AsterioBid 10. DataBeat 11. Digiday 12. Digiday Research 13. AdPushup 14. Mile.tech 15. Criteo 16. US v. Google 17. Axios 18. Raptive 19. Kluwer 20. Roxot 21. Enlyft 22. 6sense 23. Marketing Brew 24. Prebid Server Docs
AI Disclosure: Research conducted with Claude (Anthropic). Human editorial direction and domain expertise.