Previous plaintiffs have argued unsuccessfully that Google’s Jedi Blue agreement with Facebook is anticompetitive and illegal. The agreement grants Facebook preferential access to Google’s dominant digital advertisement system in exchange for not building competing technologies. The plaintiffs’ challenges to Jedi Blue would have been on stronger ground had they argued that Jedi Blue is compelling evidence of illegal monopoly maintenance, as occurred in Microsoft, writes Joshua B. Gray.
During the course of their investigation into Google’s practices in digital advertising technology markets, the group of state attorneys general led by Texas uncovered a network bidding agreement between Google and Facebook, which Google called Jedi Blue. Jedi Blue offers Facebook privileged treatment inside Google’s digital advertising stack, through which advertisers and publishers buy and sell advertising space, in return for not building a competitor. The attorneys general, and subsequent private plaintiffs, challenged Jedi Blue in lawsuits against Google alleging claims of collusion and deception, but a judge dismissed the antitrust claims, finding that they failed to plausibly demonstrate that the agreement was anticompetitive. Even if that decision was correct, the court never heard the most compelling antitrust challenge to Jedi Blue.
At the heart of the Jedi Blue agreement is an explicit monopolistic bargain: Google grants Facebook special benefits on the condition that Facebook not support new technologies that could enable web publishers to escape Google’s control. Facebook’s absence as a competitor, and as the only potential competitor with the size and resources to compete with Google’s market power, enables Google to monopolize the market, which it does by also preventing publishers from using and sharing their own data to build competing ad exchanges. Google’ use of its control over data and auction rules to selectively block off a disruptive competitive threat entrenches its monopoly power. This is classic exclusionary conduct that follows the precedent set by Microsoft.
In a 1998 op-ed “What Antitrust is All About,” Robert Bork explained his support for a government enforcement action against Microsoft through the lens of his favored Supreme Court precedent construing Section 2 of the Sherman Act, Lorain Journal Co. v. United States, 342 U.S. 143 (1951):
When a monopolist employs practices and makes agreements that exclude competitors and does so without the justification that the practices and agreements benefit consumers, the company is guilty, as was The Lorain Journal, of an attempt to monopolize in violation of Section 2 of the Sherman Act. When its own documents display a clear intent to monopolize through such means, the case is cold.
Bork had concluded that Microsoft’s conduct was “predatory” by his definition because Microsoft executives intended to preserve the company’s monopoly in personal computer operating systems through practices that excluded, or severely hindered, rivals and an emerging technology—specifically Netscape and Java—but did not benefit consumers. Microsoft had “forced” its trading partners dependent on its operating system to agree to terms that conflicted with their commercial interest. Thereby, Microsoft “denied an important distribution channel” to Netscape and Java for their cross-platform technologies that could have opened up new pathways to competition against the Windows operating system. Bork’s support of Microsoft galvanized a generation of conservative antitrust lawyers trained in Chicago economics, many of whom came to view Microsoft as a paradigm of the wise use of Section 2.
With hindsight, the Microsoft case, following Bork’s construction of Lorain Journal, anticipated the next generation of economics emanating from the Toulouse School of Economics concerning what we now call “platforms” and a set of questions about monopoly maintenance that we now group together under the rubric of “entrenchment.” That post-Microsoft learning should have become the lens through which critics of Google’s monopolies in digital advertising evaluated the Jedi Blue agreement between Google and Facebook. So far, we have not analyzed the Jedi Blue this way.
Facebook’s threat to Google was less conjectural than the nascent threat to Windows from diminutive Netscape. Facebook was, and remains, a peer to Google, the only potential rival in with control over a unique demand source (Facebook Audience Network) that approaches Google Ads in scale. Facebook, like Netscape in the Microsoft story, had the power to create a new path for competition, and executives at Google unambiguously recognized this possibility as had Bill Gates in the Microsoft case.
Microsoft intentionally squelched the competitive threat to its operating system from Netscape and Java. Google acted with similar intent when it used its control over the publisher ad server to conditionally grant a uniquely capable Facebook enhanced user-identification and auction advantages on the express understanding that Facebook not compete, directly or indirectly, against Google stack—its integrated system of digital ad services.
Google’s publisher ad server confers control over data about web users and how impressions, the number of times an ad is seen, are routed. Google’s decisions about data access and auction rules determine who can see high quality user signals, and, importantly, who must bid blind. That bottleneck role is the functional equivalent of the distribution channels at issue in Lorain Journal and Microsoft: Google’s monopoly ad server is the gate through which effective competition must pass, and a monopolist can use that gate to favor some paths and to cripple others.
In 2017, header bidding, in which ad publishers auction space to multiple bidders at once, was a primitive technology with obvious potential. Previously, ad auctions used a waterfall method, in which space was sold sequentially to advertisers in a predefined order, or through Google’s exchange AdX, which favored Google Ads. Publishers loved header bidding because, as a former Google sales manager said, it increased the yields from their advertising inventory by allowing “all sources of demand to compete fairly, in real-time (or as close to real-time) as possible.” The manager further observed that “While I can help our Sales team build a story around latency and lost impressions… the fact remains that if we allowed for real-time competition across all demand sources, we wouldn’t even need to have that conversation!”
Header bidding threatened to reduce Google’s role in digital ad auctions by enabling publishers to solicit bids from multiple exchanges before routing impressions through Google. Publishers could have promoted header bidding and successor technologies to strengthen independent exchanges, build up new demand sources, and reduce Google’s gatekeeping power. If Facebook contributed its demand to independent exchanges supporting these emerging technologies, they could have caught up with Google quickly to become the foundation of an open, interoperable system.
As such, Google bought off Facebook with promises to share access to data, lower its fees, and grant exclusive insider status. The non-price advantages Google conferred on Facebook—enhanced user identification, richer signals, and preferential integration into Google’s auctions—are not generally available to publishers or to independent exchanges. Publishers and their advertisers also continue to pay the monopoly fee or take rate.
The Jedi Blue agreement is explicit about this plan:
1. Volume commitments guaranteed that Facebook demand flowed through Google rather than header bidding channels.
2. Data restrictions prohibited Facebook from using Google user or platform data to develop competing programmatic advertising platform services or to bid outside the Google-defined and controlled framework.
3. Auction protections promised not to disadvantage Facebook in Google’s “Final Clearinghouse Auction,” but provided no comparable assurances to other bidders.
4. Structural limits prevented Facebook from acting as an intermediary that could route demand to non-Google inventory.
In these ways, Jedi Blue traded valuable identity and auction advantages for Facebook’s commitment to stand down. It made these benefits conditional on Facebook continuing to exclusively use Google’s stack. If the details of this agreement sound familiar, it is because Google entered an economically similar pact with Apple to induce Apple to back away from any form of competition against Google’s search and search advertising monopolies. A court found this behavior violated Section 2 in 2024.
Framed in Bork’s terms, the problem is not that Google integrates with a major partner. Rather, it is the conditions Google imposes for what Bork would call a “predatory” purpose. The conditional nature of the benefits Google grants to Facebook alone reveals their true purpose: not improving product performance for users generally, but purchasing the withdrawal of a disruptive, competitive threat. That is precisely the sort of bargain Microsoft treats as exclusionary: trading selective favors for a key counterparty’s agreement not to compete in a manner that could have redounded to the benefit of consumers.
Google took further steps to “force”—Bork’s term—weaker parties to act against their own commercial interest in promoting competition. Tellingly, Google denies to its comparatively dependent publisher customers any user-identity tool for use with rival exchanges. If enhanced identity matching and integration were necessary to improve auction efficiency, Google could have made similar capabilities available to publishers, or directly to rival exchanges. The far more favorable commercial terms Google grants to Facebook, conversely, are a good proxy for the types of commercial terms major publishers could have reached in a but-for world where publishers are permitted to use their own data to solicit bids from several advertising platforms.
Google took a second additional step by also preventing publishers themselves from creating similar tools with their own data to share with independent exchanges, because (again) such tools could be used to support competition against Google’s stack. The discrimination is thus two-fold: Facebook gets superior identity data and auction treatment for its acquiescence, and everyone else is denied those tools from any source where the tools might revive the competitive threat Facebook has agreed to abandon.
The anticompetitive effect follows directly. By locking Facebook’s demand inside Google’s privileged route, the Jedi Blue agreement reinforced Google’s dominance: advertisers concentrate their spending within Google; publishers became more dependent on Google’s auctions; and rival exchanges struggle to achieve necessary scale. Weakened potential rivals did not, and could not, receive equivalent data access, auction integration, reduced fees, or performance protections. The result is that the primary path by which independent exchanges could have grown into serious competitors—publisher-driven header bidding backed by a large demand source—is intentionally weakened. The market tips further toward a single integrated stack, not because others cannot match Google, but because the most capable sponsor of an alternative architecture was paid to forfeit the game.
Author Disclosure: The author’s firm, Sperling, Kenny Nachwalter, occasionally counsels clients who are customers and rivals of Google about rights and remedies in both ad tech and search. The author has worked on several such matters over the last few years.
Articles represent the opinions of their writers, not necessarily those of the University of Chicago, the Booth School of Business, or its faculty.
Subscribe here for ProMarket’s weekly newsletter, Special Interest, to stay up to date on ProMarket’s coverage of the political economy and other content from the Stigler Center.
