Day 2 of US v. Google was jam-packed. We heard from four witnesses, three of which were from the buy-side.

James Avery, CEO at Kevel

Kevel is hosted on AWS, employs about 50 engineers, and receives 20K to 100K requests per second. They implement rolling updates to their systems with the expectation of zero downtime for customers.

DOJ walks through several of their remedies proposals, beginning with the proposal for AdX to provide real-time bids into rival ad servers and to participate in the pre-bid header bidding auction. Kevel currently has APIs with many big ad exchanges, and enabling access to AdX demand in a fair way would enable Kevel to compete for traditional publishers. Avery worries that AdX may bid differently into Kevel than DfP, and thus feels divestiture is necessary to mitigate the conflict of interest that creates the incentive for Google to deliver better bids to DfP.

We talk about software migration and begin to get into timelines and feasibility. Kevel is currently migrating from AWS to also run on Google Cloud Platform (GCP) at a customer’s request.  This started 3-4 months ago and should be done by Q1 2026; so the process will have taken about a year. They’ve also rebuilt their pipeline tech, changing the architecture and programming language. They’ve acquired Nexta, a self-serve advertiser front-end that is on Microsoft Azure and also slated to migrate to AWS and GCP. All of this is done or planned with no customer downtime. The connector for DfP to receive bids from pre-bid is something Kevel has - it took a month, with two to three engineers to build.

He says IAB Tech Lab, W3C or Prebid are all equipped to be administers of the Final Auction Logic, but reiterates the importance of a full DfP divestiture. Kevel would be interested in acquiring either AdX or DfP. There’s been no change since liability in the challenges Kevel faces with securing open-web display publisher customers, no plans to focus on open-web display in the absence of divestiture, and no impact of generative AI on the publisher ad server market.

On cross, Google attempts and fails at a “gotcha,” pointing to Avery’s deposition to indicate that in the Nexta acquisition, there are code dependencies that wouldn’t be moved over. In fact, Avery said the opposite in his deposition. There are no dependencies that would not be moved over. Google points to Avery’s perspective that AdX and DfP should be divested together, as he says without the exchange business, it would not be feasible to offer the ad server for free. Nevertheless, Avery confirms that he’d be interested in either company on its own, even with DfP’s final auction logic code being open-source. On redirect, DOJ asks him to explain why it’d be in Google’s interest to help with the migration onto GCP. Because Google makes money from that, too.

Luke Lambert, Partner, Head of Reputation Marketing & Insights at Omnicom Confluence

Lambert, who testified during liabilities phase, is now in a new role at Omnicom’s Confluence, the unit dedicated to Amazon’s media planning and buying. DOJ asks about Google’s statement that open-web display is already in rapid decline, and Luke says that he will continue to purchase as much as he can because display is most efficient and effective for his client’s upper-funnel needs.  It may slow, but won’t disappear.

He talks about transparency pain-points from the buyer’s perspective. He needs to optimize budget going toward working media, which he can’t do without transparency.  He says absent divestiture, it is unlikely that Google will provide more transparency - past behavior is a great predictor of future behavior.

Also, we learn that AdX take rates are no longer outlined in the Omnicom contract, thought they were 10 years ago, and are not negotiable. AdX effectively lives outside of Omnicom’s SPO decisions as well.

On AdX divestiture, he explains he’s seen exchanges change hands without service disruption to buyers. He couldn’t even tell when the technical migration actually happened, as it was seamless. No concerns about malware and fraud, as they rely on third parties for that. No concerns about privacy, either, from a sale of AdX to a new owner.

He says that the final auction logic is a black box. Open-sourcing it alleviates that concern for publishers, but not for buyers. He’d still want access to the same type of log-level data that prebid makes available to publishers. Asked about the provision prohibiting Google from using buy-side tools like DV360 to discriminate against non-Google products, he says “I thought that was the case already.” DSPs should be favoring efficient paths that are best for the advertiser; not for their own business.

Again, we see Google’s depiction of AdX bidding into prebid, where it isn’t clear that this is the only path. Lambert says there’s a difference between “enabling” AdX bidding into prebid, versus making it the way it is done, period.

Since liabilities phase, there have been no changes in competitor exchange offerings, no impacts of generative AI on the need for an exchange or on ad targeting. He mentions that his teams do not use Google Ads and Performance Max (PMax), Google’s black-box AI buying product, for display ads. Buyers want more control, he says, adding that Google actually recommended that their teams move their spend into DV360.

Cross-examination was once again, a lot of attempts at impeachment; but the main substantive revelation that stood out is that neither DOJ’s, nor Google’s proposals give buyers or their advertisers the full transparency that they are looking for.

Jay Friedman, Strategic Advisor and former CEO at Goodway Group

Direct examination quickly delves into the impacts of Generative AI. Friedman notes that there will be little effect on ad servers, and that there have yet to be effects on ad exchanges. He notes that GenAI would, however, help Google dramatically, because nobody has a bigger trove of data than Google.

We get into remedies proposals. First up, AdX divestiture. Divestiture would likely create fee competition, as Google’s rates have been an “anchor” at the top of the market.  There is not fair competition in the ad exchange market today, as the world seems to have accepted that “there’s Google, then there’s everyone else.” He says that remedies would need to address this, such that the ad server would stand on its own and operate in the interest of the publisher. Friedman says he wants to see AdX and DfP divested. Otherwise it’s like “taking half a dose of antibiotics.”

Google’s proposal for AdX to provide real-time bids to rival ad servers is more “ripe for problem than for solution.” Will DfP still be preferences? Will the bids come 25 seconds later? Similarly, prebid interoperability alone doesn’t address the concerns about self-preferencing. He also says his ideal world would see DfP divested to two companies, or made open-source. Further to the point, he says that a prohibition on self-preferencing would make advertisers more likely to increase use of Google’s tools, and would make him more likely to recommend them. On 1st party data self-preferencing restrictions, he is not concerned as long as advertisers can still use the data.

Cross-examination, like Friedman’s last time on the stand, drilled into public statements and blog posts in a way that wasn’t particularly fruitful. He is asked about “minimizing” the effect of AI on the market, and he rightly clarifies that his statements were specific to the publisher ad server and ad exchange markets. Through cross and redirect, Friedman explains there are too many ad exchanges in the market, so the idea of AdX divestiture failing is missing the point. Once it is no longer under Google’s ownership, this is effectively success - there are plenty of other exchanges that are now unlocked and able to compete. There are some effectively superfluous questions about the retail media network ad server market being crowded, AI agents, and other AI-driven ad tech offerings that have nothing to do with the open-web display ad server or ad exchange markets.

Cross-examination ends with a bewildering question from Dunn about whether Friedman has any evidence Google is using its 1st party data for ad targeting. I would hazard a guess that most of the industry would be shocked to hear Google answer “no” to this one.

Jed Dederick, CRO at The Trade Desk (TTD)

Dederick describes the current state of the open-web display market as anemic. There is a lack of transparency, and thus a lack of trust that hinders investment and innovation. He is hopeful that remedies here will allow for that advertiser trust to be restored.  He says that what is necessary to achieve this is severing the ownership between Google and its sell-side tools.

On impacts of AI, he says that he’s seen far more impact on the Search market.  AI impact is radically different in the Search market as it is changing consumer behavior. Not so in the adtech ecosystem. Nevertheless, AI disruption will benefit Google most as they have the most data, he notes. He says that Google’s proposal to bid into rival ad servers is insufficient. When an entity owns the ad selection process, he says, trying to name all of the ways you can use that to manipulate the auction or take undisclosed fees is impossible - the problem is the incentive. He wants a fair, competitive, transparent auction for the advertisers TTD represents.

Dederick states that if DOJ’s remedies are not imposed, he says there will be a sense that “they got away with it.” Isaacson objects on hearsay grounds here, and Wood asks him not to interrupt her witness. Judge Brinkema voices her frustration with cumulative testimony.

This frustration reaches a fever pitch during cross-examination, when Isaacson seemingly attempts to re-litigate market definition for the sake of the record. Judge Brinkema notes that industry witnesses are at this point “window dressing,” and what she will be looking closely at are the DOJ’s technical experts, and what Google employees have to say about technical feasibility.

To that end, DOJ advised at the end of Day 2 that they would like to rearrange their witnesses. While the next witness was set to be another publisher, Matthew Wheatland of Daily Mail, they now intend to call technical experts, pending Google’s confirmation that they can accommodate the change. This would have DOJ wrapping their case in chief on Thursday, unless cross-examination runs long.

There was a point where Isaacson was trying to get into the record deposition statements by Dederick that Pubmatic, Magnite, and Index engage in undisclosed markups. It is unclear what relevance this has to the remedies trial, and seemed more an attempt to rattle Dederick and put him in an uncomfortable position. Isaacson represents the statements as definitive, whereas Dederick says it was a suspicion - that’s the nature of undisclosed markups.

Despite all of this, Dederick is the winner of today’s levity prize. At one point, a frustrated and frazzled Isaacson was trying to get at what Jed meant by “final say” in the auction and if it’s the same or different as “final auction logic.” Wood jumps in as well, before Judge Brinkema interjects to ask if Dederick has any idea what they’re talking about. “No, I’m just the window dressing.”

During redirect, he reminds us all why this trial is critical. Journalism relies on open-web display advertising revenue. If buyers were more confident in where their money was going, more money would go to fund quality journalism and news. Today, open-web display advertising is starving. Remedies would bring more trust, which would breed investment and innovation. This trial gives him hope, for the first time in a long time, and TTD isn’t giving up on open-web display any time soon.

What’s to come:

As noted, DOJ would like to call technical experts and a Google employee tomorrow.

  1. Jon Weissman
  2. Tim Craycroft (Google)
  3. Goranka Bjedov (ex-Google & ex-Meta)

DOJ hopes to rest their case in chief on Thursday, unless cross-examination pushes them into Friday. Rocket Docket, indeed.

Further, the parties have proposed November 3rd for submission of post-trial briefs, and November 17th for closing statements.