Why Breaking Up Google Is Hard to Do

Attention, Dispatch Premium members (and those of you on the fence about becoming Dispatch Premium members): We’re hosting our first Dispatch Premium Town Hall with Steve Hayes and Jonah Goldberg on Tuesday, December 10, at 8 p.m. ET, providing members with an opportunity to interact directly with The Dispatch’s co-founders in an ask-us-anything Zoom call. What do they think of Trump’s Cabinet nominations so far? What’s on the horizon for the 119th Congress? How are they sourcing up for a very different Washington?

We’ll send out more details in the coming days, including an RSVP link exclusively for Dispatch Premium members. If you’d like to attend this special live town hall, be sure to join Dispatch Premium before next Tuesday!

Welcome back to Techne! Orca whales might be going through their own retro era. An orca off Washington’s coast was spotted balancing a salmon on its head, reviving a behavior last seen in the 1980s. Some scientists think the salmon on the head should be viewed as a fashion statement. 

There’s an image that haunts me, in a good way, because it visualizes a tendency that I want to avoid. 

It comes from a group of researchers at the University of South Australia that asked 1,036 people in the United States how they felt about music from across various decades. The people varied in age from 18 to 84 years, but they all followed a similar pattern. They thought very highly of the music of their adolescence, teenage years, and early adulthood. But as they approached age 30, their approval of current music declined. 

To me this is evidence of the adventure window. Once we hit 30 or so, our willingness to try new things begins closing. Writer and researcher Adam Thierer introduced me to the concept back in 2011, just as I was hitting my mid-20s and beginning to notice this within myself.

The phrasing comes from Robert Sapolsky, the famous neuroscientist, who was cataloging the tendency in humans and all mammals to stop trying new things later in life. As Thierer explained it, Sapolsky “came to study this phenomenon after growing increasingly annoyed with his young male research assistant, who would come to work every day of the week listening to something new and quite different than the day before. Meanwhile, the much older neuroscience professor lamented the fact that he had been listening to the same Bob Marley tape seemingly forever.”

The lesson I take from Sapolsky is this: The world changes, so it is important that you keep yourself open to change as well. To the detriment of consumers, competition, and good governance, this is a lesson not well understood by the Department of Justice in its ongoing antitrust case against Google. How people access information is changing, the search market is changing, and yet the DOJ still clings to an outdated map of the world. 

The August 8 edition of Techne focused on the DOJ’s case against Google, which argued that the tech giant was acting anticompetitively in the search engine market. In the heat of the summer, a judge agreed and ruled that Google “has acted [through exclusive deals] to maintain its monopoly.” Despite acknowledging that Google is the best search engine, the court found issues with the contracts it negotiated to be the default search engine in Apple’s iPhone and iPad, the Android operating system, the Firefox browser, telephone operators like Verizon, and a range of different players in the ecosystem. 

Two weeks ago, the DOJ put forth its proposal for remedies— and they are by far the strictest and most onerous available. The government wants to break up the company by forcing Google to cleave off the Chrome browser and the open-source Chromium project that the company has helped build. The DOJ also wants to end all the default deals, including those in the Android operating system that Google helped grow. The government also wants to limit the company’s investment in AI that might affect search as well as stop any future deals involving this space. And on top of this, Google must share data with rival search engines. 

It is a grab bag of horrible ideas that are also incredibly hard to implement. In 2018, I explained the difficulties in breaking up a tech company, noting that “the government has to break up the firms’ teams and underlying technology. It also has to put into place a legal and regulatory system that keeps the targeted firms separate from other markets.” 

That means that in breaking up Google the DOJ is asking the court to decide the fate of Polymer, an open-source JavaScript library for web applications; AngularJS, a web application framework; AMP, which improves the performance of web content and advertisements; Dart for making applications on Internet of Things devices; Flutter for Android and iOS mobile apps; Borg for cluster management; Colossus, the scalable storage system; BigTable for databases; Spanner, another database management system; Chubby, the lock service; Blobstore, which runs the Google Cloud Storage; Stackdriver, which monitors Google Cloud; and a bunch of others I’m sure I have missed. 

The disconnect between the DOJ’s proposed remedies and the current market landscape is striking. Indeed, it was a bit of a shock when I read these opening lines from the agency:

The purposes of the following remedies are to unfetter the monopolized markets from Google’s exclusionary practices, pry open the monopolized markets to competition, remove barriers to entry, and ensure there remain no practices likely to result in unlawful monopolization of these markets and related markets in the future by prohibiting contracts that foreclose or otherwise exclude Competitors, including by raising their costs, discouraging their distribution, or depriving them of competitive access to inputs.

And yet, when I look at the broader search market, I see it transitioning. Yes, Google is still dominant in search engine queries, and yes, Google is still projected to grow. But Google’s nearest competitors and the broader search advertising segment are both growing even faster. That’s why Google’s share of the search ad market is expected to drop below 50 percent next year for the first time in over a decade. 

Search is a bigger cohort that includes search engine queries. Likewise, search ads include more than just Google searches. Customers are beginning their searches elsewhere, like on Amazon and other e-commerce platforms. Amazon is expected to have 22.3 percent of the market this year, with a 17.6 percent growth rate, compared to Google’s 7.6 percent growth, according to the eMarketer research firm. And for reasons I also don’t completely understand, there is a growing segment of searches being conducted on TikTok. TikTok’s share of the search ad market is small at 3.4 percent, but that revenue is projected to grow 38% this year.  

To emphasize the fast-changing environment, OpenAI put itself in direct competition with Google when it rolled out its long-anticipated search engine in October, between the judge’s decision in the first part of the Google case, the liability phase, and the second part of the case, the remedy phase that’s going on now. When it comes to searching for basic things these days, though, I use Perplexity instead of either OpenAI or Google.  

While I’m skeptical, let’s for a moment assume that courts got it right and that Google is locking up competition through its default contracts. What then? Well, you could apply a default-option intervention, which has already happened in Europe.

As part of a 2018 case against Google, the European Commission forced Google to implement a choice screen for general search providers on all new Android phones and tablets sold in the European Economic Area and in the United Kingdom after March 2020. During initial device setup, users were presented with four search engine options, allowing them to choose their preferred default search provider. Economists love this type of action process because it helps to reveal the true price of a good.

And how did the change affect the market? According to a trio of economists, it moved the market share about 1 percentage point. It’s not nothing, but it does show just how small an impact we are talking about.

Even more fundamental, the economic foundations of the DOJ’s case are shifting, as perspectives on behavioral economics continue to evolve. The case was built on a body of literature about the value of defaults. But more subtle and thoughtful experiments have gotten closer to replicating the real world. I’ve got a note on my website that explains more details. 

Economists from Yale, Harvard, and Berkeley just last month released a report re-estimating the effect of automatic enrollment defaults in 401(k) savings plans. They find that the effect was smaller than previously thought. Auto enrollment was thought to have boosted savings rates by 2.2 percent of income, but the new paper cuts that down to just 0.6 percent. This example shows a lot more caution is warranted on lessons from behavioral economics, upon which the Google case is fundamentally built. 

If you want to do a deeper dive on the legal side of this case, I’d direct you to Geoffrey Manne’s “A Critical Analysis of the Google Search Antitrust Decision.” Manne used to be my boss, so blame him for any errors I make in antitrust law. But this piece does a good job going into all of the legal minutiae, and especially the problem of bifurcation, which is when a case is split into two parts, the liability phase and the remedy phase. With Google, the DOJ separated the arguments over harm from the arguments over remedies. Manne summarized the big takeaway on X:  

You don’t get to “remedy” conduct that you didn’t prove in the first place was anticompetitive by claiming you have the right to ensure “there remain no practices likely to result in [future harm].” That’s not how the law works. 

This is exactly why many of us cried foul when they bifurcated the liability phase from the remedy phase of trial. The reason the DOJ wanted that was because it knew it would be easier to “prove” its case if it didn’t have to address the obvious arguments that rectifying the challenged conduct would accomplish nothing — thus strongly suggesting that the challenged conduct wasn’t anticompetitive.

But now having gotten that bifurcation, the DOJ isn’t given carte blanche to seek to impose any remedy the DOJ can think of, addressing conduct that was in no way proven — or even claimed — to be anticompetitive in the liability phase.

In an interesting twist of fate, the last time the DOJ tried to break up a tech company was in the 2000 Microsoft antitrust case, which concerned Microsoft’s supposed dominance in browsers. It’s ironic, then, that if the DOJ gets its way, Google will no longer be able to support Chrome or the Chromium project and Firefox will probably fail because it gets nearly 80 percent of its revenue from the default deal. An entire ecosystem will collapse, leaving Microsoft in a position of dominance. 

The Department of Justice’s adventure window has closed. While the world of search evolves rapidly, the DOJ remains fixated on an outdated understanding of the market. Its proposed remedies are built on shaky behavioral economics grounds that are being revamped. The cruel irony is that in trying to pry open competition, the government may instead close the window on innovation, leaving us with fewer choices than we had originally. Sometimes the most dangerous thing isn’t being unwilling to try new things, it’s clinging so tightly to old ideas that you break what’s working.

• As was largely expected, the Commerce Department introduced new export controls directed at the Chinese semiconductor industry on Monday. Here is my X thread about it. The big changes:
  • Restrictions on the sale to China of high-bandwidth memory.
  • Restrictions on electronic computer aided design (ECAD) and technology computer aided design (TCAD) software to design chips.
  • More than 100 Chinese companies have been added to the entity list, including tool manufacturers, fabs, and investment companies. Shenzhen Pengjin High-Tech, which makes semiconductor equipment, and ChangXin Memory Technologies, which is working to develop AI memory chip technology, were not added to the entity list, as initially expected.
• Gail Slater, a former economic adviser to Vice President-elect J.D. Vance and well-known antitrust expert, is Donald Trump’s pick to lead the Department of Justice’s antitrust division. Trump also put forth Jared Isaacman as the next NASA administrator. My little corner of the internet cheered the nomination of the respected aerospace founder. 
• What an opener of a paragraph in the New York Times: “In a guarded compound at the foot of the Rockies, government scientists are working on a new kind of global alarm system: One that can detect if another country, or maybe just an adventurous billionaire, tries to dim the sun.” My August 15 edition of Techne is all about geoengineering.
• I wrote about DOGE and the broader effort to make government efficient in the November 14 edition of Techne. Since then, here’s some commentary I’ve enjoyed on the subject: Cass Sunstein, the former as head of the Office of Information and Regulatory Affairs, offers up six things to consider when reforming government; Nicholas Bagley, a legal expert in administrative law, explains why bureaucrats are needed to get things done in The Atlantic; economist James Broughel explains how DOGE can cut thousands of regulations; and economist Arnold Kling lays out what he’d do if he was in charge of DOGE.        
• Earlier this year, writer Chris Arnade summed up an argument you often hear about economic growth: “When there’s a need to convince people, ‘actually things are good!’, it kinda raises the question, maybe there’s a disconnect between what we think is good, and what makes people content, fulfilled, & happy.” Writer and researcher Kevin Drum recently pulled together data dating back to 1976 on the percentage of Americans who say they feel financially satisfied or dissatisfied. People’s sentiments didn’t change that much even with all the economic growth.
• I knew the ocean contained large amounts of thorium, but I wasn’t aware of all of the other metals down there. “There is more metal contained in polymetallic nodules on the ocean floor, than in all known terrestrial reserves combined. Long-term, sourcing metal from the ocean will be the most significant impact on abundance and cost of metal.” That’s from Matt Gray, who’s “working on developing equipment for seabed research specifically because there’s so many unknowns we have to figure out before we can extract responsibly.” I couldn’t find much on Gray’s background, so I plan to learn more about polymetallic nodules, also known as manganese nodules due to their high manganese content, and report back.  
• Economist Alex Tabarrok is right: “H5N1 will likely pass us over—but only the weak rely on luck. Strong civilizations don’t pray for mercy from microbes; they crush them. Each new outbreak should leave us not relieved, but better armed, better trained and better prepared for nature’s next assault.”

In case you missed it, I wrote about three lessons from Aaron Wildavsky’s classic Searching for Safety. The political scientist’s timeless wisdom on risk offers a clear way to think about AI safety. Safety isn’t chosen, it’s searched for. 

Last month, OpenAI filed comments in a government inquiry on “Bolstering Data Center Growth, Resilience, and Security.” The nine pages are a quick read and lay out the central tension in the data center debate: “With an estimated $175 billion in global infrastructure funds waiting to be committed, the question is not whether that funding will flow, but where. If it doesn’t flow into U.S.-backed global infrastructure projects that advance a global AI that spreads the technology’s benefits to the most people possible, then it will flow to China-backed projects that leverage AI to cement and expand autocratic power. There is no third option.”

Holly Elmore, the executive director of the PauseAI advocacy group, details the tensions within the AI safety crowd in an extended X thread. I think the thread does a good job of explaining the internal politics driving the AI safety crowd.