Page 2On Appeal from the United States District Court for the District of New Jersey (D.C. No. 1:23-cv-02536) District Judge: Honorable Karen M. Williams

Argued: September 17, 2025

(Opinion filed: July 29, 2026)

Christopher J. Cormier Burns Charest 2445 M Street NW Suite 740 Washington, DC 20037

Joseph J. DePalma Catherine B. Derenze Lite DePalma Greenberg & Afanador 570 Broad Street Suite 1201 Newark, NJ 07102

Joseph Z. Fields [ARGUED] Susman Godfrey One Manhattan West 395 9th Avenue, 50th Floor New York, NY 10001Page 3 Shawn Raymond Susman Godfrey 1000 Louisiana Street Suite 5100 Houston, TX 77002

Mindee J. Reuben Hausfeld 325 Chestnut Street Suite 900 Philadelphia, PA 19106 Counsel for Appellants

Jason S. Rathod Migliaccio & Rathod 412 H Street NE Washington, DC 20002 Counsel for Amicus Appellant Open Markets Institute

Joshua P. Davis Matthew I. Summers [ARGUED] Berger Montague 505 Montgomery Street San Francisco, CA 94111

David O. Fisher American Antitrust Institute 1025 Connecticut Avenue NW Washington, DC 20036 Counsel for Amicus Appellant American Antitrust Institute Page 4Sam Auld Boris Bershteyn [ARGUED] Michael H. Menitove Andrew Muscato Kenneth Schwartz Tansy Woan Skadden Arps Slate Meagher & Flom One Manhattan West New York, NY 10001 Counsel for Appellees Caesars Entertainment, Inc., Boardwalk Regency, LLC, DBA Caesars Atlantic City Hotel & Casino, Harrah’s Atlantic City Operating Company, LLC, DBA Harrah’s Resort Atlantic City Hotel & Casino, and Tropicana Atlantic City Corp., DBA Tropicana Casino and Resort, Atlantic City Bethany W. Kristovich Munger Tolles & Olson 350 S Grand Avenue 50 th Floor Los Angeles, CA 90071

Justin P. Raphael Munger Tolles & Olson 560 Mission Street 27 th Floor San Francisco, CA 94105 Counsel for Appellees MGM Resorts International, Marina District Development Company, LLC, DBA Borgata Hotel Casino & Spa Page 5Jennifer L. Del Medico Laura W. Sawyer Jones Day 250 Vesey Street Floor 31 New York, NY 10281

David C. Kiernan Matthew Silveria Jones Day 555 California Street 26 th Floor

San Francisco, CA 94104 Counsel for Appellees Hard Rock International, Inc., Seminole Hard Rock Support Services, LLC

Craig Carpenito David S. Lesser King & Spalding 1290 Avenue of the Americas New York, NY 10104 Counsel for Appellee Boardwalk 1000, LLC, DBA Hard Rock Hotel & Casino Atlantic City

Melissa Arbus Sherry Christopher Brown Lawrence E. Buterman Graham B. Haviland Anna M. Rathbun Latham & Watkins 555 11 th Street NW Suite 1000Page 6 Washington, DC 20004

Sadik Huseny Brendan A. McShane Latham & Watkins 505 Montgomery Street Suite 2000 San Francisco, CA 94111 Counsel for Appellee Cendyn Group, LLC

Thomas J. Sullivan, Jr. Shook Hardy & Bacon 2001 Market Street Two Commerce Square Philadelphia, PA 19103 Counsel for Amicus Appellee International Center for Law & Economics

McKEE, Circuit Judge.

Plaintiffs, casino-hotel guests, appeal the District Court’s dismissal of their Complaint alleging that Defendants, select casino-hotels in Atlantic City and their algorithmic software provider, Cendyn Group, LLC, have conspired to fix prices of hotel rooms in violation of Section 1 of the Sherman Antitrust Act. For the reasons that follow, we will reverse the District Court's dismissal of the Complaint and remand for further proceedings consistent with this opinion.

Page 7Section 1 of the Sherman Antitrust Act is deceptively simple. It states: 'Every contract, combination . . ., or conspiracy, in restraint of trade or commerce . . ., is declared to be illegal.” 1 But every contract imposes some restraint on trade; indeed, that is the very point of entering into a contractual relationship. Accordingly, very early on, the Supreme Court interpreted Section 1 to apply only to “unreasonable” restraints on trade.²

Few would doubt that today’s business world is unrecognizable from the world in 1890 when Congress enacted the Sherman Act. Today, software programs automated with artificial intelligence ('AI') can help businesses optimize operations, respond rapidly to demand fluctuations, reduce transaction costs, and enhance market transparency and efficiency in ways that were unimaginable in 1890.³

It is therefore not surprising that researchers have cautioned that in today’s business environment, AI programs that provide “dynamic pricing” algorithms can also facilitate Page 8anticompetitive behavior.⁴ Historically, successful collusion may have been hindered by gaps in communication and the cost of ensuring compliance. Today, these algorithms have the capacity to bridge any such gaps. 5 This makes it possible for firms to collude in ways that were inconceivable in 1890. Collusion among competitors in today's world need not be characterized by handshakes (or a wink and a nod) in smokefilled rooms or by tell-tale telephone conversations. Instead, AI can enable participants in the marketplace to coordinate pricing and thereby collude in ways that reduce competition at the expense of consumers who bear the brunt of higher prices and reduced output or supply. This is what the Sherman Act is intended to prevent.

This appeal asks us to consider whether Plaintiffs have plausibly alleged that Defendants have colluded to fix the prices of casino-hotel guest rooms in violation of Section 1 of Page 9the Sherman Act.⁶ According to Plaintiffs' Consolidated Amended Complaint ('CAC'), the casino-hotels send their 'current, non-public room pricing and occupancy data' to Cendyn’s Rainmaker software, an AI-powered dynamic pricing program, which then processes each casino-hotel's non-public data, “along with” similar data provided by their competitors, and thereby generates suggested room rates for each of the participating casino-hotels.⁷ Rainmaker, according to Plaintiffs, then functions as a coordinating mechanism, using the collective data to generate anticompetitive prices across participating casino-hotels. Plaintiffs allege that the resulting anticompetitive rates are automatically uploaded into each casino-hotel’s room-selling platform, causing consumers to pay anticompetitively high prices for guest rooms. This practice is allegedly a “stark change from how [casino-hotels] independently priced rooms for years” at “significantly reduced room rates” in order to draw guests to their casinos, since “gambling has always been the main revenue driver and profit center for casino-hotels.” 8 According to the allegations in the CAC, Cendyn’s software allows casino-hotel Defendants to maintain the resulting inflated room rates without fear that potential hotel guests will be induced to book rooms with one of their competitors because the colluding casino-hotels have collectively agreed to charge the Page 10anticompetitively high rates recommended by Cendyn’s Rainmaker software. According to the CAC, Cendyn notes that its clients, including Defendant casino-hotels, do in fact charge the room rate recommended by Rainmaker 90 percent of the time. According to the CAC, the Defendants can thereby charge inflated rates for their hotel rooms without fear that their prices will be undercut by a competitor in the Atlantic City market.

The District Court dismissed the CAC because it concluded that Plaintiffs did not sufficiently plead the existence of a hub-and-spoke price-fixing scheme. The court reasoned that Plaintiffs’ allegations failed to show a “rim,” i.e., an agreement between the casino-hotel defendants. We disagree. We hold that the well-pleaded allegations in the CAC are sufficient to support a finding that casino-hotel Defendants have conspired to fix prices through Cendyn's software. Accordingly, we will reverse the District Court’s order dismissing the Complaint.

As we have already noted, advancements in technology have resulted in software capabilities dubbed “artificial intelligence” or “AI”—program- and machine-based systems comprised of multiple algorithms. 9 An algorithm is a set of

9 Leah Friedman, Nancye Blair Black, Erin Walker & Jeremy

Roschelle, Safe AI In Education Needs You , COMMC’NS OF

THE ACM: BLOG@CACM (Nov. 8, 2021),

https://web.archive.org/web/20211113002115/https://cacm.ac

m.org/blogs/blog-cacm/256657-safe-ai-in-education-needs-

you/fulltext; see also White & Cowart, supra note 5, at 292.

Page 11rules programmed to produce a defined output based on specific inputs.¹⁰ It sometimes functions as a group of step-bystep instructions dictating how to process incoming data. 11 Software programs using AI process data through a series of algorithms in order to make a prediction, solve a problem, interpret conditions, or actuate something, such as activating autopilot control of a system. 12

One feature of highly advanced AI programs is machine learning—a process by which an AI program uses the large amounts of data available to it to continuously learn and improve on previous outputs. 13 'The function of a machine learning system can be descriptive , meaning that the system uses the data to explain what happened; predictive , meaning the system uses the data to predict what will happen; or Page 12prescriptive , meaning the system will use the data to make suggestions about what action to take.” 14 The more voluminous the data inputs, the more refined the program's outputs will be. 15

Dynamic pricing algorithms are prescriptive AI systems that automate pricing for sellers in marketplaces. 16 These algorithmic programs are typically trained on data specific to a vendor or market to suggest (and in some cases implement) prices. 17 They learn through an iterative process of 'trial and error and through finding patterns from a great volume and variety of data.” 18 They may use data 'related to past, present, and future supply and demand conditions,” including data on competitors' public prices, to allow vendors to adjust prices frequently at a lower transaction cost. 19 Although dynamic pricing algorithms vary in settings, parameters, function, and sophistication, they share the common purpose of “optimiz[ing] business processes, [thereby] allowing businesses to gain a competitive advantage by . . . setting Page 13optimal prices that effectively respond to market circumstances.'²⁰ Indeed, these algorithms have become a common feature of many markets, including '[r]ide-sharing apps, airlines, and [major online retailers]” and thus shape the marketplace for millions of Americans every day. 21

It is important to note that “[t]here is nothing inherently wrong [or anticompetitive] with using [algorithms] to engage more effectively in commercial activity, regardless of whether that activity is participation in the financial markets or the selling of goods and services.” 22 Nevertheless, the potential for abuse is obvious. It is therefore not surprising that many leading economists and legal scholars have raised concerns about the potential drawbacks of dynamic pricing algorithms on consumers and market competition, such as increased prices and reduced output. 23 Some suggest that these programs may be able to “facilitate price-fixing and … collusion” and thus must be “treated with skepticism” to ensure that they do not Page 14cross the line established by the Sherman Act.²⁴ Here, Plaintiffs allege that the casino-hotels' use of Cendyn's dynamic pricing algorithm crossed the line, from procompetitive efficiency to anticompetitive collusion, and thereby unreasonably interfered with market competition.

Section 1 of the Sherman Act provides: “Every contract, combination in the form of trust or otherwise, or conspiracy, in restraint of trade or commerce among the several States, or with foreign nations, is declared to be illegal.” 25 As this court has emphasized, “Section 1 only prohibits contracts, combinations, or conspiracies that unreasonably restrain trade.” 26 We have interpreted the terms 'contract, combination, . . . or conspiracy' collectively to require some form of agreement or “concerted action.” 27 In other words, a plaintiff Page 15alleging a violation of Section 1 must plausibly show that defendants had a “unity of purpose,” a “common design and understanding,” a 'meeting of minds,” or a “conscious commitment to a common scheme.'²⁸ A plaintiff may support allegations of a collusive agreement either with direct evidence or circumstantial evidence. Because direct evidence, 'the proverbial 'smoking gun,'' is usually unavailable to those outside the alleged conspiracy, “plaintiffs have been permitted to rely solely on circumstantial evidence (and the reasonable inferences that may be drawn therefrom) to prove a conspiracy.” 29 Indeed, they must be permitted to do so if the law is to reach more sophisticated or surreptitious agreements.

Secondly, a plaintiff must show that this agreement or concerted action “imposed an unreasonable restraint on trade.” 30 Most restraints of trade are analyzed under the 'rule of reason,' which requires comprehensively looking at the market in which the defendant operates and then weighing the alleged anticompetitive effects against the proffered procompetitive benefits. 31 However, a few categories of Page 16restraints are said to be so manifestly anticompetitive that they are “conclusively presumed to unreasonably restrain competition.' 32 These agreements are deemed as a matter of law per se illegal, because they almost always tend to restrict competition and decrease output.³³ 'Paradigmatic examples [of per se illegal restraints] are ‘horizontal agreements among competitors to fix prices or to divide markets.’” 34 Horizontal agreements between competitors form a key part of what has been dubbed a “hub-and-spoke” conspiracy, a form of conspiracy analogized to the shape of a wheel.³⁵ Such a conspiracy involves a hub at the center, usually an agent, supplier, or purchaser with whom all the competitors, the spokes, have a relationship.³⁶ Establishing a per se violation with the hub-and-spoke model requires showing that the spokes are connected to each other via a horizontal agreement.³⁷ This horizontal agreement connecting the competitors is the “rim” of the wheel. 38

to the consumer and restraints stimulating competition that are in the consumer’s best interest.”).

32 In re Flat Glass AntitrusPage 17t Litig. , 385 F.3d at 356 (citation To state a claim of horizontal agreement under the Sherman Act with circumstantial evidence, plaintiffs must plead “something more than merely parallel [conduct]” among the defendants.³⁹ This is because parallel conduct may occur without any coordination and result merely from each competitor independently working to gain a competitive advantage or maximize its market position.⁴⁰ Accordingly, plaintiffs alleging an illegal conspiracy to restrain trade based on circumstantial evidence must also allege that certain “plus factors” exist beyond the mere fact of parallel conduct.⁴¹ We have identified at least three types of “plus factors” that “tend to demonstrate the existence of an agreement: '(1) evidence that the defendant had a motive to enter into a price fixing conspiracy; (2) evidence that the defendant acted contrary to its interests; and (3) evidence implying a traditional conspiracy.’” 42 These are nPage 18ot exhaustive, as many antitrust claims arise in unique contexts and market conditions. This is particularly true given the evolving complexity and sophistication of the technology that has become integral to today’s marketplace.

modified).

“The main threat of horizontal agreements is that they can enable participants collectively to reduce the output of goods in some market, thus causing higher prices, inefficient substitutions, and the resultant losses in consumer or labor welfare.” 43 In a competitive market, the presence of many firms competing for market can make such agreements impractical.⁴⁴ And '[f]irms that seek to [coordinate prices] through the conscious parallelism of oligopoly must rely on uncertain and ambiguous signals to achieve concerted action.” 45 However, AI-enabled algorithms are capable of solving these problems while simultaneously evading traditional detection methods.⁴⁶ AI software can facilitate collusion by enabling competitors to coordinate prices and share information without ever communicating with each other. And real-time price monitoring enables cartels to more effectively police each other's pricing behavior and adjust Page 19accordingly. Our review of Plaintiffs’ Section 1 claims must proceed against this background.

Whether Plaintiffs have stated a claim upon which relief can be granted is a question of law. Accordingly, '[w]e exercise plenary review of the District Court’s order[] granting [D]efendants’ motion to dismiss under Federal Rule of Civil Procedure 12(b)(6).'⁴⁷ To state a claim, 'a complaint must contain factual allegations that, taken as a whole, render the plaintiff's entitlement to relief plausible.'⁴⁸ The plausibility standard at the pleading stage is not a “probability requirement.'⁴⁹ It simply requires the reviewing court to determine whether a complaint which pleads a Section 1 claim includes sufficient factual allegations “to raise a reasonable expectation that discovery will reveal evidence of [an] illegal agreement.' 50 In reviewing the complaint, we take the wellpleaded allegations to be true, and we construe them in the light most favorable to plaintiffs.⁵¹ Twombly recognized that although antitrust cases can be complex and expensive, we Page 20must not require heightened fact pleading.⁵² Rather, we review the well-pleaded allegations 'in light of common economic experience.” 53 This is a 'context-specific task' that requires us to draw on our judicial experience and common sense in assessing the totality of the circumstances. 54

This putative class action is a consolidation of three similar cases with identical price-fixing claims filed before the District Court. Plaintiffs are consumers who directly rented Page 21guest rooms from the Defendant casino-hotels, allegedly at anticompetitively high prices, during a period which began no later than June 28, 2018, to the present (the 'class period'). They allege that, although the casino-hotels started using Rainmaker products 56 at various times preceding the class period (as early as 2004 and as recently as 2018), all of them were using Rainmaker's prescriptive algorithm for their pricing decisions by the start of the class period.⁵⁷ Rainmaker allegedly acts as an exchange hub and a shared pricing agent for the casino-hotels. Plaintiffs allege that the 'products comprising the Rainmaker pricing algorithm platformGuestREV, REVCaster, and GroupREV—enable casino-hotel clients to achieve higher rates and profits on guest rooms” by forecasting market demand for individual and large group bookings and recommending to each client the optimal price Page 22rate to charge.⁵⁸ According to Plaintiffs, the process works like this:

Rainmaker’s software is installed directly into casino-hotels’ on-site room pricing and occupancy data systems, such that each casinohotel client provides its current, non-public room pricing and occupancy data to the Rainmaker platform on a continuous basis. In turn, the algorithm continuously processes and analyzes this non-public, real-time information, along with the same type of non-public, realtime data the client’s participating competitors also submit to the platform, and other relevant supply and demand-related data. The algorithm utilizes this continuous flow of real-time data to obtain a clear and complete picture of market supply and demand and competitive dynamics at any given time. The algorithm ultimately uses this information to generate “optimal” room rates, updated multiple times per day, for each client to charge guests. 59

Since the software is allegedly “integrated directly into a casino-hotel's property management system,' the pricing recommendations generated by the algorithm are “directly and Page 23automatically uploaded into the casino-hotel’s system.” 60 It is undisputed among the parties that the casino-hotels retain final pricing authority and may choose to change the algorithm's suggested prices. However, Plaintiffs allege that Cendyn constrains the casino-hotels' ability to do so by requiring a special override to be used only in times of “need and extreme circumstances” and by scoring each casino-hotel on how often it overrides the algorithm’s price and forecast recommendation.⁶¹ The Defendant casino-hotels are alleged to accept Cendyn's recommendations 90 percent of the time, resulting in the alleged anticompetitive guest room rates in Atlantic City during the class period.

Cendyn allegedly “conveyed [to the casino-hotel Defendants] that uniform adoption would enable their access, under the auspices of a single shared algorithm, to one another's real-time, non-public pricing and occupancy data, [and] that the Rainmaker platform[] . . . would generate significantly higher prices for each [casino-hotel Defendant] than if each one did so independently without use of that platform.” 62

The CAC also purports to describe the structural features of the Atlantic City casino-hotels market and the financial difficulties faced by casino-hotels in this market in the years preceding the class period. Plaintiffs posit that Defendants had the motive to conspire to fix prices and that Rainmaker provided the opportunity to do so. They allege that “[s]tarting in mid-2018 . . . [the] Casino-Hotel Defendants . . .

Page 24collectively had market power and incentive to recoup years of losses.” 63 The CAC includes aggregate statistics from Atlantic City casino-hotel market reports suggesting that during the class period, casino-hotels’ guest room rates and corresponding revenue significantly increased in parallel as room occupancy levels decreased. 64 For example, between 2017 and 2019—before the COVID-19 pandemic temporarily shut down hotels in 2020—casino-hotel Defendants’ collective room occupancy allegedly began to trend downward with an 8 percent decrease in occupancy, while their room revenue began to trend upward with about a 22 percent increase in room revenue. 65 And in 2022, the casino-hotels are alleged to have collectively rented 5 percent fewer rooms but charged 25 percent more for those rooms as compared to data from 2019.⁶⁶ Accordingly, Plaintiffs allege that this parallel conduct during the class period, and other “plus” factors, like the purported exchange of non-public information through Cendyn’s algorithm, constitute circumstantial evidence of a horizontal price-fixing agreement.

Page 25Defendants moved to dismiss the Complaint for failure to state a claim under Federal Rule of Civil Procedure 12(b)(6). The District Court granted the motion to dismiss, reasoning that the CAC contained several factual deficiencies and did not establish a plausible price-fixing agreement among the casinohotel Defendants.⁶⁷ The District Court understood Plaintiffs' allegations of collusion to be premised merely on the casinohotel Defendants' 'knowing' and 'purposeful' use of the Rainmaker products.⁶⁸ It found that the CAC was substantially similar to a complaint from an action in the District of Nevada which was dismissed because its allegations of a horizontal agreement did not 'cross[] the line from conceivable to plausible.” 69 Among many factors the Court determined to be fatally insufficient, it noted that the parallel conduct pled was not parallel because the casino-hotel Defendants signed up to use Rainmaker over a fourteen-year period.⁷⁰ It also found that Plaintiffs did not adequately show how Cendyn used the data it received from the casino-hotels—namely that the CAC did not allege that the information was 'pooled or otherwise commingled” or somehow improperly exchanged —and that the casino-hotel Defendants continued to retain and exercise pricing authority.⁷¹ Accordingly, the District Court concluded Page 26that these factors militated against finding an agreement among the Defendants such that the conspiracy Plaintiffs allege fatally “lacks a rim.” 72 We disagree with the District Court.

The CAC contains many allegations that, taken together as true, lead to a plausible inference that Defendants have agreed to fix their room rates through Cendyn's dynamic pricing algorithm and thereby inflate room rates and avoid competing. Moreover, and quite significantly, Plaintiffs allege that even when it would have been in the casino-hotels' economic interests to reduce room rates to increase occupancy so as to capture more guests and thereby generate more casino revenue when occupancy was declining, the casino-hotels overwhelmingly adhered to Cendyn’s price recommendation. Even more significantly, the CAC alleges that Cendyn's former Vice President of Data Science and Analytics (who also served as Rainmaker’s Vice President of Revenue Analytics) encouraged casino-hotels to “avoid the infamous ‘race to the bottom’ when competition inevitably becomes fierce within a market.'⁷³ Cendyn is also alleged to have 'led discussions involving industry executives and managers, including personnel from [c]asino-[h]otel Defendants, on the best practices for maximizing room revenue and profitability while avoiding price wars , including through use of the Rainmaker platform .” 74 But of course, that is just one way of indicating that Rainmaker could help the casino-hotel Defendants avoid having to lower prices to compete with one another. Thus, the Page 27District Court erred in failing to appreciate or recognize the nature of the conspiracy alleged by Plaintiffs.

Plaintiffs’ Complaint relies primarily on circumstantial evidence to state a claim of collusion. Accordingly, Plaintiffs must sufficiently allege parallel conduct and plus factors.⁷⁵

Plaintiffs' two theories of parallel conduct include allegations of Defendant casino-hotels’ (1) contemporaneous use of Cendyn’s software and (2) synchronous price and output movement during the class period. First, Plaintiffs allege that at some point between 2004 and 2018, each of the casino-hotel Defendants adopted Rainmaker and “knowing[ly] and purposeful[ly]” used it to set prices.⁷⁶ In doing so, the casinohotel Defendants allegedly gave Rainmaker not only information that was generally available to the public but also proprietary information knowing that competitors would benefit from it. Cendyn’s dynamic pricing algorithm therefore functioned as a 'shared pricing agent' for defendants and “generate[d] recommended room rates for each of them using non-public pricing and occupancy data shared by each casinohotel Defendant with the platform in real-time.'⁷⁷ A key Page 28allegation is that during the relevant class period, all of the casino-hotel Defendants were delegating their pricing decisions to Cendyn by accepting Rainmaker’s pricing recommendations 90% of the time. This alleged parallel behavior among casino-hotel Defendants after 2017 purportedly replaced a “historically independent room pricing system” with an “interdependent, collusive one.'⁷⁸ The anticompetitive effects are alleged to be consistently higher room rates despite declining occupancy, and this resulting anticompetitive upward pressure on the price of hotel rooms during the class period is clearly alleged in the CAC. Cendyn purportedly relied on the resulting higher room rates to show how its software product kept revenue high despite the downward pressure that would otherwise have resulted in lowering room rates.

Plaintiffs also point to the alleged synchrony of “pricing and occupancy rate movement [among the casino-hotels] during the class period,” for which the CAC includes numerous data points in support.⁷⁹ Relying on data from the New Jersey Division of Gaming Enforcement, plaintiffs include graphs indicating that after 2017, the year preceding the class period, there is sudden upward trend in casino-hotel Defendants' average daily room rates.⁸⁰ The graphs indicate that by 2019, Page 29room rates had increased by an additional $15 to $60.⁸¹ Meanwhile, occupancy rates for all the casino-hotel Defendants (except for Hard Rock) “meaningfully decreased” during this time.⁸² The graphs purport to show that this downward trend in room occupancy began even before the outlier year of 2020 when the COVID-19 pandemic and subsequent lockdown forced the temporary closure of many businesses.⁸³ According to Plaintiffs, 'economic principles should have compelled at least some Casino-Hotel Defendants to drop room rates in order to compete,' especially in the casino-hotel business where greater occupancy is likely to result in greater overall revenue from the additional guests who would patronize the casino and other entertainment facilities.⁸⁴ However, according to the CAC, casino-hotel defendants did not reduce room pricing.

The alleged contemporaneous use of Cendyn’s software and the synchrony of the casino-hotels’ alleged upward pricing and downward room occupancy—a large departure from their traditional practice and under economic circumstances that should typically lead to competitive price-cutting—are sufficient to show conscious parallel conduct.⁸⁵

Page 30Defendants ask us to find otherwise, partly because of allegations that the casino-hotels adopted the software programs at various points within a fourteen-year period. They also point to the fact that the alleged period of collusion here is “over 50 times longer than the three-month gap in Burtch [ v. Milberg Factors, Inc. , 662 F.3d 212, 221 (3d Cir. 2011)],' which we held to be insufficiently parallel.⁸⁶ In Burtch , we found that business decisions made by defendant-financiers at various points between March 13, 2002, and September 15, 2003, to allegedly end their business relationship with the plaintiff was not sufficiently parallel to sustain a claim of conspiracy.⁸⁷ The conduct in question was the financiers' decisions regarding the plaintiff’s requests for a credit line. The Burtch complaint alleged that, during the relevant period, some defendants declined those credit requests altogether, others decreased the existing credit line, and others even increased credit to the plaintiff.⁸⁸ Thus, each of the defendants there took very different approaches toward the plaintiff at various points during the relevant time period.⁸⁹

Burtch does not support Defendants' argument, because here, the alleged relevant conduct is (1) the casino-hotels' 'continu[ous] deployment' of Cendyn's pricing agent and continuous delegation of pricing to the software, and (2) the subsequent, sudden synchronous movement of prices and output (i.e. hotel room occupancy).⁹⁰ The CAC does not allege that the conspiracy began with the adoption of Cendyn's Page 31software, but instead relies upon the continuous parallel conduct of the casino-hotels during the relevant period of the alleged conspiracy.⁹¹ Nonetheless, because parallel conduct alone cannot sustain a claim of horizontal price-fixing, we also consider Plaintiffs’ allegations of additional “plus” factors. 92

Plaintiffs allege several other 'plus' factors which further support an inference of collusion. In the CAC, Plaintiffs allege (1) that Defendants had motive to conspire because ofPage 32 “an extended period of financial hardship in the years [preceding] the class period;” (2) that the casino-hotel Defendants' adoption of Cendyn's price recommendations when they could have competed more aggressively on price was against their economic interests; and (3) non-economic evidence showing exchange of non-public commercial information, opportunities to collude, and sudden changes in longstanding business practices.⁹³

Plaintiffs allege that Defendants’ incentive to conspire stems from the financial hardship the casino-hotel market experienced due to high debt levels and depressed cash flows for many years following the 2008 recession. The CAC suggests that the structural features of the Atlantic City casinohotel market-high barriers to entry, lack of reasonable substitutes, and the high market concentration—are conducive to collusion and contributed to the casino-hotels’ motive.

The CAC also alleges that the failure of at least some Defendants to undercut each other's prices during the class period was against their individual economic self-interest in the absence of collusion.

In Lifewatch Services Inc. v. Highmark , we noted that when the market in question is alleged to be an oligopoly, that is, a highly concentrated market comprised of few dominant firms with fungible products, the first two categories of “plus” factors, motive to conspire and actions against self-interest, may simply be attributed to market transparency and not Page 33necessarily to collusive conduct.⁹⁴ This is because it is easy for competitors to monitor and be influenced by each other's behavior in oligopolistic markets. 95 Sellers in such markets might be hesitant to undercut each other’s prices as competitors can easily notice and reciprocate by reducing their own prices. 96 Thus, we said in Valspar Corporation v. E.I. Du Pont Page 34De Nemours and Company , that evidence that “the market was primed for anticompetitive interdependence” may do nothing more than describe the market characteristics in some cases. 97 However, “certain types of ‘actions against self[-]interest’ may do more than restate economic interdependence' even in alleged oligopolistic markets. 98 And here, more is alleged.

In this unique context of casino-hotels, the allegation that the casino-hotel Defendants agreed not to compete on room rates is significant. As we have discussed, if hotel rooms allegedly draw more guests into the casino facilities, casinohotels “have even more incentive than [standard] hotels to fill their hotels to capacity” as they may obtain “more income from the rental of their rooms . . . [and a] greater amount of revenue from their casinos.” 99 Taken as true, economic principles state that a casino-hotel whose room occupancy is steadily decreasing over the years would lower room rates in order to compete for more hotel guests who will then be available to venture into the casino. Moreover, common sense suggests as much. 100 But, this might spur competition and create Page 35downward price pressure thus resulting in the very “race to the bottom” that Cendyn’s former executive allegedly cautioned hotels to avoid. 101 Accordingly, the alleged scheme would only work if casino-hotels could maintain higher room prices knowing that other casino-hotels would not reduce their rates to compete for the pool of potential hotel guests. This certainly supports an inference of collusion; indeed, we can think of no other explanation.

Plaintiffs also allege evidence of a traditional conspiracy, which we have described as 'non-economic evidence that there was an actual, manifest agreement not to compete” such as “proof that the defendants got together and exchanged assurances of common action or otherwise adopted a common plan even though no meetings, conversations, or Page 36exchanged documents are shown.'¹⁰² Plaintiffs allege, among other things, de facto data exchanges through the software, opportunities for the Defendants to conspire at various industry events, knowledge of each other's relationship with Cendyn through publications and events, and the sudden change in the casino-hotels’ business practices.

The CAC alleges that the casino-hotel Defendants mutually used the Cendyn pricing software as a means to exchange non-public proprietary information and that these information exchanges directly impacted their pricing decisions. Of course, information exchanges among competitors are not per se illegal. They are, however, 'a facilitating practice that can help support an inference of a price-fixing agreement.” 103

The District Court held that 'Plaintiffs' 'failure to plausibly allege the exchange of confidential information from one of the spokes to the other through the hub’s algorithm is [a] fatal defect . . . [and it] compels the conclusion that there is Page 37no rim [in the alleged hub-and-spoke conspiracy].''¹⁰⁴ But, there are numerous paragraphs in the CAC that plausibly make such allegations in the context of AI-driven dynamic pricing.¹⁰⁵ The CAC alleges that each casino-hotel knew and 'understood' that 'the recommended room rates they were receiving from [Cendyn’s software] were based on real-time, non-public pricing and occupancy data [that] they and their codefendants all were providing to the platform.” 106 It is further alleged that they 'understood that their co-defendants also knew that the recommended room rates they each were receiving from [Cendyn’s software] were based on real-time, non-public pricing and occupancy data [that] they and their codefendants were all providing to the platform.” 107 The CAC adequately alleges that the casino-hotels understood that each of them was committed to a common plan of setting room rates based on the recommended rates received from Cendyn's software, the hub for their collective data, while also “knowing that their competitors would not lower their room rates to take market share.” 108

The District Court required plaintiffs to plead with more specificity how the algorithm functions to facilitate the exchange of information.¹⁰⁹ This is tantamount to expecting Plaintiffs to explain how Cendyn’s proprietary software works Page 38without affording the discovery required to do that. Moreover, at this stage of litigation, such level of detail into the software’s operations is neither required nor appropriate.¹¹⁰ Thus, at this stage, Plaintiffs have sufficiently alleged that a material exchange is likely occurring and facilitating collusion.

Defendants also argue that because each casino-hotel retained final pricing authority and could override Cendyn’s price recommendation, Plaintiffs have not sufficiently shown that the casino-hotels enforced their agreement by delegating their respective pricing decisions to Cendyn. However, “[p]rices are fixed when they are agreed upon,” irrespective of whether conspirators always adhere to them. 111 Although an antitrust plaintiff's allegations may well fail if the alleged colluders routinely varied from recommended prices, that is not what is alleged here. The alleged 90 percent compliance rate certainly supports an inference of an agreement between casino-hotel defendants to price rooms consistent with Cendyn’s suggestions. This is especially true given the alleged practical difficulties of deviating from Cendyn’s rates, such as requiring 'override permissions' accessible to select staff at each casino-hotel. 112 Moreover, as we have explained, dynamic pricing algorithms tend to rely on the collective input Page 39of data to draw patterns and make suggestions. This, coupled with the well-pleaded allegations in the Complaint, further supports an inference that information exchanges have resulted in collusive pricing and diminished competition.

To be clear, we do not make any assumptions or conclusions about how Cendyn's dynamic pricing software works. We only conclude that, taking Plaintiffs’ allegations as true, the software is, in effect, facilitating collusive conduct by receiving from each client non-public commercial information, and in return, giving each client the benefit of their competitors’ non-public data in formulating a price recommendation which Defendants purportedly agreed to comply with.¹¹³ Such exchanges involving current pricing and Page 40occupancy data have great potential for anticompetitive collusion. 114 While there are independent business reasons to use dynamic pricing software, there are rarely legitimate business justifications for affording competitors the benefit of commercially sensitive proprietary information under the circumstances alleged here.

In conclusion, we find that the District Court gave inadequate consideration to the allegations in the CAC in context with the complexity and novelty of dynamic pricing algorithms. Nevertheless, the point made by amicus curiae International Center for Law & Economics (ICLE) cautioning us to avoid criminalizing industry-wide use of the same algorithmic software is well taken. They ask, “[i]f multiple gas stations use Excel spreadsheets with the same pricing formulas, is that an antitrust violation? If retailers use the same market research firm’s pricing surveys, have they joined a hub-and- Page 41spoke conspiracy? If manufacturers rely on the same forecasting software to set production levels, are they unlawfully coordinating output?'¹¹⁵ But these questions oversimplify the issues and ignore many of the specific allegations in the CAC that are unique to this litigation. The allegations here imply more than using identical spreadsheets or pricing formulas without colluding on the prices that will be charged as a result.

The CAC involves many more factors that go beyond merely using the same independently-operated software to set production levels. For example, it alleges an exchange of nonpublic commercial information from the spokes that is consolidated in the hub and then shared back to the spokes in the form of price recommendations which automatically determine room rates under circumstances where each spoke is confident that the resulting rate will not be undercut by the competing spokes. Far from ICLE's proffered examples of software programs used separately and independently to help businesses compete against one another, Rainmaker is alleged to operate as a single decision-maker or hub, coordinating pricing for a majority of the market.

Some researchers also caution that existing antitrust jurisprudence may not be fully equipped to tackle future developments in technology and the new issues that may subsequently arise. 116 We do not disagree. As we noted at the Page 42outset, Section 1 of the Sherman Act was enacted well over a century ago in 1890. Nevertheless, the goal of our antitrust jurisprudence remains the same: to ensure the continued existence of “independent centers of decision-making.” 117

As former acting chair of the U.S. Federal Trade Commission, Maureen K. Ohlhausen, inquired:

“Is it ok for a guy named Bob to collect confidential price strategy information from all the participants in a market, and then tell everybody how they should price? If it isn’t ok for a guy named Bob to do it, then it probably isn’t ok for an algorithm to do it either.” 118

If Cendyn’s algorithm is in effect collecting non-public commercial information from defendants and utilizing the collective pot of data to 'suggest' prices to each, under the circumstances alleged here, plaintiffs have surely raised a plausible inference of collusion under Section 1 of the Sherman Act even though the alleged hub is named “Rainmaker” rather Page 43than 'Bob.'¹¹⁹ Accordingly, we will reverse and remand for further proceedings in accordance with this Opinion. 120