AMI - GOVERNMENT EMPLOYEES PROVIDENT FUND MANAGEMENT COMPANY LTD., et al. v. ALPHABET INC., et al.

Opinion

UNITED STATES DISTRICT COURT NORTHERN DISTRICT OF CALIFORNIA

AMI - GOVERNMENT EMPLOYEES Case No. 23-cv-01186-RFL PROVIDENT FUND MANAGEMENT COMPANY LTD., et al., ORDER GRANTING MOTION FOR Plaintiffs, CLASS CERTIFICATION v. Re: Dkt. Nos. 134, 149

ALPHABET INC., et al., Defendants.

This is a securities case against Alphabet Inc., its subsidiary Google LLC, and Google CEO Sundar Pichai. Plaintiffs allege Defendants misrepresented that Google’s digital advertising auctions were agnostic to the channel by which a bid was received. According to Plaintiffs, Google’s advertising auctions actually favored bids submitted by the Facebook Advertising Network. Plaintiffs allege that Defendants’ misrepresentation artificially maintained Alphabet’s stock price, thereby violating Sections 10(b) and 20(a) of the Securities Exchange Act of 1934, and Rule 10b-5, 17 C.F.R. § 240.10b-5. Plaintiffs now move to certify a class of investors. After a minor tweak to the class definition, Plaintiffs’ motion is GRANTED.1 I. BACKGROUND A detailed description of Plaintiffs’ allegations is contained in the order on Defendants’ motion to dismiss. AMI - Gov’t Emps. Provident Fund Mgmt. Co. Ltd. v. Alphabet Inc., No. 23-

1 Defendants’ motion for leave to file a surreply is GRANTED. Plaintiffs appropriately raised additional information in their reply brief as to the issue of price impact to which Defendants were entitled to respond. CV-01186-RFL, 2025 WL 899959, at *2–3 (N.D. Cal. Mar. 24, 2025). To briefly summarize: Alphabet and its subsidiary Google (collectively, “Google”) are in the digital advertising business. (Dkt. No. 87 (“SAC”) ¶¶ 51–57.) Online media companies, alternatively called publishers, use ad servers to sell their ad space. (Id. ¶¶ 39, 44.) Google operates an ad server. (Id. ¶ 44.) Ad servers can solicit bids to fill ad space through ad exchanges and ad networks. (Id. ¶¶ 45, 48.) Facebook operates the Facebook Advertising Network (“FAN”), an ad network. (Id. ¶ 95.) According to Plaintiffs, Google caused its ad server to give FAN preferential treatment, so bids submitted through FAN had a higher probability of winning. (Id. ¶¶ 99–106.) Plaintiffs contend that Defendants misrepresented that practice, thereby artificially maintaining Alphabet’s stock price. Only one of Defendants’ statements has been found to give rise to a viable claim in this case. AMI, 2025 WL 899959, at *1. In response to Questions for the Record from the House Judiciary Committee, Pichai submitted an answer on September 14, 2020 that “[t]he channel through which a bid is received does not otherwise affect the determination of the winning bidder.” (SAC ¶¶ 165–66.) Plaintiffs say that statement was false because bids submitted through FAN were more likely to win. (Id. ¶ 167.) Plaintiffs contend that six corrective disclosures revealed the falsity of Pichai’s statement. First, on December 16, 2020, Texas and nine other states filed an antitrust lawsuit against Google concerning its advertising technology business. (Id. ¶ 260.) The lawsuit alleged that Google formed an agreement with Facebook to give it “information, speed, and other advantages” in mobile app advertising inventory auctions. (Dkt. No. 144-1 ¶ 14.) Second, on September 1, 2021, the news reported that the Department of Justice (“DOJ”) was readying an antitrust lawsuit against Google concerning its advertising technology business. (SAC ¶ 263.) Third, on October 22, 2021, an unredacted copy of the Texas lawsuit was filed. (Id. ¶ 266.) Fourth, on February 11, 2022, the European Publishers Council filed an antitrust complaint with the European Commission against Google. (Id. ¶ 269.) Fifth, on January 24, 2023, the Department of Justice and eight states filed an antitrust lawsuit against Google concerning its advertising technology business. (Id. ¶ 271.) Finally, on April 17, 2023, nine more states joined the Department of Justice lawsuit. (Id. ¶ 275.) Plaintiffs initially proposed to certify a class of persons or entities who acquired Alphabet’s stock between September 14, 2020 and January 23, 2023. (Dkt. No. 134 at 7–8.)2 At oral argument, Plaintiffs acknowledged that since the alleged misrepresentation was not released until after the close of trading on September 14, 2020, the class period should start on September 15, 2020. In connection with this motion, Plaintiffs submitted two expert reports from Dr. Zachary Nye, a financial economist at Stanford Consulting Group. (Dkt. No. 134-2 (“Nye Report”) ¶ 1; Dkt. No. 148-2 (“Nye Reply Report”).) In opposition, Defendants submitted two expert reports from Professor Allen Ferrell, an economist and law professor at Harvard University. (Dkt. No. 146-1 (“Ferrell Report”) ¶ 1; Dkt. No. 150-3 (“Ferrell Surreply Report”).) II. LEGAL STANDARD Class certification requires a plaintiff to show, by a preponderance of the evidence, that the four prerequisites of Federal Rule of Civil Procedure 23(a) and at least one of the bases for certification under Rule 23(b) are met. Wal-Mart Stores, Inc. v. Dukes, 564 U.S. 338, 350 (2011); Olean Wholesale Grocery Coop., Inc. v. Bumble Bee Foods LLC, 31 F.4th 651, 665 (9th Cir. 2022) (en banc). A class can be certified only if the court concludes, after a “rigorous analysis,” that these requirements have been satisfied. Wal-Mart, 564 U.S. at 351; Comcast Corp. v. Behrend, 569 U.S. 27, 34 (2013). The Rule 23(a) prerequisites are commonly referred to as “numerosity,” “commonality,” “typicality,” and “adequacy.” That means: “(1) the class is so numerous that joinder of all members is impracticable; (2) there are questions of law or fact common to the class; (3) the claims or defenses of the representative parties are typical of the claims or defenses of the class; and (4) the representative parties will fairly and adequately protect the interests of the class.” Fed. R. Civ. P. 23(a). If those prerequisites are satisfied, the analysis turns to the three Rule 23(b) bases for certification. Here, Plaintiffs invoke Rule 23(b)(3). Rule 23(b)(3) requires plaintiffs to prove the

2 All citations to page numbers in filings on the docket refer to ECF pagination. elements of predominance and superiority, such that “questions of law or fact common to class members predominate over any questions affecting only individual members, and . . . a class action is superior to other available methods for fairly and efficiently adjudicating the controversy.” Fed. R. Civ. P. 23(b)(3). Predominance requires only “a showing that questions common to the class predominate, not that those questions will be answered, on the merits, in favor of the class.” Amgen Inc. v. Conn. Ret. Plans & Tr. Funds, 568 U.S. 455, 459 (2013). A common question is one where “the same evidence will suffice for each member to make a prima facie showing [or] the issue is susceptible to generalized, class-wide proof.” Tyson Foods, Inc. v. Bouaphakeo, 577 U.S. 442, 453 (2016) (citation omitted). III. DISCUSSION As Defendants do not contest, Plaintiffs satisfy the Rule 23(a) prerequisites of numerosity, commonality, typicality, and adequacy. First, since Alphabet has “millions of shares trading on a national exchange,” numerosity can be inferred. (See In re Twitter Inc. Sec. Litig., 326 F.R.D. 619, 626 (N.D. Cal. 2018) (citation omitted); Nye Report ¶ 28.) Second, this case will involve several common questions of fact and law, including whether Defendants’ statement was false or misleading, made with scienter, and artificially maintained Alphabet’s stock price. Third, Plaintiffs are typical because, like all other class members, they purchased Alphabet stock at allegedly artificially inflated prices. Finally, Plaintiffs have no apparent conflicts of interest, have been actively involved in this litigation, and are represented by counsel with significant securities litigation experience. Plaintiffs also satisfy the Rule 23(b)(3) requirement of superiority, as Defendants do not contest. See In re Twitter, 326 F.R.D. at 631. The only remaining requirement is predominance, which the parties dispute. Defendants contend that individualized questions of reliance and damages will predominate over common questions. A. Reliance Requiring a plaintiff to directly prove reliance “would place an unnecessarily unrealistic evidentiary burden on the Rule 10b–5 plaintiff who has traded on an impersonal market.” Basic Inc. v. Levinson, 485 U.S. 224, 245 (1988). Thus, if stock trades in an efficient market, an investor is rebuttably presumed to have traded in reliance on public material misrepresentations. Id. at 246–48. To invoke the Basic presumption, a plaintiff must prove: “(1) that the alleged misrepresentation was publicly known; (2) that it was material; (3) that the stock traded in an efficient market; and (4) that the plaintiff traded the stock between the time the misrepresentation was made and when the truth was revealed.” Goldman Sachs Grp., Inc. v. Ark. Tchr. Ret. Sys., 594 U.S. 113, 118 (2021) (citation omitted). If they do so, the defendant can rebut that presumption through “[a]ny showing that severs the link between the alleged misrepresentation and either the price received (or paid) by the plaintiff, or his decision to trade at a fair market price.” Id. (quoting Basic, 485 U.S. at 248). Courts must be “open to all probative evidence on that question—qualitative as well as quantitative—aided by a good dose of common sense.” Id. at 122 (citation omitted). For instance, when there is a “mismatch” between the misrepresentation and corrective disclosure, like a generic misrepresentation but specific corrective disclosure, the inference that “back-end price drop equals front-end inflation . . . starts to break down.” Id. at 123. Ultimately, defendants “bear the burden of persuasion to prove a lack of price impact by a preponderance of the evidence.” Id. at 117. Plaintiffs have undisputedly satisfied their burden to trigger the Basic presumption of reliance. The alleged misrepresentation was publicly posted on Congress’ website. (Ferrell Report ¶ 12.) Alphabet’s stock trades on the NASDAQ, a “quintessentially efficient market.” See Junge v. Geron Corp., No. 20-CV-00547-WHA, 2022 WL 1002446, at *4 (N.D. Cal. Apr. 2, 2022). Plaintiffs and class members purchased Alphabet stock between the time the alleged misrepresentation was made and the time the truth was revealed. And at this stage, Plaintiffs are not required to prove materiality. See Amgen, 568 U.S. at 470. The burden therefore shifts to Defendants to rebut the Basic presumption by showing a lack of any price impact from the alleged misrepresentation. Defendants, however, fail to disprove price impact from at least two corrective disclosures. 1. Statistical Evidence As an initial matter, the parties dispute what statistical evidence is sufficient to prove price impact, or lack thereof. The 95% confidence level is indisputably the “conventional[]” level used to classify a result as statistically significant. (Ferrell Report ¶ 38; see also Nye Reply Report ¶ 34.) But academics (including Ferrell) routinely note results that would be significant at a 90% confidence level. (Nye Reply Report ¶¶ 33–34, 36.) And it is sometimes appropriate to use sub-95% confidence levels to determine significance. (Id. ¶ 34 (quoting the Reference Manual on Scientific Evidence as stating that “[o]ne might argue . . . when regression analysis is used in a price-fixing antitrust case to test a relatively specific alternative to the null hypothesis (e.g., price fixing), a somewhat lower level of confidence (a higher level of significance, such as 10%) might be appropriate”).) Scholars note that if the fraud is relatively small in comparison to the stock’s total market value, or if the stock has high price volatility, a confidence level of 95% can lead to “a high frequency of . . . finding there was no effect of an alleged corrective disclosure that really did reduce firm share price.” Jill E. Fisch & Jonah B. Gelbach, Power and Statistical Significance in Securities Fraud Litigation, 11 Harv. Bus. L. Rev. 55, 67, 82–84 (2021) (cited in Defendants’ Opp., Dkt. No. 142 at 20). Thus, statistical evidence can sometimes be probative of a price impact even if it is not significant at a 95% confidence level. Borderline statistical results are still relevant to determining if there is a price impact, and courts must be open to “to all probative evidence on that question.” See Goldman, 594 U.S. at 122. There is no good reason to treat the 95% confidence level as a magic threshold, such that a p-value of 0.0499 is conclusive quantitative evidence of a price impact and a p-value of 0.0501 is conclusive quantitative evidence of the opposite. See In re Tesla, Inc. Sec. Litig., No. 18-CV-04865-EMC, 2022 WL 7374936, at *13 (N.D. Cal. Oct. 13, 2022) (collecting cases denying Daubert motions); In re Chi. Bridge & Iron Co. N.V. Sec. Litig., No. 17-CV-01580, 2019 WL 5287980, at *13 & n.91 (S.D.N.Y. Oct. 18, 2019) (“A conclusion does not immediately become ‘true’ on one side of the divide and ‘false’ on the other . . . .”). To be sure, a result significant at a slightly lower confidence level is “obviously less comfort than a result that is statistically significant at a confidence level of 95%.” Pirnik v. Fiat Chrysler Autos., N.V., 327 F.R.D. 38, 46 (S.D.N.Y. 2018). But that goes to the result’s weight in favor of a price impact, not whether it is evidence of a price impact at all. That does not mean that statistical evidence can never be evidence of a lack of price impact. For instance, a “finding” of price impact at a 50% confidence level could provide evidence of a lack of price impact. If half of such stock drops constitute an abnormal return but the other half would not, that could provide some indication of no price impact. (Cf. Ferrell Surreply Report ¶¶ 15, 19–21 (noting that “event study methodology has been used to draw conclusions of ‘no effect’ from lack of statistical significance”).) Other intermediate results might similarly provide evidence of a lack of price impact. See Fisch & Gelbach, supra, at 104– 09 (discussing potential considerations). But the precise contours need not be fleshed out in this case. In sum, though weaker evidence, borderline statistical evidence of a price impact can still be evidence of a price impact. With that in mind, the analysis turns to the corrective disclosures. 2. 2020 Texas Lawsuit Starting with the Texas lawsuit, Defendants have not shown a complete lack of price impact. The lawsuit was filed on December 16, 2020, at 3:35 p.m. (Nye Reply Report ¶ 45 n.124.) The statute of limitations does not bar considering this as a corrective disclosure since it only partially disclosed the facts needed to discover a securities violation. See Merck & Co. v. Reynolds, 559 U.S. 633, 653 (2010). There is also no mismatch, as it is closely related to the alleged misrepresentation. For instance, the lawsuit includes specific allegations about how Google advantaged bids submitted through Facebook in its advertising auctions, including that Google gave Facebook “information, speed, and other advantages” in its auctions. (Dkt. No. 144-1 ¶ 14.) That information is a partial disclosure that bidders using Google’s favored channels were more likely to win. Defendants seemingly agree, as they do not contend that there is a mismatch between this corrective disclosure and the alleged misrepresentation. The parties’ statistical evidence weighs in favor of finding a price impact. Nye’s model, using a 24-hour intraday event window (that is, for the 24 hours following the filing of the Texas lawsuit), found an abnormal price drop with a confidence level of 94.92% and 93.49% for Alphabet’s two classes of stock, respectively. (Nye Reply Report ¶ 48.) Using a two-day event window, Nye’s model shows an abnormal price drop at the 97.13% and 96.47% confidence levels, respectively. (Id. ¶ 46.) Ferrell’s model—evaluating either December 16 (with only 25 minutes of relevant trading) or only December 17 (excluding 25 minutes of relevant trading)— found no statistically significant impact at the 95% confidence level from this disclosure. (Ferrell Report ¶¶ 61–63.) The model demonstrates a similar result when considering December 16 at 3:34 p.m. through open on December 17. (Id. at 88.) Nye’s 24-hour intraday analysis is the most persuasive.3 The lawsuit was filed late in the trading day. (See Nye Reply Report ¶ 45 n.124; Monroe Cnty. Emps.’ Ret. Sys. v. S. Co., 332 F.R.D. 370, 392 (N.D. Ga. 2019) (approving a two-day event window when information was released late in the trading day).) And the disclosure’s complexity in the form of a hundred-page legal complaint suggests that market participants would take longer to digest its information than an earnings announcement coming out on an expected date. (See Nye Reply Report ¶¶ 24–30.) That is particularly so because understanding the full impact of the disclosure for purposes of Google’s antitrust liability requires some specialized knowledge from legal professionals. Moreover, while market participants may have expected some form of ad tech antitrust lawsuit, there is no evidence that they expected the disclosure of the Facebook agreement or the facts concerning Google’s alleged manipulation of the auctions. (Cf. Ferrell Surreply Report ¶ 34.) Ferrell’s approach of either limiting the event window to less than half an hour or excluding 25 minutes of relevant trading right after the announcement is less persuasive in this context. (Cf. id. ¶ 29 n.60.) Nevertheless, it is not clear that a two-day event window is necessary, since that would include nearly an entire day of irrelevant trading on December 16 while adding less than a

3 It is reasonable to exclude earnings days from the model, as by definition, earnings days are not typical days with unknown random price variation but instead are expected to have strong price volatility. (See Nye Reply Report ¶¶ 17–21.) half hour of relevant trading on December 17. Though the intraday price drop from this disclosure was not statistically significant at the 95% confidence level, it reflects an abnormal price drop at confidence levels of 94.92% and 93.49%, which still provides some evidence of a price impact. (See Nye Reply Report ¶ 48.) Furthermore, there appears to be a strong connection between the disclosure’s correction of the alleged misrepresentation in this case and the price impact. Analysts and news outlets extensively reported on the lawsuit, focusing on the alleged auction manipulation scheme with Facebook. (Id. ¶¶ 55–59.) For example, reports flagged allegations that Facebook received “special treatment” when using Google’s ad auctions pursuant to its agreement and that Google used its tools “to manipulate” the auction process to “favor itself” over “rivals.” (Id.) And the record does not indicate that there was other news about Alphabet at the time that would explain the price drop. (Id. ¶ 50; Ferrell Report ¶ 60 n.111 (noting only that the lawsuit itself contained allegations beyond auction manipulation).) At oral argument, Defendants contended that a December 17 Colorado antitrust lawsuit against Google was confounding information that severed the link between the misrepresentation and price decline. (See Ferrell Report ¶ 17.) That argument was never raised in Defendants’ papers, so it has been waived. Even if it was considered on the merits, the Colorado lawsuit would not provide sufficient evidence to sever the link. First, it was filed after the close of trading on December 17, which is after the 24-hour intraday window ended. (See id.) Though a news source reported some of its allegations two hours before market close, the extent of that reporting is not clear. (See id. ¶ 17 n.32.) Presumably because of that, Ferrell treated December 18 as the “reaction date” to the Colorado lawsuit. (Id. ¶¶ 47 n.95, 52.) Second, the antitrust lawsuit was reported as concerning both Google’s “search and advertising businesses,” which suggests it is not confounding after all but instead relates to similar allegations. (Id. ¶ 17 n.32. (emphasis added).) Finally, unlike with other potentially confounding information, Ferrell did not attempt to test whether it explained any of the stock price decline. (See id. ¶¶ 74–79.) Taken as a whole, Defendants have not carried their burden to disprove price impact from the 2020 Texas lawsuit. 3. 2023 DOJ Lawsuit It is a similar story for the January 24, 2023 DOJ lawsuit. To start, there is no mismatch because the lawsuit disclosed new information to the market, even assuming its allegations completely overlapped with those in the Texas lawsuit. First, a truth-on-the-market defense requires information to be available “with a degree of intensity and credibility sufficient to effectively counter-balance any misleading impression.” In re Apple Computer Sec. Litig., 886 F.2d 1109, 1116 (9th Cir. 1989) (citing Basic, 485 U.S. at 249). For that reason, when a statement reveals the truth about a matter after an investigation is announced, both the statement and the announcement are treated as partial corrective disclosures. See Lloyd v. CVB Fin. Corp., 811 F.3d 1200, 1210 (9th Cir. 2016). Similarly, even some information “nominally available to the public” is still new “if the market has not previously understood its significance.” In re BofI Holding, Inc. Sec. Litig., 977 F.3d 781, 794 (9th Cir. 2020). After both the DOJ under the Biden administration and a group of Republican-led states filed substantially similar lawsuits against Google, there was a bipartisan consensus that Google rigged its advertising auctions in favor of Facebook. That provided additional information to the market about the strength of those allegations. Prior articles generically reporting that the DOJ was “[r]eadying” such a lawsuit did not fully reveal that information. (See Dkt. No. 145-10 at 2; Lloyd, 811 F.3d at 1210.) Second, an event can be corrective “when it merely discloses a consequence of concealed information (for example, an earnings miss), even if the connection between the cause and the consequence is not made explicitly in the disclosure.” Jaeger v. Zillow Grp., Inc., 746 F. Supp. 3d 1025, 1036 (W.D. Wash. 2024) (citing Mineworkers’ Pension Scheme v. First Solar Inc., 881 F.3d 750, 753–54 (9th Cir. 2018)), aff’d, No. 24-6605, 2025 WL 2741642 (9th Cir. Sept. 26, 2025). When the DOJ filed suit, market participants learned that a break-up remedy was significantly more likely. Plaintiffs’ theory is that investors would have already understood the level of this risk if Google had been forthright that bidders using certain channels were more likely to win bids. Accordingly, when the lawsuit revealed a break-up remedy was on the table, it disclosed a consequence of the alleged misrepresentation in this case. Additionally, Defendants have not shown other evidence of a lack of price impact from this disclosure. Nye’s model, using a 24-hour intraday event window, shows a statistically significant price drop when using the 95% confidence level. (Nye Reply Report ¶ 92.) The lawsuit was filed in the middle of the trading day at 12:18 p.m., so it is reasonable to use an intraday event window to fully capture the price impact, for the same reasons as applied to the Texas lawsuit. (See id. ¶¶ 24–30, 92.) Even Ferrell’s model with a one-day close-to-close event window, if earnings dates are excluded, shows an abnormal price drop at a confidence level slightly lower than 95%. (See Ferrell Surreply Report ¶ 26.) Like the Texas lawsuit, analysts and news outlets extensively covered this complaint, and the record does not indicate other news about Alphabet that would explain the price movement. (Nye Reply Report ¶¶ 93, 97–98.) Accordingly, Defendants have not met their burden to show a lack of any price impact from this corrective disclosure. * * * Defendants have not disproven back-end price impact from the first and last corrective disclosures within the class period. And those two incidents of back-end price impact support Plaintiffs’ theory that the alleged misrepresentation maintained the initial stock price at the front end. Some of the other corrective disclosures are similarly situated, while others have weaker evidence of a price impact. But it is unnecessary to disentangle those disclosures at the class certification stage, given the back-end impact covering both ends of the class period. See Pardi v. Tricida, Inc., No. 21-CV-00076-HSG, 2024 WL 4336627, at *8 & n.6 (N.D. Cal. Sept. 27, 2024). Instead, those questions of loss causation and damages are properly reserved for the merits phase. B. Damages To gain certification of a Rule 23(b)(3) class, a plaintiff must generally show that “damages are susceptible of measurement across the entire class.” Comcast, 569 U.S. at 35. “Calculations need not be exact” at this stage. Id. Instead, “plaintiffs must show that they will be able to prove their case through common proof at trial.” Lytle v. Nutramax Lab’ys, Inc., 114 F.4th 1011, 1024 (9th Cir. 2024), cert. denied, 145 S. Ct. 1308 (2025). So “plaintiffs may rely on an unexecuted damages model to demonstrate that damages are susceptible to common proof so long as the district court finds, by a preponderance of the evidence, that the model will be able to reliably calculate damages in a manner common to the class at trial.” Id. Plaintiffs have made such a showing. Nye proposes using the “out-of-pocket” method, which uses an event study to calculate classwide damages. (Nye Report ¶¶ 68–73.) This is the “standard measurement of damages in Section 10(b) securities cases.” City of Mia. Gen. Emps.’ & Sanitation Emps.’ Ret. Tr. v. RH, Inc., No. 17-CV-00554-YGR, 2018 WL 4931543, at *3 (N.D. Cal. Oct. 11, 2018) (citations omitted). It “isolate[s] Company-specific price movement caused by the revelation of true facts related to the alleged fraud from price movement caused by other factors.” (Nye Report ¶ 70.) If any confounding information was released on the same day as the corrective disclosures, “it is widely recognized that an event study can be fashioned . . . to isolate the effects of such information.” (Nye Reply Report ¶ 108.) After calculating the total price inflation caused by the misrepresentation, “a Class member’s actual trading activity in the security can be used to mechanically calculate damages on an individual basis.” (Nye Report ¶ 72.) Courts routinely accept such an outline of the out-of-pocket method as sufficient to show damages are susceptible of classwide calculation. See Junge, 2022 WL 1002446, at *6 (collecting cases); City of Mia., 2018 WL 4931543, at *3–4. Defendants protest that Plaintiffs are only inadequately “gesturing at a model.” See Lytle, 114 F.4th at 1032. But the Ninth Circuit’s decision in Lytle demonstrates the opposite. There, plaintiffs in a false advertising case sought class certification. Id. at 1019. Their expert proposed measuring classwide damages with a conjoint survey. Id. at 1021. The survey hadn’t been conducted yet. Id. Instead, the expert explained the process of conducting a conjoint survey and his opinion that it was “well-suited to the facts of this case.” Id. The Ninth Circuit found that explanation was sufficient for class certification. Id. at 1034. It noted that “conjoint analysis is a well-accepted technique” in false advertising cases. Id. at 1033. Novel techniques “demand a greater degree of specificity and completeness.” Id. But that was unnecessary for a well- accepted technique. Id. Additionally, the “speculative possibility that [plaintiffs’ expert] might slip up in executing his model, standing alone, is insufficient to defeat class certification.” Id. The situation here is analogous. Nye proposes to use the standard out-of-pocket method to calculate damages. He is clearly familiar with conducting event studies, having done so for this motion to show market efficiency and price impact. (See Nye Report ¶ 52; Nye Reply Report ¶ 16.) As detailed above, his event study was able to associate the corrective disclosures with declines in Alphabet’s stock price. (Cf. Ferrell Report ¶ 95.) And the model can be modified to isolate confounding information. (See Nye Reply Report ¶ 108; cf. Ferrell Report ¶¶ 96–97.) The possibility that Nye will “slip up in executing his model” is purely speculative. See Lytle, 114 F.4th at 1033. In sum, Plaintiffs have sufficiently shown that Nye’s model will be able to reliably calculate classwide damages. IV. CONCLUSION For the foregoing reasons, Plaintiffs’ motion for class certification is GRANTED subject to the modified class definition agreed to at oral argument. Thus, the following class is certified:

All persons and entities other than Defendants, members of the immediate family of the Individual Defendant, Alphabet Inc.’s (“Alphabet” or the “Company”) subsidiaries and affiliates; any person who is or was an officer or director of the Company or any of the Company’s subsidiaries or affiliates during the Class Period; any entity in which any Defendant has a controlling interest; and the legal representatives, heirs, successors, and assigns of any such excluded person or entity, who purchased or otherwise acquired Alphabet’s Class A and/or Class C shares between September 15, 2020 and January 23, 2023, both dates inclusive (the “Class Period”) on (i) any stock exchanges located in the United States, (ii) on any alternative trading systems located in the United States, or (iii) pursuant to other domestic transactions, seeking to recover damages caused by Defendants’ violations of the federal securities laws and to pursue remedies under Sections 10(b) and 20(a) of the Securities Exchange Act of 1934 (the “Exchange Act”) and Rule 10b-5 promulgated thereunder, against Alphabet, Google LLC (“Google”), and Sundar Pichai.

(See Dkt. No. 134 at 7–8.) Plaintiffs (AMI - Government Employees Provident Fund Management Company Ltd.; City of Fort Lauderdale Police & Fire Retirement System; Menora Mivtachim Insurance Ltd.; and Menora Mivtachim Pensions and Gemel Ltd.) are appointed as class representatives. Pomerantz LLP is appointed as class counsel. IT IS SO ORDERED. Dated: May 18, 2026

RITA F. LIN United States District Judge