GlobalFoundries U.S. Inc. v. International Business Machines Corporation

Opinion

UNITED STATES DISTRICT COURT SOUTHERN DISTRICT OF NEW YORK GLOBALFOUNDRIES U.S. Inc.,

Plaintiff, No. 23-CV-3348 (KMK) v. ORDER INTERNATIONAL BUSINESS MACHINES CORPORATION,

Defendant. Appearances: Clement Joseph Naples, Esq. James Ellis Brandt, Esq. Jason Craig Hegt, Esq. Corey Anne Calabrese, Esq. Meredith A. Cusick, Esq. Latham & Watkins LLP New York, NY Counsel for Plaintiff

Maximilian A. Grant, Esq. Sami Al-Marzoog, Esq. Latham & Watkins LLP Washington, DC Counsel for Plaintiff

Brett M. Sandford, Esq. Blake R. Davis, Esq. Thomas Yeh, Esq. Latham & Watkins LLP San Francisco, CA Counsel for Plaintiff

Catherine Nyarady, Esq. Robert A. Atkins, Esq. Anna Gressel, Esq. Ariane Rockoff-Kirk, Esq. Darren Wright Johnson, Esq. Kripa Anand Raman, Esq. Pietro John Signoracci, Esq. Paul, Weiss, Rifkind, Wharton & Garrison LLP New York, NY Counsel for Defendant

KENNETH M. KARAS, United States District Judge: Before the Court is GF’s Objection to the denial of its May 31, 2024, letter motion to compel production of a single document. (See Not. of Obj. (Dkt. No. 400); Letter Mot. to Compel (Dkt. No. 384).)1 Magistrate Judge Andrew Krause denied the motion on the record at a June 14, 2024, discovery hearing, finding that the disputed document was neither relevant nor necessary to GF’s case. (See generally Decl. of Corey A. Calabrese, Esq., in Supp. of Mot. (“Calabrese Decl.”), Ex. D (“June Hearing Tr.”) (Dkt. No. 404-4.) For the following reasons the objection is sustained. I. Background A. Factual Background The Court assumes the Parties’familiarity with the underlying facts, so it reviews only those relevant to the Objection. Earlier on in this case, GF sought production of a chip design manual that it believed contained GF trade secrets. (SeeLetter Mot. to Compel. (Dkt. No. 283).) At a hearing on the motion, IBM represented, among other things, that “nothing in the design manual at issue was based on technology GF alleges it has the right to control” and, similarly, that nothing “predate[d] the start of the IBM/Intel Collaboration.” (Tr. of Proceedings (“3/7 Hearing Tr.”) at 55:12–15(Dkt. No. 301); see also id. at 63:10–12(“There’s nothing in here that suggests that anything that is GF related was used ... .”).)

1 To reduce reference to sealed information, the Court does not refer to the document by its name and uses the Parties’ agreed-upon acronym where necessary, as Judge Krause did on the record in multiple discovery hearings. Judge Krause took the matter under advisement, (see3/7 Hearing Tr. at 78:11–15), but soon after, GF found an inadvertently produced version of the document attached to an email included in an IBM document production. (See Supp. Letter Mot. to Compel. at 1 (Dkt. No. 309).) GF reupped its motion to compel, arguing that, contrary to IBM’s representations, the document did in fact contain material from derived GF’s protected information. (See generally

id.) Judge Krause revisited the issue at an April 17, 2024, hearing. Specifically, he found that the document appeared to contain “a substantial amount of information” traceable to one of GF’s earlier draft manuals and, generally, that there was “a lot of overlap” between the two. (Tr. of Proceedings (“4/17 Hearing Tr.”) at 21:17–24; 23:1–4 (Dkt. No. 394).)2 IBM chalked the similarity up to the Parties using the same template to create both documents. In other words, any shared material was insignificant boilerplate. (Seeid. at 15: 14–20.) Judge Krause explained, however, that determinations of a documents legal significance were not “the right questions for discovery.” (Id. at 21:21–24.) And he therefore ordered IBM to revisit its

production to see if similar documents were withheld for the same reason. (Id. 31:7–15.) Judge Krauseincluded an important caveat to that holding. IBM was to produce only “ultimate [draft]that [was]provided to Intel and Rapidus,” as opposed to “every [purely internal] version” of the design manual. (Id. at 43:1–6, 21–24.) There was not “enough of a basis” to produce the latter group of drafts at that time both because those internal drafts were never disclosed to the relevant partiesand because the ultimate documents that were disclosed would

2 To be clear, Judge Krause did not make any finding of bad faith. Instead, he recognized that IBM made a “discovery call,” albeit the wrong one. (4/17 Hearing Tr. at 22:5–14.) Nevertheless, he was still troubled by IBM’s “aggressive statements” at the March 7 hearing, a sentiment the Court shares. (Id.at 22:15–25.) “show what ha[d] been changed or improved or derived from whatever GlobalFoundries is asserting was the trade secret.” (See id. at 43:3–6.) More broadly, Judge Krause referenced concerns that the Parties’ positions with respect to these documents were too absolute—with GF seeking production of all related internal drafts, and IBM arguing that it need not produce any such drafts. (See id. at 14:16–25.) IBM produced at least three ultimate or final drafts in

response to Judge Krause’s order, which “resolved many of [GF’s] issues.” (SeeTr. of Proceedings (“6/14 Hearing Tr.” at 10:1–8) (Dkt. No. 422).) Later, GF happened upon another IBM email about one of those internal drafts (the “GDM Draft”). The email stated that the GDM Draft contained multiple “carry-overs from old documents.” (SeeCalabrese Decl., Ex. C at 2 (Dkt. No. 402-2).) And another email in the chain stated that a reference to “7nm” should be changed to “NS3 technology.” (See id.) GF contends that “7nm” refers to one of its protected processes and that NS3 is the version of that process ultimately sold by IBM to Intel and Rapidus. (Mem. of Law in Supp. (“GF Mem.”) 3 (Dkt. No. 401).) At bottom, GF’s theory is that IBM took GF’s document, “took [their] names off of it and

then used that document to create other documents with intel and rapidus.” (6/14 Hearing Tr. at 16: 9–16.) GF again moved to compel, (Dkt. No. 384), and Judge Krause took the matter up at a June 14 hearing. GF argued, as it does here, that the GDM Draft was highly relevant because it shows that “the basis for a lot of the documents that IBM is giving to Intel and Rapidus came directly from a GlobalFoundries document.” (6/14 Tr. at 10:18–23.) Judge Krause again denied the motion reasoning that such purely internal documents were outside the scope of the “parameters drawn at the April 17th conference.” (Id. at 16:3–6.) He added that GF had a “clear path” to arguing that its documents formed the basis for IBM’s collaboration with Intel and Rapidus via “inference” by comparing GF’s original drafts to the drafts IBM ultimately disclosed (which have since been produced to GF). (Id. at 15:8–13, 10:24–11:3.) B. Procedural History GF filed its Objection on June 28, 2024. (Not. of Obj.; GF Mem.; Calabrese Decl. (Dkt. Nos. 402, 404).) IBM filed its Opposition on July 17, 2024. (Mem. of Law in Opp. (“IBM

Mem.”) (Dkt. No. 416).) On July 22, 2024, GF requested leave to file a reply, which the Court granted. (See Order (Dkt. No. 423); Reply Mem. of Law (“GF Reply”) (Dkt. No. 426).)3 II. Discussion A. Standard of Review A district court reviewing a decision from a magistrate judge addressing a dispositive motion “may accept, reject, or modify, in whole or in part, the findings or recommendations made by the magistrate judge.” 28 U.S.C. § 636(b)(1). However, “[u]nder Rule 72 of the Federal Rules of Civil Procedure, ‘when a pretrial matter not dispositive of a party’s claim or defense is referred to a magistrate judge to hear and decide,’ the district court ‘must consider timely objections and modify or set aside any part of the order that is clearly erroneous or is contrary to law.’” Michelo v. Nat’l Collegiate Student Loan Tr. 2007-2, Nos. 18-CV-1781, 18-

CV-7692, 2022 WL 153183, at *2 (S.D.N.Y. Jan. 18, 2022) (alteration adopted). “Orders involving discovery are considered nondispositive.” Pac. Life Ins. Co. v. Bank of New York Mellon, 571 F. Supp. 3d 106, 112 (S.D.N.Y. 2021). “‘An order is clearly erroneous only when the reviewing court on the entire evidence is left with the definite and firm conviction that a mistake has been committed’ and is ‘contrary to law if it fails to apply or misapplies relevant statutes, case law[,] or rules of procedure.’” Id. (quoting Blackrock Allocation Target Shares:

3 The Court granted letter motions to seal the filings related to this Objection. (See Dkt. Nos. 405, 428.) Series S. Portfolio v. Wells Fargo Bank, Nat’l Ass’n, No. 14-CV-10067, 2018 WL 3863447, at *3 (S.D.N.Y. Aug. 13, 2018))(internal quotation marks omitted). Because “‘[a] magistrate judge is best qualified to judge the entire atmosphere of the discovery process,’” his “‘rulings on discovery matters are entitled to substantial deference.’” Michelo, 2022 WL 153183, at *2 (quoting U2 Home Entm’t Inc. v. Hong Wei Int’l Trading Inc., No. 04-CV-6189, 2007 WL

2327068, at *1 (S.D.N.Y. Aug. 13, 2007)). B. Analysis GF now objects to the June 14 ruling (the “Ruling”)as contrary to Rule 26(b)(1). (See generally GF Mem.) IBM responds that the Objection was waived and that, in any event, it should be denied on the merits. (See generally IBM Mem.) 1. Waiver As an initial matter, IBM argues that GF waived this argument by failing to object to Judge Krause’s April 17 ruling. (IBM Mem. 2–3.) IBM’s view is that the April 17 ruling encompassed the instant dispute and that GF’s time to object to that ruling has since expired. GF’s response is that Judge Krause’s ruling was limited to the documents at issue in those hearings and that he expressly left open the discoverability of internal drafts. (SeeGF Reply 1.)

The law regarding a party’s ability to object in this scenario is somewhat unsettled. On the one hand, anything in the four corners of Judge Krause’s April 17 ruling is off limits. Several cases have held pursuant to Rule 72(a)’s clear terms that courts “may not ‘assign as error’” a defect an earlier discovery ruling that “was ‘not timely objected to.’” See Royal Park Invs. SA/NV v. Deutsche Bank Nat’l Tr. Co., No. 14-CV-4394, 2017 WL 9802844, at *5 (S.D.N.Y. Sept. 15, 2017) (quoting Fed. R. Civ. P. 72(a)) (explaining an unobjected-to non- dispositive ruling was “not at issue” at a subsequent conference), report and recommendation adopted, 2018 WL 1750595(S.D.N.Y. Mar. 29, 2018); Partminer Worldwide Inc. v. Siliconexpert Techs. Inc., No. 09-CV-586, 2011 WL 587971, at *5 (D. Colo. Feb. 9, 2011) (finding objection untimely where party had opportunity to raise legal issue in prior discovery objection); Maryland Cas. Co. v. W.R. Grace & Co.-Conn, No. 88-CV-2613, 1994 WL 592267, at *6 (S.D.N.Y. Oct. 26, 1994) (holding that the magistrate judge properly adhered to a prior and un-objected to privilege ruling as the law of the case, “to be followed in successive stages of the

same litigation”). That is especially so where the objecting party offers no special circumstance explaining why it “could not [have] raised [its] objections sooner.” See Lewis v. United States, No. 19-CV-79, 2021 WL 5304190, at *2 n.4 (S.D. Ga. Nov. 15, 2021). On the other hand, if two discrete discovery rulings commit the same legal error, it is not clear that a party waives that ground for objection by failing to raise it in the first instance. Indeed, the Court is aware of one case stating expressly that Rule 72 “does not contemplate waiver by failing to object to a previous order discussing the same issue.” See S.E.C. v. McNaul, 277 F.R.D. 439, 442 (D. Kan. 2011). And in this District, Judge Wood, citing McNaul, considered an objection in similar circumstances. See Thai Lao Lignite (Thailand) Co., Ltd. v.

Gov’t of Lao People’s Democratic Republic, 924 F. Supp. 2d 508, 518 (S.D.N.Y. 2013). A party in Thai Lao Lignite objected to a discovery ruling intended to “obtain . . . compliance” with an earlier unobjected-to order. See id. Even though Judge Wood was persuaded that the second order “imposed no new obligations” on the parties, she considered the objection on the merits given the “conflicting case law regarding waiver for failing to object to a prior order.” Id. Here, GF’s objection is timely and the Court is not convinced that a waiver rule would apply in this posture to the extent it exists in the first place. First, the Court views the April 17 and June 14 rulings as distinct, meaning the second ruling started a new clock for purposes of Rule 72. Even though the former ruling referred to “internal” drafts, it did not address the document now in dispute or GF’s particular arguments about that document. Additionally, the April 17 ruling did not make a categorical determination that internal drafts would never be discoverable. Instead, Judge Krause stated he did not think production of such drafts would be appropriate “at [that] point” and left “for another day” the “exact parameters” of the derivative drafts IBM must produce. (4/17 Hearing Tr. at 43:12–20.) Second, and for similar reasons, it is

not clear that both rulings addressed the “same issue.” See McNaul, 277 F.R.D. at 442. Again, it seems that the April 17 ruling set a working parameter for discovery based on information available at the time. And the June 14 ruling teed up a similar, but materially different, “issue” when it applied that parameter to new facts. To the extent the rulings did address the same issue, the Court follows Judge Wood’s approach and exercises its discretion to consider the objection on the merits. See Thai Lao Lignite, 924 F. Supp. 2d at 518.4 3. GF’s Objection As to the merits of the Objection, Federal Rule of Civil Procedure 26(b) instructs that parties “may obtain discovery regarding any nonprivileged matter that is relevant to any party’s claim or defense . . . considering, among other factors, whether the burden or expense of the

proposed discovery outweighs its likely benefit.” In the Matter of the Complaint of Specialist LLC, No. 16-CV-2515, 2016 WL 6884919, at *3 (S.D.N.Y. Nov. 22, 2016) (quotation marks and emphasis omitted) (quoting Fed. R. Civ. P. 26(b)). In other words, the Court considers the “marginal utility of the discovery sought,” Vaigasi v. Solow Mgmt.Corp., No. 11-CV-5088,

4 Although the Court exercises this discretion, nothing about this Order should be read as an endorsement of piecemeal objections to Judge Krause’s rulings. The Court is wary that GF’s arguments here can be applied to virtually any single document that later appears relevant after it is initially excluded from production—in a vacuum, it will always be the case that the burden of producing “a single, readily available document” seems trivial by comparison. (See GF Mem. 12.) The Parties should endeavor to raise any future objections at the earliest possible opportunity and account for the full context of discovery parameters in this case. 2016 WL 616386, at *14 (S.D.N.Y. Feb. 16, 2016) (emphasis added), taking into account the dangers of “overly broad, burdensome, or oppressive” requests, Alaska Elec.Pension Fund v. Bank of America Corp., No. 14-CV-7126, 2016 WL 6779901, at *2 (S.D.N.Y. Nov. 16, 2016) (internal quotation marks omitted). “Although not unlimited, relevance, for purposes of discovery, is an extremely broad concept,” id., and “the [C]ourt should and ordinarily does

interpret ‘relevant’ very broadly to mean matter that is relevant to anything that is or may become an issue in the litigation,” see Oppenheimer Fund, Inc. v. Sanders, 437 U.S. 340, 351 n.12 (1978) (emphasis added) (internal quotation marks omitted). To set the stage, Judge Krause made two points in support of the June 14 Ruling: (1) that the GDM Draft was outside the scope of the no-internal-drafts parameter set at the April 17 hearing(6/14 Hearing Tr. at 15:24–16:6); and (2) that GF could adequately argue misappropriation via inference, (see id. at 15: 8–13). The Court views this Ruling primarily as a relevance determination. Although Judge Krause previously referenced concerns that GF’s stance on producing internal communications, writ large, was “overbroad,” (see 4/17 Hearing Tr.

at 21:12–14), he did not make a finding that production of this document would be unduly burdensome, nor does IBM make such an argument here. GF’s objection is that “[a]ny use of GF’s trade secrets,” even a purely internal and incremental draft copying some of GF’s protected material, is direct evidence of misappropriation, and thus highly relevant. (GF Mem. 9 & n.3.) For similar reasons, it claims that the GDM Draft helps establish IBM’s ultimate breach of the ACA, helps establish that IBM acted willfully, and helps rebut IBM’s independent development defense. (Id.9–11) Because there is no countervailing burden to outweigh the GDM Draft’s relevance, GF views the June 14 Ruling as contrary to law. IBM responds that this purely internal draft is irrelevant because GF’s theory of misappropriation is limited to “allegations of improper disclosure” and, as Judge Krause ruled, GF can adequately argue misappropriation via inference. (See IBM Mem. 3.) Both IBM and the Ruling appear to reflect the view that the scope of GF’s misappropriation claim is determinative. (See 6/14 Tr. at 15:4–7 (indicating that the GDM Draft “was not shared and not intended to be shared”).) And in some sense, this is a fair point. Based

on the terms of the ACA—which only grants GF exclusive rights to license or disclose certain information—there is an argument that IBM’s internal use of GF trade secrets is privileged and the only thing that could constitute misappropriation is disclosure to third parties not named in that agreement. (See IBM Mem. 3–4.) The missing piece, though, is why nondisclosure renders the GDM Draft irrelevant, or of minimal relevance. Assuming GF’s claim is limited to disclosure, the GDM Draft—with the accompanying email—is relevant if for no other reason than it is probative of what IBM disclosed.5 To prove a DTSA disclosure claim, “a plaintiff must show an unconsented disclosure . .. of a trade secret by one who . . . at the time of disclosure, knew or had reason to know that the trade secret was

acquired through improper means, under circumstances giving rise to a duty to maintain the secrecy of the trade secret, or derived from or through a person who owed such a duty.” Nielsen Consumer LLC v. Circana Grp. L.P., No. 22-CV-3235, 2023 WL 5917751, at *8 (S.D.N.Y. Sept. 11, 2023) (alteration adopted) (internal quotation marks omitted); see also Catalyst Advisors, L.P. v. Catalyst Advisors Invs. Glob. Inc., 602 F. Supp. 3d 663, 676 (S.D.N.Y. 2022) (same) (quoting 18 U.S.C. § 1839(5)). The providence of the trade secret information therefore plays a key role. Defendants often argue, as IBM does here, that they acquired the information

5 Because it takes as given IBM’s more limited view of GF’s claims, the Court need not address IBM’s arguments that this case does not fit the “typical” misappropriation mold. (See IBM Mem. 6–7 & n.4.) through “independent development,” not through a confidential relationship. See ScentSational Techs., LLC v. PepsiCo, Inc., No. 13-CV-8645, 2017 WL 4403308, at *13 (S.D.N.Y. Oct. 2, 2017) (outlining the independent development defense). (See also Answer at 46 (Dkt. No. 213 (asserting an independent development defense).) Or they may claim—as IBM does—that they did not disclose a trade secret at all and any similarity between a plaintiff’s and a defendant’s

documents results from something else, like template software. (See 4/17 Hearing Tr. at 23:9–19 (IBM representing that “[s]aying that the GDM is traceable to the GF test chip” is inaccurate because “they are both traceable to a program . . . a template that IBM uses”).) In this case, the GDM along with its accompanying email tends to refute both theories. The reference to “carry- overs from old documents” in the GDM—particularly the use of “7nm”—suggests that IBM may not have injected all of its own independent ideas into the drafts ultimately disclosed to Intel and Rapidus.6 And it can also be used to refute the notion that overlap between the Parties documents merely results from similar software, as opposed to copying and the like. At bottom, the Court agrees that the GDM Draft may act as a “link” between the ultimate drafts disclosed to

Intel and Rapidus—something that helps the factfinder determine where the disclosed information came from. In that sense it is relevant to one or more of the claims or defenses alleged in this case.7

6 Assuming GF’s original design manual contains “Specific Results,” the Court agrees with GF that the GDM Draft would also be probative of breach of contract to the extent it contains that same material. (SeeGF Mem. 10.)

7 Although not determinative, the Court does not view the GDM Draft as “direct evidence” of misappropriation. GF’s direct evidence argument appears to assume that both the GDM Draft and its initial draft contain trade secret material. (See GF Mem. 9–10.) And any such determination is premature at this stage. IBM’s assertions that the GDM Draft does not contain trade secret material are similarly non-dispositive. (See IBM Mem. 4 n.3.) The Court has no real way to assess the veracity of those statements, and regardless, the alleged overlap between the GF’s initial draft, the GDM Further, the ability to argue misappropriation via inference does not reduce the GDM’s value. Of course, it is “well-recognized” that plaintiffs need not rely on direct evidence and can instead “construct a web of perhaps ambiguous circumstantial evidence from which the trier of fact may draw inferences [of misappropriation].” ScentSational Techs., 2017 WL 4403308, at *13 (alteration adopted)(quotation marks omitted); Stanacard, LLC v. Rubard, LLC, No. 12-CV-

5176, 2016 WL 462508, at *19 (S.D.N.Y. Feb. 3, 2016) (same). In that sense, the inquiry is something of a black box—a battle of inferences to see whether “it is more probable than not that what plaintiff alleges happened did in fact take place.” Catalyst Advisors, LP v. Catalyst Advisors Invs. Glob. Inc., No. 21-CV-4855, 2024 WL 522751, at *24 (S.D.N.Y. Feb. 9, 2024) (alterations adopted) (quotation marks omitted). But while sufficient in some circumstances, that type of argument hardly inspires confidence for plaintiffs. Indeed, this “web” of inferences is “often [a] delicate construction” that fails when faced with defense witnesses “who directly deny everything.” See id. And it may not be enough to rely on “mere overlap,” or the “similar nature” of two processes, as IBM suggests GF should here. See Parchem Trading, Ltd. v. DePersia, No.

17-CV-1618, 2020 WL 764211, at *11 (S.D.N.Y. Feb. 14, 2020) (granting summary judgment to defendant on misappropriation claim). (See IBM Mem. 6–7.) Given how tenuous that inquiry can be, an additional “link” allowing GF to take smaller inferential leaps, (see GF Mem. 14), may make a material difference in its case. Accordingly, because the GDM is relevant and there was no finding that the burden of producing it would outweigh any likely benefit, see In the Matter of the Complaint of Specialist LLC, 2016 WL 6884919, at *3, GF’s objection is sustained.

Draft, and the documents ultimately disclosed to Intel and Rapidus are what establish the GDM Draft’s likely relevance. There are two caveats to this ruling, however. First, this Order is limited to the GDM draft referenced in Exhibit C to the declaration accompanying GF’s Objection, and nothing about the Order should affect the analysis of other internal drafts or communications. As the above analysis indicates, the GDM Draft’s relevance depends in significant part on the accompanying email which characterizes what the draft was, what it contained, and edits IBM made to it. See supra Section L.A. Production of additional internal drafts may raise a host of other issues, including being burdensome or cumulative, none of which the Court has had occasion to address here. Second, the Court expresses no opinion about whether the GDM Draft constitutes a trade secret, a derivative of a trade secret, or 1s otherwise comprised of trade secret material. As Judge Krause indicated, that type of question 1s inappropriate for purposes of discovery, and the Court is keenly aware that the scope of discovery regarding derivatives remains a live issue. Il. Conclusion For the foregoing reasons, the Objection is sustained. SO ORDERED.

Dated: October 9, 2024 White Plains, New York iy agin on KENNETH M. KARAS United States District Judge