The opinion
1
2
3
4
5
6
7
8 UNITED STATES DISTRICT COURT
9 NORTHERN DISTRICT OF CALIFORNIA
10 San Francisco Division
11 TODD CROWDER, et al., Case No. 22-cv-00237-HSG (LB)
12 Plaintiffs,
DISCOVERY ORDER
13 v.
Re: ECF Nos. 267-3, 268
14 LINKEDIN CORPORATION,
15 Defendant.
16
17 INTRODUCTION
18 This is a putative antitrust class action about the market for professional social networking.
19 The plaintiffs served a Rule 30(b)(6) deposition notice on LinkedIn with seven topics. After a
20 court-ordered meet-and-confer, the parties resolved Topics 1 through 4 (subject to document
21 production by a date certain) and now dispute the scope of LinkedIn’s testimony on Topic 5
22 (LinkedIn’s integration of its infrastructure and AI development with Microsoft Azure) and Topics
23 6 and 7 (the systems and methodology for pricing LinkedIn’s paid subscription products). Fact
24 discovery closes on September 18, 2026.1
25
26
27 1 Disc. Letter Br. – ECF No. 267-3 at 1–3 (sealed brief); Scheduling Order – ECF No. 143; see Disc.
1 Three disputes remain. First, LinkedIn will produce a witness on Topic 5 to testify “at a
2 || general level” about the contemplated Azure migration, its pause, and the extent of LinkedIn’s use
3 || of Azure, but (as clarified at the meet-and-confer) it will not prepare the witness on the details of
4 || LinkedIn’s current Azure workloads, its capacity purchases, or its Azure pricing.” Second, the
5 || plaintiffs contend that LinkedIn will limit its testimony on Topics 6 and 7 to 2018 forward, even
6 || though the court set a January 2014 lookback for the associated pricing discovery.’ Third,
7 || LinkedIn declines to provide testimony on Topic 7(c), which concerns Ls
|
9 The court can decide the dispute without oral argument, Civil L. R. 7-1(b), and rules as
10 || follows. First, LinkedIn must prepare its designees to testify to Topic 5 as written. The court’s
11 earlier orders made this deposition the plaintiffs’ vehicle for testing whether LinkedIn’s two-
12 || document production fully illuminates “the extent of its use of Azure,” and a witness who cannot
13 || address current workloads, capacity purchases, spending, and future plans cannot serve that
14 || function.? Second, the designees must be prepared to testify to Topics 6 and 7 for January 2014 to
3 15 || the present, which is the period that the court fixed for the associated document discovery and one
a 16 || that LinkedIn’s own statement appears to accept.° Third, LinkedIn must prepare its designee to
2 17 || testify to Topic 7(c). Finally, the court construes the “general level” qualifier:’ the designees must
Z 18 || testify to information known or reasonably available to LinkedIn on each topic as scoped by this
19 || order, without unilateral subject-matter carve-outs, but the deposition is not a memory contest, and
20 || the witnesses may reasonably rely on (and explain) the documents and data that LinkedIn has
21 produced.
22
23
24 * Disc. Letter Br. — ECF No. 267-3 at 2-3.
25 || 3 Jd. at 3 (citing Order — ECF No. 170 at 14).
26 || ‘i
° Order — ECF No. 170 at 9.
27 ll 6 Disc. Letter Br. — ECF No. 267-3 at 4.
28 || ‘Id. at3.
1 STATEMENT
2 The operative complaint alleges that LinkedIn’s professional-social-networking business 1s
3 || protected by a “Data, Machine Learning, and Inference Barrier to Entry” and that LinkedIn
4 || integrated its operations with its parent Microsoft’s Azure cloud platform in a way that deterred
5 |} competition.* From 2019 to 2022, LinkedIn said publicly that it would move its entire business
6 || onto Azure and scrap its own datacenters (a plan known internally as Blueshift). In December
ee
7 || 2023, it was reported publicly that the plan had been paused;
8 ” The plaintiffs contend that the
9 || pause did not end the relationship:
10
ll 10
a 12 The parties have litigated Azure discovery before. The plaintiffs sought Azure documents
& 13 || through RFP Nos. 13-16 and moved to add a custodian. The court denied the custodian and
14 || adopted a compromise that LinkedIn proposed: LinkedIn would produce documents showing “(1)
2 || the migration was never restarted and (2) the extent of its use of Azure,” followed by a Rule
Q 16 |} 30(b)(6) deposition after “full illumination through . . . document production of the core requests
=
s 17 || in RFP Nos. 13-16.”!! LinkedIn produced two documents: a five-page email thread from 2022
Z 18 || (with one email, dated April 2026, redacted) and a spreadsheet. !? The court then denied the
19 || plaintiffs’ motion to compel further documents without prejudice, directed the parties “to proceed
20 || with a Rule 30(b)(6) deposition on the Azure topic,” and preserved the plaintiffs’ right to “renew
21 their motion to compel upon a showing that LinkedIn’s Azure-related production and Rule
22 || 30(b)(6) deposition testimony is inadequate.” ? It added: “To the extent the plaintiffs believe that
23
2A 8 Disc. Letter Br. — ECF No. 267-3 at 1 (citing First Am. Compl. — ECF No. 65 (4 279-325)).
* Id.
1 7a. at 1-2 (citing LI CROWDER 00145822, LL CROWDER 00181356,
26 LI CROWDER _ 00474939, and LI CROWDER 00212333).
1! Td. at 2 (quoting Order — ECF No. 170 at 9).
27 2 7g
28 |! Id. (quoting Order —ECF No. 222 at 7).
1 LinkedIn’s two-document production is not a ‘full illumination’ of the core requests in RFP Nos.
2 |} 13-16, they can test that theory through the Rule 30(b)(6) deposition.” !*
3 As directed by the court’s July 17, 2026 discovery order, the plaintiffs served their Rule
4 || 30(b)(6) notice on July 31, 2026.!° Topic 5 seeks testimony about “LinkedIn’s integration of its
5 || infrastructure, applications, workloads, and artificial-intelligence development with Microsoft
6 || Azure, under any internal project name, including (a) which LinkedIn systems, applications, or
7 || workloads have run on or been moved to Azure, and when; (b) LinkedIn’s purchases of Azure
8 || services and of GPU and other AI computing capacity from Microsoft; (c) LinkedIn’s Azure
9 || spending, commitments, and forecasts; and (d) LinkedIn’s plans for future use of Azure.” !°
10 Topics 6 and 7 concern the systems and methodology that LinkedIn uses to price its paid
11 subscription products (Premium Career, Premium Business, Premium Company Page, Sales
ee
« 12 || Navigator, and Recruiter Lite).!’ The plaintiffs contend that discovery revealed that
15
£ a
s 17 a 1° The plaintiffs also say that LinkedIn has not produced historical pricing data housed in
Z 18 ee that LinkedIn’s own witnesses identified, and that they will move
19 || separately to compel that data and for sanctions; that issue is not before the court, and this order
20 || does not address it.7°
21
22
23 oT
4 Td. (quoting Order — ECF No. 222 at 6).
24 45 1a at 1, 3 (citing Order — ECF No. 212 at 8).
25 |) "Id. at 1.
17
26 Id. at 2, 4.
"8 Td. at 2.
27 Wl 74. at 3.
28 || 7° Id. at 2-3.
1 LinkedIn served responses and objections on August 18, 2026, and on August 27 designated
2 || Raghu Hiremagular and Alek Shoykhet (both also noticed in their individual capacities) as its
3 || designees for Topics 5 through 7.7! Lead counsel met and conferred on September 3, 2026, and
4 || the parties filed this letter brief the same day.”
5 LinkedIn objects that Topic 5 “is disproportionate to the needs of the case because LinkedIn ha
6 || already produced documents sufficient to show its Azure usage and spending” and offers a witness
7 || subject to “proportionality considerations.” On the plaintiffs’ account of the meet-and-confer, that
8 || means no preparation on the details of LinkedIn’s current Azure workloads, capacity, or pricing, ar
9 || account that LinkedIn’s statement does not deny and that its proffer (testimony about the
10 || contemplated integration, the pause, and the extent of Azure use to “illuminate the document
11 || production”) corroborates.” LinkedIn responds that testifying “at a general level” means only that
12 || will prepare its witnesses in good faith and consistent with Rule 26 proportionality.** For Topics 6
13 || and 7, LinkedIn commits to identify the databases, systems, and repositories where it keeps
2 15 , and to address pricing strategy and methodology — including its □
a 16 ee — for the five Premium products, with dates “consister
2 17 || with the date ranges in the Court’s order on the associated document requests.””° The plaintiffs say
Z 18 || that LinkedIn told them it would limit its testimony to 2018 to the present.*°
19
20 LEGAL STANDARD
21 Parties may obtain discovery regarding any nonprivileged matter that is relevant to any party’s
22 || claim or defense and proportional to the needs of the case, considering the importance of the
23
24 21 Id. at 3.
25 || 7? Jd. at 1.
26 || 7 dd. at 244.
4 Td. at 3-4.
27 Wl 2s 14. at.
28 || 7° Jd. at 3.
1 issues at stake in the action, the amount in controversy, the parties’ relative access to relevant
2 information, the parties’ resources, the importance of the discovery in resolving the issues, and
3 whether the burden or expense of the proposed discovery outweighs its likely benefit. Information
4 within this scope of discovery need not be admissible in evidence to be discoverable. Fed. R. Civ.
5 P. 26(b). “Pretrial discovery is ordinarily accorded a broad and liberal treatment.” Peng v. Nw.
6 Mut. Life Ins. Co., No. 17-cv-01760-SI, 2017 WL 3007030, at *1 (N.D. Cal. July 14, 2017)
7 (cleaned up) (quoting Shoen v. Shoen, 5 F.3d 1289, 1292 (9th Cir. 1993)). Moreover, “[t]he test
8 for relevance is not overly exacting: evidence is relevant if it has ‘any tendency to make . . . more
9 or less probable . . . [a] fact [that] is of consequence in determining the action.’” In re Volkswagen
10 “Clean Diesel” Mktg., Sales Pracs., & Prods. Liab. Litig., MDL No. 2672 CRB (JSC), 2018 WL
11 340640, at *1 (N.D. Cal. Jan. 9, 2018) (quoting Fed. R. Evid. 401).
12 The party moving to compel discovery “has the initial burden of establishing that the
13 information sought is relevant to any party’s claim or defense and proportional to the needs of the
14 case.” Impinj, Inc. v. NXP USA, Inc., No. 19-cv-03161-YGR (AGT), 2022 WL 16586886, at *2
15 (N.D. Cal. Nov. 1, 2022) (cleaned up). The party resisting discovery bears the burden of showing
16 that the discovery should not be allowed and of supporting its objections with competent evidence.
17 Lofton v. Verizon Wireless (VAW) LLC, 308 F.R.D. 276, 281 (N.D. Cal. 2015). The court must limit
18 discovery that is “unreasonably cumulative or duplicative,” obtainable from a less burdensome
19 source, or where the burden “outweighs its likely benefit.” Fed. R. Civ. P. 26(b)(1), (b)(2)(C).
20 Upon the noticing of a corporation’s deposition, it must designate one or more witnesses who
21 are knowledgeable on the noticed topic to testify on its behalf. Fed. R. Civ. P. 30(b)(6); Apple Inc.
22 v. Samsung Elecs. Co., No. C 11-1846 LHK (PSG), 2012 WL 1511901, at *2 (N.D. Cal. Jan. 27,
23 2012). A Rule 30(b)(6) notice must describe the matters for examination “with reasonable
24 particularity.” Fed. R. Civ. P. 30(b)(6). While a corporation must make a good-faith effort to
25 prepare a Rule 30(b)(6) witness to “fully and unevasively answer questions about the designated
26 subject matter,” the rule does not extend to “burdening the responding party with production and
27 preparation of a witness on every facet of the litigation.” Apple, 2012 WL 1511901, at *2.
1 ANALYSIS
2 1. The Scope of Topic 5 (Azure)
3 The issue is whether LinkedIn may confine its preparation to a general account of the
4 contemplated Azure migration and its pause, excluding the details of LinkedIn’s current Azure
5 workloads, its purchases of computing capacity from Microsoft, its spending and commitments, and
6 its future plans. It may not: LinkedIn must prepare its designees to testify to Topic 5 as written.
7 The court’s earlier orders frame — and largely resolve — the issue. The compromise the court
8 adopted, at LinkedIn’s suggestion, obligated LinkedIn to illuminate two subjects: that “the
9 migration was never restarted” and “the extent of its use of Azure.”27 When the plaintiffs
10 challenged the resulting two-document production, the court declined to compel more documents
11 and made this deposition the testing ground: the plaintiffs “can test that theory through the Rule
12 30(b)(6) deposition,” and they may renew their motion to compel only “upon a showing that
13 LinkedIn’s Azure-related production and Rule 30(b)(6) deposition testimony is inadequate.”28
14 LinkedIn commits to testimony about the extent of its use of Azure, but (per its objections and its
15 meet-and-confer position), it would exclude the details that give that phrase content: what
16 currently runs on Azure, what LinkedIn buys and spends, and what it plans. A designee prepared
17 only to narrate the pause cannot be examined meaningfully about the extent of LinkedIn’s use of
18 Azure and so cannot serve the function that the court assigned to this deposition. LinkedIn’s
19 proposal would collapse the compromise’s second prong into its first.
20 LinkedIn’s reading of the court’s July 31 order does not change the analysis. LinkedIn points
21 to the order’s observations that the “core requests” concern “the scope and status of the Azure
22 migration” and that there is “no rule about how many documents it takes to make such a
23 showing.”29 Those observations addressed the adequacy of LinkedIn’s document production (an
24 issue the order reserved) in the same order that directed a deposition “on the Azure topic” as the
25
26
27 Disc. Letter Br. – ECF No. 267-3 at 2 (quoting Order – ECF No. 170 at 9).
27
28 Id. (quoting Order – ECF No. 222 at 6–7).
1 mechanism for testing the extent of LinkedIn’s Azure use. Reading “migration” to exclude
2 || LinkedIn’s present and planned use of Azure would nullify the second prong of the compromise
3 || that LinkedIn itself proposed and would leave the plaintiffs no way to make the inadequacy
4 || “showing” that the order contemplates.
5 Topic 5 also is proportionate to the needs of the case. It is one subject (LinkedIn’s integration
6 || with and use of Azure) divided into four particularized subparts that track the court’s earlier
7 || orders: what runs on Azure and since when (5(a)); what LinkedIn buys from Microsoft, including
8 || GPU and other AI computing capacity (5(b)); what it spends, commits, and forecasts (5(c)); and
9 || what it plans (5(d)). It is not a catch-all demand for testimony about every aspect of LinkedIn’s
10 business. Cf Concurrent Ventures, LLC v. Advanced Micro Devices, Inc., No. 25-cv-09567-JST
11 (LB), 2026 WL 2445547, at *4 (N.D. Cal. Aug. 9, 2026). Its relevance is concrete: the complaint
12 || alleges that LinkedIn’s integration with its parent’s cloud platform deterred competition, and the
& 13 || plaintiffs cite produced documents suggesting that
2 15 ee. 30 Whether those facts ultimately support the claim is a merits question:
Q 16 || for discovery purposes, they show that the deposition examines a live, particularized issue. And
s 17 || the burden is modest: LinkedIn has already designated two witnesses, and the commercial
6
Z 18 |} sensitivity of capacity and pricing information can be managed under the protective order and the
19 || sealing procedures the parties are already using.*!
20
21
30
22 Id. at 1-2.
31 The closest question in the letter brief is whether the deposition must reach the prices and commercia
23 terms of LinkedIn’s Azure purchases. If the court reads its earlier orders as preserving deposition testin
only of the “core requests” (the scope and status of the migration and the extent of LinkedIn’s use of
24 || Azure), it could conclude that the negotiated terms of LinkedIn’s purchases from Microsoft lie beyond
that pale. On that view, the court would compel Topic 5 with one limit: LinkedIn must prepare its
25 designees fully on subparts (a) and (d) and on the categories, magnitudes, and aggregate amounts of its
Azure purchases, spending, commitments, and forecasts (subparts (b) and (c)), but need not prepare
26 them on unit pricing or the negotiated terms of its agreements with Microsoft. Pricing-specific □□□□□□□□
would await a renewed motion under the July 31 order’s without-prejudice reservation. But the court
27 does not read its earlier orders that narrowly. The notice defines the deposition, the DMIBE theory
makes intercompany terms probative, the court intended the depositions to allow the plaintiffs to test
28 their theories, and the discovery is proportional to the case’s needs.
1 LinkedIn’s designees therefore must be prepared to testify, to the extent the information is
2 || known or reasonably available to LinkedIn, to each subpart of Topic 5 as written, including
3 || LinkedIn’s current Azure workloads and systems; its purchases of Azure services and of GPU and
4 || other AI computing capacity (including the prices and terms of those purchases); its Azure
5 || spending, commitments, and forecasts; and its plans for future use of Azure. The preparation
6 || standard in Section 4 applies.
7
8 2. The Temporal Scope of Topics 6 and 7
9 The plaintiffs ask that testimony on Topics 6 and 7 cover January 2014 to the present; they say
10 || that LinkedIn would limit its witnesses to 2018 forward.*” The court set a January 2014 lookback
11 for the associated pricing document discovery, and the prices charged immediately before the class
«12 || period are material to the economic analysis of the prices charged during it.*? LinkedIn’s own
& 13 || statement, moreover, agrees to testimony “consistent with the date ranges in the Court’s order on
. ” . : 34 ‘aad .
14 || the associated document requests,” which is the January 2014 lookback.’* To eliminate any doubt:
2 15 || LinkedIn must prepare its designees to testify to Topics 6 and 7 for the period from January 1,
QA 16 || 2014, to the present, to the extent the information is known or reasonably available to it.
=
17
es
Z 18 || 3. Topic7(c) i)
19 LinkedIn agreed to provide testimony on Topic 6 and on most of Topic 7, but it declines to
ee
20 || provide testimony on Topic 7(c). The topic subpart concerns a. which (on the plaintiffs’
22 -* LinkedIn’s statement offers no argument for the carve-out. Testimony about i
23 ee at issue falls within topics addressed to LinkedIn’s pricing systems and
24
25
26 || * Disc. Letter Br. — ECF No. 267-3 at 3.
33 Td. (citing Order — ECF No. 170 at 14).
27 Wl 3474. ata.
28 || * Id. at 3.
1 methodology, and LinkedIn identifies no disproportionate burden. LinkedIn must prepare its
2 || designee to testify to Topic 7(c).
3
4 4. Witness Preparation and the “General Level” Qualifier
5 Two constructions govern the depositions. First, “at a general level” is not a license to
6 || withhold noticed subjects. This order fixes the scope of Topics 5 through 7. Within that scope,
7 || LinkedIn must prepare its designees to testify to information known or reasonably available to it,
8 || without unilateral subject-matter carve-outs. Second, the plaintiffs must take the witnesses as Rule
9 || 30(b)(6) provides them: the deposition “is not designed to be a memory contest,” Williamson v.
10 || Google LLC, No. 15-CV-00966 BLF (NC), 2018 WL 11414609, at *2 (N.D. Cal. May 24, 2018),
11 and the rule does not require “production and preparation of a witness on every facet of the
12 litigation,” Concurrent Ventures, LLC v. Advanced Micro Devices, Inc., No. 25-cv-09567-JST
13 || (LB), 2026 WL 2568608, at *2 (N.D. Cal. Aug. 31, 2026) (cleaned up). A prepared designee may
14 || reasonably rely on and explain the documents and data that LinkedIn has produced rather than
3 15 || recite figures from memory. If precise data points matter, the documents are the vehicle, and, for
a 16 || the historical pricing data that the plaintiffs say is missing, whatever relief the court may order.
2 17 || The absence of that data is not a reason to leave a witness unprepared on the pricing systems and
Z 18 || methodology themselves.
19 CONCLUSION
20 The court resolves the letter brief as follows.
21 1. Topic 5. LinkedIn must designate and prepare one or more witnesses to testify, to the extent
22 || the information is known or reasonably available to it, to Topic 5 as written, subparts (a) through
23 (d), including LinkedIn’s current and historical Azure workloads and systems; its purchases of
24 || Azure services and of GPU and other AI computing capacity from Microsoft, including prices and
25 terms; its Azure spending, commitments, and forecasts; and its plans for future use of Azure.
26 2. Topics 6 and 7. The designees must be prepared to testify to Topics 6 and 7 for the period
27 || from January 1, 2014, to the present.
28 3. Topic 7(c). LinkedIn must prepare its designee to testify about a.
] 4. Preparation. The designees must testify to information known or reasonably available to
2 || LinkedIn on the topics as scoped by this order; they may reasonably rely on and explain
3 LinkedIn’s produced documents and data and need not memorize granular data.
4 5. Timing. The court understands that the depositions will take place on September 8, 2026.
5 This resolves ECF Nos. 267-3 and 268.
6 IT IS SO ORDERED.
7 Dated: September 3, 2026 LAE
8 ee
LAUREL BEELER
9 United States Magistrate Judge
10
1]
as 12
«14
16
17
Oo
Z 18
19
20
21
22
23
24
25
26
27
28