Roberto Cohen, et al v. Nvidia Corp., et al

Filing

FILED OPINION (RICHARD C. TALLMAN, SANDRA S. IKUTA and BEVERLY REID O'CONNELL) AFFIRMED. Judge: BRO Authoring. FILED AND ENTERED JUDGMENT. [9262516]

Case: 11-17708

10/02/2014

ID: 9262516

DktEntry: 44-1

FOR PUBLICATION

UNITED STATES COURT OF APPEALS
FOR THE NINTH CIRCUIT

IN RE: NVIDIA CORPORATION
SECURITIES LITIGATION,

ROBERTO COHEN; NEW JERSEY
CARPENTERS PENSION AND ANNUITY
FUNDS, on behalf of themselves and
all others similarly situated,
Plaintiffs-Appellants,

No. 11-17708
D.C. No.
3:08-cv-04260RS
OPINION

v.
NVIDIA CORP.; JEN-HSUN HUNAG;
MARVIN D. BURKETT,
Defendants-Appellees.

Appeal from the United States District Court
for the Northern District of California
Richard Seeborg, District Judge, Presiding
Argued and Submitted
January 14, 2014—San Francisco, California
Filed October 2, 2014

Page: 1 of 41

Case: 11-17708

10/02/2014

2

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.
Before: Richard C. Tallman and Sandra S. Ikuta, Circuit
Judges, and Beverly Reid O’Connell, District Judge.*
Opinion by Judge O’Connell

SUMMARY**

Securities Fraud
The panel affirmed the district court’s dismissal of a
securities fraud action against NVIDIA Corp., a publicly
traded semiconductor company, and other defendants under
§§ 10(b) and 20(a) of the Securities Exchange Act of 1934
and Securities Exchange Commission Rule 10b-5.
The amended complaint alleged that NVIDIA should
have informed investors of product defects earlier and that
absent such a disclosure, the company’s intervening
statements regarding its financial condition were misleading
to investors.
The panel held that the plaintiffs failed adequately to
allege scienter by stating with particularity facts giving rise
to a strong inference that the defendants acted with the
required state of mind, as required by the Private Securities
Litigation Reform Act, because they intentionally misled

*

The Honorable Beverly Reid O’Connell, United States District Judge
for the Central District of California, sitting by designation.
**

This summary constitutes no part of the opinion of the court. It has
been prepared by court staff for the convenience of the reader.

Page: 2 of 41

Case: 11-17708

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Page: 3 of 41

3

investors, or were at least deliberately reckless. Agreeing
with the Third Circuit, the panel held that the district court
did not err by failing to consider plaintiffs’ allegations of
scienter in the context of Item 303 of Regulation S-K, C.F.R.
§ 229.303, because Item 303’s disclosure duty is not
actionable under § 10(b) and Rule 10b-5. The panel held that
none of the plaintiffs’ allegations of scienter created a strong
inference of scienter individually and that, together, they did
not give rise to a strong inference of scienter holistically. The
panel concluded that neither the corporate scienter doctrine
nor the core operations doctrine supported a strong inference
of scienter.

COUNSEL
David Brower (argued), Brower Piven, New York, New
York, for Plaintiffs-Appellants.
James Kramer (argued) and Michael Torpey, Orrick,
Herrington & Sutcliffe, LLP, San Francisco, California, for
Defendants-Appellees.

OPINION
O’CONNELL, District Judge:
This case involves allegations of securities fraud.
Defendant NVIDIA Corporation is a publicly traded
semiconductor company. In the spring of 2008, it disclosed
to investors information about defects in two of its products.
A little over one month later, it further disclosed that it would
be taking a $150–$200 million charge to cover costs arising

Case: 11-17708

4

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

from those product defects. As a result, NVIDIA’s share
price dropped 31% and its market capitalization contracted by
$3 billion. According to Plaintiffs, who had purchased
NVIDIA’s stock in the preceding eight months, the company
knew it would be liable for the defective products long before
its 2008 disclosures. They claim that NVIDIA should have
informed investors about the defects as early as November
2007. They further contend that, absent a disclosure about
the product defects, NVIDIA’s intervening statements
regarding its financial condition were misleading to investors,
and consequently in violation of Section 10(b) of the
Securities Exchange Act of 1934 and corresponding
Securities Exchange Commission (“SEC”) Rule 10b-5.
The district court below dismissed Plaintiffs’ amended
complaint without further leave to amend, holding that it
failed to adequately allege scienter, a necessary element for
a claim under either Section 10(b) or Rule 10b-5. We have
jurisdiction under 28 U.S.C. § 1291.
On appeal, Plaintiffs essentially raise three distinct
arguments, all directed to the element of scienter. First, they
argue that the disclosure duty under Item 303 of Regulation
S–K, 17 C.F.R. § 229.303, is actionable under Section 10(b)
and Rule 10b-5. A proper analysis, they contend, should
ascertain whether Defendants acted with scienter in violating
Item 303’s disclosure duty. Second, Plaintiffs assert that the
district court failed to consider their allegations holistically.
They contend that, when considered holistically, their
allegations give rise to a strong inference of scienter. Third,
Plaintiffs argue that the district court erred in finding that
neither the corporate scienter doctrine nor the core operations
doctrine supports a strong inference of scienter.

Page: 4 of 41

Case: 11-17708

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Page: 5 of 41

5

For the reasons discussed below, we affirm.
I
A.
NVIDIA Corporation is a publicly traded semiconductor
company founded in 1993 by Jen-Hsun Huang, its current
CEO. Its core business involves the design and sale of two
similar semiconductor chips. One is a graphics processing
unit (“GPU”); the other is a media and communications
processor (“MCP”). In essence, GPUs are designed to
process the vast amount of data necessary to render images to
a computer’s visual display. MCPs are similar to GPUs in
that they function as a GPU in addition to various other
devices, such as a system memory interface, Ethernet
communications controller, and audio signal processor.
Original equipment manufacturers (“OEMs”), such as
Hewlett-Packard (“HP”) and Dell Computer (“Dell”),
purchase these chips and incorporate them into the
motherboards of computers they assemble and sell to
consumers.
In addition to their similar functions, GPUs and MCPs
also share a similar configuration, which comprises two main
parts: (1) a “die,” or the silicon chip itself; and (2) a
“substrate,” or wafer, which is a green circuit board that
ultimately connects the die to the motherboard’s electrical
components. To manufacture the GPUs and MCPs, the die is
mounted onto the substrate.
Importantly, the die
electronically connects to the substrate through “bumps” of
solder that relay electrical signals between the die and the rest
of the computer. The bumps are attached to the substrate
using a solder paste. Between the die and substrate is an

Case: 11-17708

6

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

“underfill,” which is a glue-like material that acts as an
additional bonding agent to fortify the connection between
the die and substrate. Together, the solder and underfill are
referred to as the “Material Set.”
Given the highly complex and technical nature of
NVIDIA’s GPU and MCP products, there is an inherent risk
that some will fail. As a result, NVIDIA routinely includes
in its SEC forms a statement explaining that “[its] products
may contain defects or flaws,” and warning investors that
“[it] may be required to reimburse customers for costs to
repair or replace the affected products.” To cover costs
relating to inevitable defects, NVIDIA automatically records
a reduction to revenue as a cash reserve. As product return
and replacement costs accrue, NVIDIA withdraws cash from
that reserve.
B.
According to the complaint, in September 2006, NVIDIA
began experiencing problems with certain of its GPU and
MCP products, particularly with those products’ Material Set.
Plaintiffs allege that some of NVIDIA’s chips experienced
cracks in the solder bumps when subjected to excessive
pressure during product testing. At that time, NVIDIA had
been using a “eutectic” solder1 (which has a relatively low
lead content) together with eutectic solder paste. In an
attempt to remedy the cracking problem, NVIDIA switched
some of the solders used in the chips from a eutectic solder to

1

Solder is a compound mixture of lead and tin. A particular mixture has
a “eutectic” composition if it has a specific ratio of lead to tin. See Donald
Askeland & Wendelin Wright, Essentials of Materials Science &
Engineering 359–63 (2013).

Page: 6 of 41

Case: 11-17708

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Page: 7 of 41

7

a high-lead solder, which is more malleable and therefore less
susceptible to cracking from the pressure in product testing.
It continued to use the eutectic solder paste, however.
According to Plaintiffs, varying thermal properties of the
new, high-lead solder and the eutectic solder paste
contributed to new problems with NVIDIA’s chips.
Specifically, because the two materials undergo thermal
expansion at varying rates, the high-lead solder is susceptible
to fatigue and cracking over time.
At some point, these new problems began manifesting in
laptop computers incorporating NVIDIA’s GPU and MCP
products that were made using high-lead solder. After HP
(and later Dell) began investigating these problems, it
observed new cracking of the solder bumps connecting the
die to the substrate (the “Material Set Problem”). At first,
NVIDIA attributed the problem to “‘customer-induced
damage or [OEM] design issues.’” HP hypothesized that heat
cycling was the root cause of the problem.2 Specifically, HP
believed that the solder bumps would weaken over time due
to repeated thermal expansion caused by heat cycling.
To reduce the stress on the chips’ solder bumps, and thus
ameliorate the cracking problem, HP and Dell, with the help
of NVIDIA, issued software updates (“BIOS”3 updates) to
their laptop computers. These BIOS updates altered a

2

Heat cycling is a fluctuation of a computer’s internal temperature. As
the computer’s internal components generate heat from usage, the internal
temperature rises. When the computer’s sensors detect a certain,
preprogrammed temperature, the computer’s program activates internal
fans to lower the temperature.
3

BIOS stands for Basic Input/Output System.

Case: 11-17708

8

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

computer’s fan algorithm, causing the internal cooling fans to
run continuously, thereby eliminating heat cycling.
Evidently, HP believed that by maintaining a fairly constant
temperature, the solder bumps would not undergo thermal
expansion as often and thus not be as susceptible to fatigue
and failure.
Ultimately, after significant testing, HP concluded that the
root cause of GPU and MCP failures in its computers was not
caused by cracking due to heat cycling, but by cracking due
to operation of the chips within a narrow temperature range.
Apparently, the stress on the solder bumps caused by varying
thermal properties of the high-lead solder and eutectic solder
paste was especially acute in this temperature range. HP
shared with NVIDIA its data demonstrating this problematic
thermal profile, and, at some point, NVIDIA reproduced the
data in its own laboratories.
In May and June 2008, NVIDIA issued to its OEM
customers Product Change Notifications (“PCNs”), indicating
that it would be transitioning back to eutectic solder.
C.
Between November 8, 2007, and May 22, 2008, NVIDIA
filed several forms with the SEC, as required by law.
According to Plaintiffs, those forms contained materially
false and misleading statements, principally because they
omitted information regarding the Material Set Problem.
For example, in the November 8, 2007 Form 8-K,
Plaintiffs point to NVIDIA’s claim that “[its] core businesses
are continuing to grow as the GPU becomes increasingly
central to today’s computing experience.” In NVIDIA’s

Page: 8 of 41

Case: 11-17708

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Page: 9 of 41

9

February 13, 2008 Form 8-K, it highlights the assertion that
“Fiscal 2008 was another outstanding and record year for us.
Strong demand for GPUs in all market segments drove our
growth.” Plaintiffs argue that these statements and others
made in NVIDIA’s March 21, 2008, Form 10-K and May 8,
2008, Form 8-K are materially false and misleading because
NVIDIA failed to disclose reported defects in its products as
well.
D.
On May 22, 2008, NVIDIA disclosed in its quarterly
report that it had received claims for reimbursement from one
of its OEMs for incremental costs due to an “‘alleged
die/packaging material set defect.’” The report also indicated
that the product was included in a significant number of the
customer’s computer products and had been shipped to other
customers in significant quantities. NVIDIA explained that
it was “evaluating the potential scope” of the problem “and
cause of the alleged defect and the merits of the customer’s
claim.” It further indicated that it was “unable to estimate the
amount of costs that may be incurred” at that time.
Just over one month later, on July 2, 2008, NVIDIA filed
an SEC Form 8-K indicating it would be taking “a $150 to
$200 million charge to cover warranty, repair, return,
replacement, and other costs ‘arising from a weak
die/packaging material set in certain versions of [its] previous
MCP and GPU products used in notebook systems.’”4 After

4

Nevertheless, the form also stated that NVIDIA had “not been able to
determine a root cause for these failures, [but] testing suggests a weak
material set of die/package combination, system thermal management
designs, and customer use patterns are contributing factors.”

Case: 11-17708

10

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

NVIDIA’s July 2, 2008 disclosure, the market reacted
accordingly, causing a 31% decline in NVIDIA’s share price
and a decrease of over $3 billion in its market capitalization.
E.
Plaintiffs invested in NVIDIA’s stock between November
8, 2007 and July 2, 2008 (the “class period”). They allege
that, beginning November 8, 2007, NVIDIA knew of the
defect in the GPU and MCP Material Set, this knowledge was
material to investors, and failure to disclose it made other
statements in NVIDIA’s SEC filings misleading.
Believing that Defendants violated federal securities laws,
Plaintiffs filed three separate lawsuits, which the district court
consolidated into a single action. In the consolidated
complaint, Plaintiffs allege three distinct but related counts.
In the first and second counts, they allege that NVIDIA and
Huang, respectively, are liable for violations of both Section
10(b) of the Securities Exchange Act of 1934 and
corresponding SEC Rule 10b-5. In the third count, they aver
that Huang is further liable for violations of Section 20(a) of
the Securities Exchange Act of 1934.
Upon Defendants’ motion, the district court dismissed
Plaintiffs’ first consolidated class action complaint with leave
to amend. Plaintiffs then filed a second consolidated class
action complaint. Upon a second motion by Defendants, the
district court dismissed that complaint without leave to
amend. In its order of dismissal, the district court specifically
held that Plaintiffs failed to sufficiently plead scienter, an
element required for each count.

Page: 10 of 41

Case: 11-17708

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Page: 11 of 41

11

II
We review dismissals under Federal Rule of Civil
Procedure 12(b)(6) de novo. In re Daou Sys., Inc., 411 F.3d
1006, 1013 (9th Cir. 2005). In doing so, we accept as true all
factual allegations and determine whether they are sufficient
to state a claim for relief; we do not, however, accept as true
allegations that are conclusory. Ashcroft v. Iqbal, 556 U.S.
662, 678 (2009). Moreover, “[f]actual allegations must be
enough to raise a right to relief above the speculative level.”
Bell Atl. Corp. v. Twombly, 550 U.S. 544, 555 (2007). There
must be “more than a sheer possibility that a defendant has
acted unlawfully.” Iqbal, 556 U.S. at 678.
In reviewing the sufficiency of a complaint, we limit
ourselves to the complaint itself and its attached exhibits,
documents incorporated by reference, and matters properly
subject to judicial notice. Tellabs, Inc. v. Makor Issues &
Rights, Ltd., 551 U.S. 308, 322–23 (2007); Lee v. City of Los
Angeles, 250 F.3d 668, 688 (9th Cir. 2001).
III
A.
1.
Section 10(b) of the Securities Exchange Act of 1934
declares it unlawful to “use or employ, in connection with the
purchase or sale of any security . . . , any manipulative or
deceptive device or contrivance in contravention of such rules
and regulations as the [SEC] may prescribe as necessary.”
15 U.S.C. § 78j(b). As the Supreme Court has indicated,
there is an “implied [] private cause of action” in

Case: 11-17708

12

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Section 10(b). Matrixx Initiatives, Inc. v. Siracusano, 131 S.
Ct. 1309, 1317 (2011). Additionally, “SEC Rule 10b-5
implements [Section 10(b)] by making it unlawful to . . .
‘make any untrue statement of a material fact or to omit to
state a material fact necessary in order to make the statements
made . . . not misleading.’” Id. (quoting 17 C.F.R. § 240.10b5). Thus, to prevail on a claim for violations of either Section
10(b) or Rule 10b-5, a plaintiff must prove six elements:
“(1) a material misrepresentation or omission by the
defendant; (2) scienter; (3) a connection between the
misrepresentation or omission and the purchase or sale of a
security; (4) reliance upon the misrepresentation or omission;
(5) economic loss; and (6) loss causation.” Stoneridge Inv.
Partners, LLC v. Scientific-Atlanta, Inc., 552 U.S. 148, 157
(2008).
Section 20(a) of the Securities Exchange Act of 1934
provides for liability of a “controlling person.” 15 U.S.C.
§ 78t(a). To establish a cause of action under this provision,
a plaintiff must first prove a primary violation of underlying
federal securities laws, such as Section 10(b) or Rule 10b-5,
and then show that the defendant exercised actual power over
the primary violator. Howard v. Everex Sys., Inc., 228 F.3d
1057, 1065 (9th Cir. 2000). Accordingly, our analysis
focuses on Plaintiffs’ allegations under Section 10(b) and
Rule 10b-5.
2.
When alleging violations of federal securities laws, a
plaintiff must satisfy the pleading requirements pronounced
in the Private Securities Litigation Reform Act of 1995
(“PSLRA”). Zucco Partners, LLC v. Digimarc Corp.,
552 F.3d 981, 990–91 (9th Cir. 2009). In passing this act,

Page: 12 of 41

Case: 11-17708

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Page: 13 of 41

13

Congress “significantly altered pleading requirements in
securities fraud cases [by] . . . requir[ing] that a complaint
plead with particularity both falsity and scienter.” Id. at 990
(internal citations and quotation marks omitted). More
precisely, now a complaint must “state with particularity facts
giving rise to a strong inference that the defendant acted with
the required state of mind.” 15 U.S.C. § 78u-4(b)(2)(A).
To assist courts in determining whether a plaintiff has
satisfied this heightened pleading standard, the Supreme
Court has provided three points of instruction: (1) “courts
must, as with any [12(b)(6)] motion to dismiss . . . , accept all
factual allegations in the complaint as true”; (2) “courts must
consider the complaint in its entirety, as well as other sources
courts ordinarily examine when ruling on Rule 12(b)(6)
motions to dismiss”; and (3) “in determining whether the
pleaded facts give rise to a ‘strong’ inference of scienter, the
court must take into account plausible opposing inferences.”
Tellabs, 551 U.S. at 322–23. In discussing the third point, the
Court explained that, although “[t]he inference [of scienter]
need not be irrefutable, . . . or even the ‘most plausible of
competing inferences,’” it “must be more than merely
‘reasonable’ or ‘permissible’—it must be cogent and
compelling, thus strong in light of other [countervailing]
explanations.” Id. at 324 (citations omitted). Ultimately, the
Court held that “[a] complaint will survive . . . only if a
reasonable person would deem the inference of scienter
cogent and at least as compelling as any opposing inference
one could draw from the facts alleged.” Id.
3.
Having outlined the lens through which we must consider
Plaintiffs’ allegations of scienter, we now turn to the

Case: 11-17708

10/02/2014

14

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

definition of scienter itself. In Ernst & Ernst v. Hochfelder,
the Supreme Court explained in a footnote that “the term
‘scienter’ refers to a mental state embracing intent to deceive,
manipulate, or defraud.” 425 U.S. 185, 193 n.12 (1976). The
Court recognized that some Courts of Appeals include within
their definition of scienter a form of recklessness, but it did
not address whether those courts are correct in doing so. Id.
As recently as in its decision in Matrixx Initiatives, the Court
stated that it “ha[s] not [yet] decided whether recklessness
suffices to fulfill the scienter requirement.” 131 S. Ct. at
1323.
In this circuit, we have “held that recklessness may satisfy
the element of scienter.” Hollinger v. Titan Capital Corp.,
914 F.2d 1564, 1568–69 (9th Cir. 1990) (en banc). We
defined recklessness “‘as a highly unreasonable omission,
involving . . . an extreme departure from the standards of
ordinary care, and which presents a danger of misleading
buyers or sellers that is either known to the defendant or is so
obvious that the actor must have been aware of it.’” Id. at
1569 (quoting Sundstrand Corp. v. Sun Chem. Corp.,
553 F.2d 1033, 1044–45 (7th Cir. 1977)).5 We added that
“the danger of misleading buyers must be actually known or
so obvious that any reasonable man would be legally bound
as knowing.” Id. at 1569–70 (internal quotation marks
omitted). In In re Silicon Graphics Inc. Securities Litigation,
we opined that the wording of our previous decisions
discussing recklessness—including our decision in
Hollinger—indicates that the standard for recklessness is
actually much closer to one of intent: “These cases indicate
that recklessness only satisfies scienter under § 10(b) to the
5

In Hollinger, we expressly “adopt[ed] the standard of recklessness
articulated by the Seventh Circuit in Sundstrand Corp.” 914 F.2d at 1569.

Page: 14 of 41

Case: 11-17708

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Page: 15 of 41

15

extent that it reflects some degree of intentional or conscious
misconduct.”
183 F.3d 970, 977 (9th Cir. 1999).
Accordingly, we explained that scienter requires “a strong
inference of, at a minimum, ‘deliberate recklessness.’” Id.
(emphasis added).6 Since then, we have consistently applied
the “deliberate recklessness” terminology and standard
articulated in Silicon Graphics. See, e.g., In re VeriFone
Holdings, Inc. Sec. Litig., 704 F.3d 694, 702 (9th Cir. 2012);
Zucco Partners, 552 F.3d at 991; In re Daou Sys., Inc., 411
F.3d at 1022.7
We now address Plaintiffs’ arguments on appeal.

6

It appears that the term “deliberate recklessness” was coined by the
district court that we affirmed in Silicon Graphics. Our opinion in that
case was the first time we ever used the term. The only other appellate
court to have used the term previously did so less than one month before
we did, and it cited to the district court’s decision that we affirmed, In re
Silicon Graphics, Inc. Securities Litigation, 970 F. Supp. 746 (N.D. Cal.
1997). See In re Advanta Corp. Sec. Litig., 180 F.3d 525, 530 (3d Cir.
1999).
7

In their reply brief, Plaintiffs cite In re Oracle Corp. Securities
Litigation, 627 F.3d 376, 390 (9th Cir. 2010), and argue that the standard
for deliberate recklessness does not include an element of intent or
conscious misconduct. Oracle, however, concerned the scienter standard
for summary judgment, and we had previously held that “[b]ecause the
PSLRA did not alter the substantive requirements for scienter under
§ 10(b), [] the standard [for establishing scienter] on summary judgment
or JMOL remains unaltered by In Re Silicon Graphics.” Howard,
228 F.3d at 1064. Unlike Oracle or Howard, this case comes to us
following a motion to dismiss. We therefore apply the standard set forth
in In re Silicon Graphics.

Case: 11-17708

10/02/2014

16

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.
B.

Plaintiffs first argue that the district court erred by failing
to consider their allegations of scienter in the context of Item
303 of Regulation S–K, 17 C.F.R. § 229.303.8 They correctly
assert that Item 303 requires disclosure of certain
information. But then they contend that, if the information is
material, failure to disclose it constitutes a material omission
for purposes of Section 10(b) and Rule 10b-5. Ultimately,
Plaintiffs argue that the district court’s analysis should have
focused on whether NVIDIA acted with scienter in failing to
make the Item 303 disclosure.
We have never directly decided whether Item 303’s
disclosure duty is actionable under Section 10(b) and Rule
10b-5. We now hold that it is not.
To prevail on a claim under Section 10(b) or Rule 10b-5,
a plaintiff must demonstrate that the defendant made a
misleading statement or omission of material fact. Matrixx
Initiatives, 131 S. Ct. at 1318. Thus, a statement might be
misleading because it affirmatively misstates information. Or
a statement might be misleading because it is made outside

8

Item 303, 17 C.F.R. § 229.303(a)(3)(ii), requires registrants to:
Describe any known trends or uncertainties that have
had or that the registrant reasonably expects will have
a material favorable or unfavorable impact on net sales
or revenues or income from continuing operations. If
the registrant knows of events that will cause a material
change in the relationship between costs and revenues
(such as known future increases in costs of labor or
materials or price increases or inventory adjustments),
the change in the relationship shall be disclosed.

Page: 16 of 41

Case: 11-17708

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Page: 17 of 41

17

the context of other material information. Yet neither
Section 10(b) nor Rule 10b-5 “create[s] an affirmative duty
to disclose any and all material information. Disclosure is
required under these provisions only when necessary ‘to
make . . . statements made, in the [sic] light of the
circumstances under which they were made, not
misleading.’” Id. at 1321–22 (alteration in original) (quoting
17 C.F.R. § 240.10b-5). In Basic Inc. v. Levinson, the
Supreme Court noted that “[s]ilence, absent a duty to
disclose, is not misleading under Rule 10b-5.” 485 U.S. 224,
239 n.17 (1988). The Court did not explain what would give
rise to a duty to disclose, but it is this language in Basic that
Plaintiffs cite in support of their argument. They assert that
Item 303 creates “a duty to disclose,” and failure to disclose
it is therefore misleading for purposes of Section 10(b) and
Rule 10b-5.
We have confronted a similar argument before. See In re
VeriFone Sec. Litig., 11 F.3d 865, 870 (9th Cir. 1993); In re
Lyondell Petrochemical Co. Sec. Litig., 984 F.2d 1050,
1053 (9th Cir. 1993); In re Convergent Techs. Sec. Litig.,
948 F.2d 507, 516 (9th Cir. 1991). In each instance, we
strongly suggested that a violation of Item 303 cannot be used
to show a violation of Section 10(b) and Rule 10b-5. We
noted that, “[w]hile § 229.303(a)(3)(ii) provides that ‘known
trends or uncertainties’ be disclosed in certain SEC filings,”
Instruction 7 to Item 303(a) “states that forward-looking
information need not be disclosed.” In re VeriFone, 11 F.3d
at 870. Without further discussion, we rejected the plaintiffs’
argument.9

9

Citing In re VeriFone, we recently noted that failure to comply with
Regulation S–K is insufficient for a claim under Rule 10b-5. Police Ret.

Case: 11-17708

18

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

In Oran v. Stafford, the Third Circuit decided this issue
more directly. 226 F.3d 275, 287–88 (3d Cir. 2000). We are
persuaded by its reasoning. There, the court explained that
Item 303’s disclosure requirement “varies considerably from
the general test for securities fraud materiality set out by the
Supreme Court in Basic Inc. v. Levinson.” Id. at 288. In
relevant part, Item 303 requires corporate management to
“[d]escribe [in 10-K and 10-Q forms] any known trends or
uncertainties that have had or that the registrant reasonably
expects will have a material favorable or unfavorable impact
on net sales or revenues or income from continuing
operations.” 17 C.F.R. § 229.303(a)(3)(ii). The SEC shed
further light on this requirement in an interpretive release:
Where a trend, demand, commitment, event or
uncertainty is known, management must make
two assessments:
(1) Is the known trend, demand, commitment,
event or uncertainty likely to come to
fruition? If management determines that it is
not reasonably likely to occur, no disclosure is
required.
(2) If management cannot make that
determination, it must evaluate objectively the
consequences of the known trend, demand,
commitment, event or uncertainty, on the
assumption that it will come to fruition.
Disclosure is then required unless
management determines that a material effect
Sys. of St. Louis v. Intuitive Surgical, Inc., 759 F.3d 1051, 1061 n.4 (9th
Cir. 2014).

Page: 18 of 41

Case: 11-17708

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Page: 19 of 41

19

on the registrant’s financial condition or
results of operations is not reasonably likely
to occur.
Management’s Discussion and Analysis of Financial
Condition and Results of Operations, Exchange Act Release
No. 34-26831, 54 Fed. Reg. 22427, 22430 (May 24, 1989).
On the other hand, in Basic, the Supreme Court stated that
materiality of forward-looking information depends “upon a
balancing of both the indicated probability that the event will
occur and the anticipated magnitude of the event in light of
the totality of the company activity.” 485 U.S. at 238
(internal quotation marks omitted).
As the court in Oran also determined, these two standards
differ considerably. 226 F.3d at 288. Management’s duty to
disclose under Item 303 is much broader than what is
required under the standard pronounced in Basic. The SEC
intimated this point as well: “[Item 303] mandates disclosure
of specified forward-looking information, and specifies its
own standard for disclosure—i.e., reasonably likely to have
a material effect. . . . The probability/magnitude test for
materiality approved by the Supreme Court in Basic, Inc., v.
Levinson, 108 S. Ct. 978 (1988), is inapposite to Item 303
disclosure.” Exchange Act Release No. 34-26831, 54 Fed.
Reg. at 22430 n.27. The SEC’s effort to distinguish Basic’s
materiality test from Item 303’s disclosure requirement
provides further support for the position that Item 303
requires more than Basic—what must be disclosed under Item
303 is not necessarily required under the standard in Basic.
Therefore, “[b]ecause the materiality standards for Rule 10b5 and [Item 303] differ significantly, the ‘demonstration of a
violation of the disclosure requirements of Item 303 does not
lead inevitably to the conclusion that such disclosure would

Case: 11-17708

20

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

be required under Rule 10b-5. Such a duty to disclose must
be separately shown.’” Oran, 226 F.3d at 288.
Plaintiffs’ reliance on Litwin v. Blackstone Group, L.P.,
634 F.3d 706 (2d Cir. 2011), and Panther Partners Inc. v.
Ikanos Communications, Inc., 681 F.3d 114 (2d Cir. 2012), is
unavailing. Those cases involved alleged violations of
Sections 11 and 12(a)(2) of the Securities Act of 1933, not
Section 10(b) or Rule 10b-5. And, as we acknowledged in
Steckman v. Hart Brewing, Inc., “Section 10(b) of the
Exchange Act . . . differs significantly from Sections 11 and
12(a)(2) of the Securities Act.” 143 F.3d 1293, 1296 (9th Cir.
1998). Liability under Sections 11 and 12(a)(2) of the
Securities Act may arise from “omitt[ing] to state a material
fact required to be stated.” See 15 U.S.C. §§ 77k(a), 77l(b).
To put it differently, liability arises from “an omission in
contravention of an affirmative legal disclosure obligation.”
Panther Partners, 681 F.3d at 120. There is no such
requirement under Section 10(b) or Rule 10b-5. As discussed
above, for purposes of Section 10(b) and Rule 10b-5, material
information need not be disclosed unless omission of that
information would cause other information that is disclosed
to be misleading. Matrixx Initiatives, 131 S. Ct. at 1321.
Furthermore, as noted in Panther Partners, scienter is not an
element of either a Section 11 or Section 12(a)(2) claim.
681 F.3d at 120. Such claims are not subject to the PSLRA’s
heightened pleading standards unless based on allegations of
fraud. Id. Accordingly, neither Panther Partners nor Litwin
affects our analysis here.
Also unavailing is Plaintiffs’ reliance on Simon v.
American Power Conversion Corp., 945 F. Supp. 416 (D.R.I.
1996). In that case, then-Chief Judge Laguex opined that
Item 303 imposes “an affirmative duty to disclose,” and

Page: 20 of 41

Case: 11-17708

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Page: 21 of 41

21

found that the defendant’s failure to disclose in that case was
actionable. Simon, 945 F. Supp. at 431. Subsequently,
however, Judge Laguex clarified his opinion: “As this Court
noted in Simon, the disclosure rules are probative of what
defendants are otherwise obliged to disclose but do not,
themselves, provide an independent duty of disclosure.”
Kafenbaum v. GTECH Holdings Corp., 217 F. Supp. 2d 238,
249 (D.R.I. 2002). He went on to say that “plaintiffs may not
rely solely upon Item 303 to prove that defendants failed to
disclose material information as a matter of law [in violation
of Rule 10b-5].” Id. at 250.
In sum, we hold that Item 303 does not create a duty to
disclose for purposes of Section 10(b) and Rule 10b-5. Such
a duty to disclose must be separately shown according to the
principles set forth by the Supreme Court in Basic and
Matrixx Initiatives.
C.
Next, Plaintiffs contend that the district court erred by not
considering their allegations of scienter holistically.
In Tellabs, the Supreme Court explained that, when
assessing allegations of scienter, a “court’s job is not to
scrutinize each allegation in isolation but to assess all the
allegations holistically . . . . [T]he reviewing court must ask:
When the allegations are accepted as true and taken
collectively, would a reasonable person deem the inference of
scienter at least as strong as any opposing inference?”
551 U.S. at 326 (citations omitted). We apply this same
standard but in a dual inquiry. First, we determine whether
any of the plaintiff’s allegations, standing alone, is sufficient
to create a strong inference of scienter. Zucco Partners,

Case: 11-17708

22

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

552 F.3d at 992. If none is sufficient alone, we then consider
the allegations holistically to determine whether they create
a strong inference of scienter taken together. Id.
We have reviewed Plaintiffs’ allegations in their entirety.
We find that none creates a strong inference of scienter
individually. And, together, they do not give rise to a strong
inference of scienter holistically. The most compelling
inference that we can reasonably draw is that NVIDIA was
first investigating the root cause, and then the scope, of the
Material Set Problem; once it determined that its liability
would exceed its normal reserves, NVIDIA disclosed the
problem to investors. Any inference of scienter requires more
than this. Thus, while the complaint may plausibly allege
knowledge of the Material Set Problem, it does not plausibly
allege that NVIDIA and Huang intentionally misled investors,
or acted with deliberate recklessness, by not disclosing the
problem sooner. We discuss our reasoning below.
Plaintiffs contend that NVIDIA acted with scienter in
withholding from investors certain information regarding the
Material Set Problem. They allege that, between November
2007 and July 2008, NVIDIA made various statements in its
SEC forms regarding its financial performance and the
quality of its products. Plaintiffs argue that those statements
were materially false and misleading because NVIDIA did
not also disclose that certain of its products were failing at an
abnormal rate and that NVIDIA ultimately would be
financially responsible for replacement costs. According to
Plaintiffs, NVIDIA first determined that it would be
financially liable for the chip failures by the middle of 2007;
nevertheless, it did not disclose this to investors until July
2008. From these facts, Plaintiffs infer that NVIDIA
intentionally misled investors by waiting to disclose its

Page: 22 of 41

Case: 11-17708

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Page: 23 of 41

23

liability for almost an entire year, until it had prepared
replacement products.
In evaluating Plaintiffs’ inference of scienter, we bear in
mind that NVIDIA includes in its SEC forms a statement
explaining that “[its] products may contain defects or flaws”
and warning investors that “[it] may be required to reimburse
customers for costs to repair or replace the affected
products.” Moreover, we are mindful that, because product
defects are so common, NVIDIA automatically records a
reduction to revenue as a cash reserve to cover costs relating
to the inevitable product failures. Bearing these facts in
mind, we must determine whether Plaintiffs’ allegations
create a strong inference of scienter that Huang and NVIDIA
intentionally misled investors, or were at least deliberately
reckless, by not disclosing NVIDIA’s liability for chip
failures prior to July 2008.
1.
Plaintiffs allege that NVIDIA first became aware of the
Material Set Problem sometime in 2006, and then determined
the root cause of the problem—and therefore that it would be
financially responsible for it—by the middle of 2007. Then
they contend that the one-year delay between NVIDIA’s root
cause determination in mid-2007 and its disclosure of the
problem in 2008 was motivated by an intent to mislead
investors until it had prepared replacement products.
Nevertheless, Plaintiffs’ allegations that NVIDIA first
determined the root cause of its chip failures by the middle of
2007 are implausible. In so alleging, Plaintiffs rely most
directly on the accounts of Confidential Witness No. 1
(“CW1”), Confidential Witness No. 7 (“CW7”), and articles
written by Charlie Demerjian.

Case: 11-17708

10/02/2014

24

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

According to CW1, “‘communications between HP and
NVIDIA began in 2006, with HP asking questions, and
asking NVIDIA to conduct [a] failure analysis.’” When
NVIDIA initially resisted, HP began to engage it at a higher,
executive, level. By early 2007, HP had determined that
NVIDIA’s chips sustained physical damage through normal
use, and therefore making NVIDIA responsible. HP
developed “‘overwhelming data demonstrating root cause’”
and “identified the thermal profile” that would cause
NVIDIA’s chips to experience cracking at the solder bumps.
CW1 asserts that HP shared the data and thermal profile with
NVIDIA sometime in early 2007 and NVIDIA reproduced it
by the middle of 2007.
The timing of CW1’s account conflicts with the account
of Richard Hunt Hodge, HP’s Director of Engineering and
Quality for the Notebook Division.10 According to Hodge,
HP itself did not determine the thermal profile that would
significantly increase the probability of chip failures until the
middle of 2008: “Ultimately, by mid-2008, HP determined
that operation of the NVIDIA part through a narrow
temperature range . . . was the cause of” chip failures due to
solder bump cracking. If HP did not determine the
problematic thermal profile until the middle of 2008, any
inference of scienter is significantly reduced, as NVIDIA
disclosed the information in July 2008.
10

Because Plaintiffs incorporate by reference Mr. Hunt’s declaration,
relying on portions of it in their complaint, we may properly consider the
declaration in its entirety. See Tellabs, 551 U.S. at 322; see also City of
Roseville Emps.’ Ret. Sys. v. Sterling Fin. Corp., 963 F. Supp. 2d 1092,
1107 (E.D. Wash. 2013) (“Once a document is deemed incorporated by
reference, the entire document is assumed to be true for purposes of a
motion to dismiss, and both parties—and the Court—are free to refer to
any of its contents.”).

Page: 24 of 41

Case: 11-17708

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Page: 25 of 41

25

The notion that HP had determined the problematic
thermal profile in early 2007 is implausible when considered
together with Plaintiffs’ other allegations. Plaintiffs allege
that HP and Dell, with the help of NVIDIA, issued BIOS
updates in November 200711 and February 2008, respectively.
Those BIOS updates altered a computer’s fan algorithm so
that its internal cooling fans would run continuously, thereby
eliminating heat cycling and maintaining a fairly constant
temperature inside the computer. Hodge explains in his
declaration that the HP BIOS update “keeps [a GPU or MCP]
outside of [the problematic] temperature range (usually
around 50ºC at idle). Because it runs constantly, it obviates
the thermal mini-cycle through the problematic narrow
temperature range.” If HP had determined the problematic
thermal profile in early 2007, as Plaintiffs and CW1 allege, it
is implausible that HP would have waited until November
2007 to issue a BIOS update that greatly reduces the
probability of a chip failure due to solder bump cracking.12
The implausibility of the timing in CW1’s account of events
further detracts from any inference of scienter.

11
According to Hodge, HP’s BIOS update “was released on or about on
December 3, 2007.”
12

Indeed, Hodge’s timing of events is much more plausible than CW1’s.
He explains that HP hypothesized in November 2007 “that repeated
temperature mini-cycling could result in solder cracks . . . and believed
eliminating that temperature cycling would correct the issue.”
Accordingly, HP developed the BIOS updates to maintain a constant
temperature within its computers. Nevertheless, as HP continued testing,
it discovered that temperature cycling itself would not induce chip
failures, but operating the NVIDIA chip within the specific problematic
thermal profile would induce chip failures. Fortuitously, the BIOS update
maintained NVIDIA’s chips at a constant temperature outside of the
problematic temperature range.

Case: 11-17708

26

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Plaintiffs’ reliance on Charlie Demerjian and CW7 to
corroborate CW1’s timing of events does not restore
plausibility to their allegations. Demerjian is “a reporter with
25 years experience working with computers.” He “covered
NVIDIA’s chip defects and spoke to dozens of people about
the problems.”
Clearly, Demerjian’s information is
secondhand. In any event, Demerjian merely indicates that
“HP started a ‘root cause analysis’ and by summer 2007, HP
employees knew that NVIDIA’s GPUs were having problems
associated with heat cycling.” This account does not indicate
that HP had determined the root cause or problematic thermal
profile by summer 2007. According to CW7, “Dell saw
problems with NVIDIA GPUs in early 2007.” CW7
indicated that NVIDIA “eventually admitted that it was their
chip causing the problem, and that it was not a Dell issue.”
Nevertheless, CW7 does not allege when NVIDIA admitted
responsibility for the problems. Thus, even if Dell had
notified NVIDIA of problems in early 2007, CW7 provides
no basis to infer that NVIDIA also knew it would be liable at
that time.
Even accepting as true CW1’s timing of events, the fact
that NVIDIA had reproduced the problematic thermal profile
by the middle of 2007—and thus had determined the root
cause of the chip failures—does not create an inference of
scienter. Plaintiffs assert that HP provided NVIDIA with
what HP believed demonstrated root cause of, and NVIDIA’s
liability for, the chip failures. Their allegations do not
provide any basis to infer what NVIDIA concluded from the
profile or other data, nor do they plausibly suggest that
NVIDIA must have determined it was at fault at that time.
Furthermore, Plaintiffs never allege that Huang or anyone
else at NVIDIA knew at that time (or any time prior to July
2008) that NVIDIA’s liability would exceed its normal

Page: 26 of 41

Case: 11-17708

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Page: 27 of 41

27

reserve set aside for costs associated with product failures.
Accordingly, these allegations do not give rise to a strong
inference that Huang and NVIDIA acted with intent to
mislead investors, or recklessly disregarded an obvious
danger of misleading investors, by not disclosing information
regarding the Material Set Problem prior to July 2008.
2.
After alleging that HP provided NVIDIA with all of its
data demonstrating root cause and the problematic thermal
profile, Plaintiffs assert that determining the cause of the
problem should have been easy for NVIDIA. They rely on
several sources to support this allegation.
CW1, Demerjian, and John Rigg, Plaintiffs’ consultant in
this case, assert that “failures of the GPU are easily
identifiable.” Confidential Witness No. 2 explains that it
“takes about a month” to determine whether a GPU or MCP
is generating more heat than specified. Hodge indicates that
“it took HP less than two months to hypothesize that
temperature cycling caused the problematic cracks in the
solder bumps.”
But even if GPU failures are easily identifiable, it does
not necessarily follow that NVIDIA would be responsible for
those failures or should have known that it would be
responsible. As NVIDIA initially contended, the failures
could have been caused by OEM design issues. And
Plaintiffs’ reliance on Hodge’s declaration for the contention
that NVIDIA should have easily determined that it was liable
for the chip failures is misplaced. Plaintiffs ignore the timing
of Hodge’s account—HP did not hypothesize temperature
mini-cycling as the root cause until November 2007. Then,

Case: 11-17708

28

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

HP did not begin to test its hypothesis until January 2008.
After 13 weeks of testing, HP determined that temperature
mini-cycling was not the root cause, as its testing did not
induce a single chip failure. It was not until sometime around
the middle of 2008 that HP ultimately discovered the specific
cause of chip failures: operation within the problematic
thermal profile.
Plaintiffs also rely on two articles for the proposition that
NVIDIA must have known that it was going to be liable
sooner than it admitted. Both articles were published after
NVIDIA’s July 2008 disclosure. One article provides
AMD’s (one of NVIDIA’s competitors) opinion on
NVIDIA’s chip failures. “According to the article, ‘AMD
thinks [high-lead bumps are] more prone to fatigue and need
“comprehensive reliability engineering to be used
successfully.”’” AMD also notes that a high-lead solder and
eutectic solder paste have varying thermal expansion
coefficients, which places significant stress on the solder
bumps. The other article discusses a research report written
by K.N. Tu, a professor affiliated with the Department of
Materials Science and Engineering at the University of
California, Los Angeles. In essence, the report discusses the
same issues discussed by AMD.
These articles do not contribute to an inference of
scienter. They were written in hindsight and do not reflect
Huang or NVIDIA’s knowledge prior to July 2008. Simply
because scientists were able to explain in retrospect the
science behind NVIDIA’s chip failures, it does not mean that
NVIDIA knew or should have known it would be liable for
those failures during the class period or that its liability would
exceed its normal reserve.

Page: 28 of 41

Case: 11-17708

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Page: 29 of 41

29

Plaintiffs also point to an email authored by HP’s Richard
Hodge in May 2007, wherein Hodge indicated that NVIDIA
had moved away from eutectic solder in its chips. From this
email, Plaintiffs conclude that there are “known industry
concerns with . . . the use of high-lead solder.” They also
conclude, “Defendants knew and/or deliberately disregarded
that their GPU and MCP problems likely stemmed from this
hasty, under-tested manufacturing change initiated in 2006.”
Nevertheless, there is no basis for this conclusion. Certainly,
Hodge’s email suggests that he had some concern about using
noneutectic solder. But there is nothing more to indicate an
industry-wide concern with noneutectic solder, nor is there
anything to suggest that NVIDIA should have known that its
use of noneutectic solder was the root cause of its chip
failures. Accordingly, this allegation does not appreciably
add to any inference of scienter.
3.
Next, Plaintiffs contend that NVIDIA delayed disclosure
of its chip failures until it had prepared replacement chips.
They support this contention with very few factual
allegations, however.
Plaintiffs rely on statements made by CW1. After HP
provided NVIDIA with data that HP believed demonstrated
the root cause of the chip failures and the problematic thermal
profile that would cause those failures, “‘NVIDIA confirmed
to HP that it was able to reproduce the problem.’” Yet,
according to CW1, NVIDIA “‘never admitted to getting to
root cause, and they peddled the same story for as long as
possible, that they were still investigating.’” CW1 also
alleges that “NVIDIA would do the ‘PhD runaway’ to appear
cooperative while trying to slow HP down.” In essence, CW1

Case: 11-17708

30

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

asserts that NVIDIA’s Ph.D. employees would ask HP to
conduct additional tests merely to buy more time.
As we discussed above, the timing of CW1’s account is
implausible, especially in light of Hodge’s more-detailed
declaration. CW1 provides no specific examples of when
NVIDIA attempted to slow down HP by having it conduct
meaningless tests or experiments, nor does CW1 explain why
those tests and experiments were meaningless. From CW1’s
assertions alone, a reasonable factfinder could not infer that
NVIDIA was conducting needless testing or attempting to
delay disclosure of its product defects. As a matter of law,
CW1’s account to this end does not add to an inference of
scienter.
Plaintiffs also rely on a statement made by NVIDIA’s
Vice President of Investor Relations, Michael Hara. In
September 2008, after NVIDIA’s disclosure, Hara explained
to investors that the failing chips were past or near the end of
their useful life and that NVIDIA was “‘not even shipping
these parts anymore.’” From this statement, Plaintiffs infer
that “Defendants stalled the disclosure of the material,
adverse facts for several months, if not a full year, until the
failing products were at the end of their life-cycle and the
Company had new products to market.”
We agree that the coincidence of NVIDIA’s disclosure
with the apparent end-of-life of its failing chips could, under
some circumstances, arouse suspicion. But this statement
alone does not create an inference that NVIDIA strategically
delayed disclosure of its chip failures until those chips were
past their useful life.

Page: 30 of 41

Case: 11-17708

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Page: 31 of 41

31

Finally, Plaintiffs rely on the fact that NVIDIA
participated in preparation of BIOS updates and issued PCNs.
As we discussed above, the uncontested evidence shows that
in November 2007, NVIDIA helped HP (and later Dell) issue
BIOS software updates that altered a computer’s fan
algorithm. In May and June 2008, NVIDIA issued PCNs that
indicated it would be changing back to a eutectic solder in its
GPUs and MCPs.
Plaintiffs allege that significant
engineering is required to transition from one Material Set to
another. From these facts, Plaintiffs conclude that NVIDIA
must have known it was liable for the product defects long
before its disclosure. But these facts do not mandate such a
conclusion. The BIOS updates were intended to ease stress
caused by heat cycling, which, as NVIDIA first contended,
could have been caused by OEM design issues. Even if
NVIDIA decided to transition back to eutectic solder long
before it issued PCNs in May and June 2008, it does not
necessarily mean that NVIDIA knew it was responsible for
the failures. A more plausible inference is that NVIDIA
believed that high-lead solder was a contributing factor and
switched back to eutectic solder to eliminate it.
4.
After alleging that NVIDIA delayed disclosure until it had
prepared replacement chips, Plaintiffs attempt to buttress that
allegation by contending that NVIDIA has a history of
delaying disclosure of known problems. Nevertheless, they
rest their contention on the accounts of several confidential
witnesses whose experiences do not contribute to an inference
of scienter. Their accounts are unspecific and speculative.
More problematic, some witnesses never worked for
NVIDIA. And those who did either stopped working there
long before the Material Set Problem arose or worked in an

Case: 11-17708

32

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

area unrelated to it, such as environmental engineering or
“Staffing Systems and Compliance.”
For example, Confidential Witness No. 11 (“CW11”)
worked for NVIDIA as a senior engineering manager
between December 2002 and March 2006, months before the
product failures arose. CW11’s declaration maintains that
NVIDIA is “arrogant, [and] think[s it] never do[es] wrong,”
and “operated with a ‘failure to recognize real problems.’”
Confidential Witness No. 12 (“CW12”) was a sales director
for one of NVIDIA’s vendors. CW12 “stated that NVIDIA
was notorious for blaming other entities for product related
problems. . . . ‘Some of the time they were right in [blaming
others] and some of the time they were wrong, but their
default was always to say, “This is not our problem. This is
your problem.”’”
Confidential Witness No. 13 (“CW13”) worked at
NVIDIA as an environmental compliance engineer. CW13
“stated that NVIDIA has a history of failing to take
responsibility for Company problems.” CW13 believes that
NVIDIA “‘had the following mentality: “Don’t say anything
to muck the waters.”’” “[Confidential Witness No. 15
(“CW15”)], a former NVIDIA Staffing Systems and
Compliance Analyst, stated that s/he was told: ‘I don’t care
what you do, just make us look good.’” CW15 provides no
context, however, as to who made the statement or why the
statement was made. Confidential Witness No. 16 (“CW16”)
worked for a website that sold NVIDIA products. CW16
stated, “‘I’ve seen a lot of hostile actions from [NVIDIA] . . .
‘NVIDIA tends to be a little “scrooge-ish” when it comes to
scrapping their failure percentage rates. . . . They’ve had a lot
of fiascos in the past with like the GeForce 2 Ultra, a lot of
overheating issues.’”

Page: 32 of 41

Case: 11-17708

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Page: 33 of 41

33

These statements do not give rise to an inference of
scienter individually. The testimony of Confidential Witness
No. 14 (“CW14”) provides the most probative evidence
indicating that NVIDIA had a culture of failing to take
responsibility for known problems. CW14 was a “Director of
IC Quality and Reliability for NVIDIA.” According to
CW14, there was an instance where employees at NVIDIA
knew of a design flaw in one of its products, but did not
reveal it until after NVIDIA’s customer had discovered and
contained the problem. CW14’s testimony is less significant,
however, when considering that his/her period of employment
was at most one year, from 2000 to 2001, long before the
product defects giving rise to this lawsuit.
Accordingly, these allegations that NVIDIA has a history
of delaying known problems do not give rise to a strong
inference of scienter.
5.
As further evidence of scienter, Plaintiffs rely on the
existence of other lawsuits filed against NVIDIA by its
insurers.
Apparently, after it disclosed to investors
information regarding its defective products, NVIDIA
submitted claims to its insurance companies to cover the
losses sustained as a result. Foreseeably, the insurance
companies did not want to pay NVIDIA’s claims. Thus, they
alleged that NVIDIA knew of the defective products before
January 2008 and had failed to provide information that the
insurers had requested. These allegations do not significantly
add to an inference of scienter, as the insurers litigating
claims would attempt to avoid liability.

Case: 11-17708

34

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Evidently, one of NVIDIA’s insurers also alleged the
existence of a class action lawsuit filed against HP in
November 2007. The insurer argued that NVIDIA knew the
lawsuit related to problems with NVIDIA’s defective chips
and that HP would seek indemnification from NVIDIA.
Nevertheless, Plaintiffs never explain with any particularity
when or how NVIDIA became aware that HP would seek
indemnification from NVIDIA. Accordingly, the existence
of this additional lawsuit does not add to an inference of
scienter.
6.
Plaintiffs also contend that, when considered with all
other allegations, the departure of some of NVIDIA’s
executives adds to the inference of scienter. We do not agree.
Plaintiffs allege that three individuals left NVIDIA after
it disclosed the defective products. On May 27, 2008,
NVIDIA’s CFO, Marvin Burkett, announced his retirement.
Significantly, Burkett continued as interim CFO until a
replacement was hired, and in February 2009, he became a
senior advisor to NVIDIA. In June 2008, NVIDIA’s Senior
Director and Head of Internal Audit left the company.
Plaintiffs do not explain, however, how this executive played
any role in allegedly delaying the disclosure of the defective
products. In January 2009, NVIDIA replaced David Kirk, its
Chief Scientist. Plaintiffs fail to provide any detail as to why
Kirk was replaced; notably, Kirk continued to work with
NVIDIA as a research fellow.
In short, Plaintiffs fail to provide any facts to connect
these departures with the problems at issue in this lawsuit.
More detrimental to their allegations, however, is that two of

Page: 34 of 41

Case: 11-17708

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Page: 35 of 41

35

the three individuals remained at NVIDIA in some type of
advisory role. Therefore, the most reasonable inference is
that these departures were benign.
7.
Plaintiffs argue that their allegations give rise to an
inference of scienter under the corporate scienter doctrine.
“‘In most cases, the most straightforward way to raise [an
inference of scienter] for a corporate defendant will be to
plead it for an individual defendant.’” Glazer Capital Mgmt.,
LP v. Magistri, 549 F.3d 736, 743 (9th Cir. 2008) (quoting
Teamsters Local 445 Freight Div. Pension Fund v. Dynex
Capital Inc., 531 F.3d 190, 195 (2d Cir. 2008)). Yet, the
collective scienter (or corporate scienter) doctrine recognizes
that it is possible to raise the inference of scienter without
doing so for a specific individual. Id. In the Ninth Circuit,
we “ha[ve] not previously adopted a theory of collective
scienter.” Id. at 744. Nevertheless, in Glazer Capital, we
opined that the doctrine might be appropriate in some cases.
Id. “For instance, as outlined in the hypothetical posed [by
the Seventh Circuit], there could be circumstances in which
a company’s public statements were so important and so
dramatically false that they would create a strong inference
that at least some corporate officials knew of the falsity upon
publication.” Id. (citing Makor Issues & Rights, Ltd. v.
Tellabs Inc., 513 F.3d 702, 710 (7th Cir. 2008)).
We do not believe the facts alleged in this case would
give rise to an inference of scienter under the collective
scienter doctrine. Here, Plaintiffs contend that statements
made by NVIDIA in its SEC filings were misleading because
NVIDIA did not also disclose information regarding the
Material Set Problem. Those statements were not “so

Case: 11-17708

10/02/2014

36

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

dramatically false”—as in the example posed by the Seventh
Circuit in Makor13—to create an inference of scienter that at
least some corporate officials knew of the falsity upon
publication.
Plaintiffs also contend that their complaint raises a strong
inference of scienter under the core operations doctrine.
Under this doctrine, the PSLRA’s pleading requirements may
be satisfied, in certain circumstances, by “a scienter theory
that infers that facts critical to a business’s ‘core operations’
or an important transaction are known to a company’s key
officers.” S. Ferry LP, No. 2 v. Killinger, 542 F.3d 776,
783–84 (9th Cir. 2008). In South Ferry, we explained that, in
light of the Supreme Court’s decision in Tellabs, the core
operations inference may be considered when weighing a
complaint’s allegations of scienter holistically. Id. at 784–85.
Nevertheless, we cautioned that, “‘absent some additional
allegation of specific information conveyed to management
and related to the fraud’ or other allegations supporting
scienter,” the core operations inference will generally fall
short of a strong inference of scienter. Id. We did, however,
13

In Makor, the Seventh Circuit illustrated its point that
it is possible to draw a strong inference of corporate
scienter without being able to name the individuals who
concocted and disseminated the fraud. Suppose
General Motors announced that it had sold one million
SUVs in 2006, and the actual number was zero. There
would be a strong inference of corporate scienter, since
so dramatic an announcement would have been
approved by corporate officials sufficiently
knowledgeable about the company to know that the
announcement was false.

513 F.3d at 710.

Page: 36 of 41

Case: 11-17708

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Page: 37 of 41

37

leave open the possibility that “in some unusual
circumstances, the core operations inference, without more,
may raise the strong inference required by the PSLRA.” Id.
at 785. One example of such unusual circumstances is
“where the nature of the relevant fact is of such prominence
that it would be ‘absurd’ to suggest that management was
without knowledge of the matter.” Id. at 786.
Here, Plaintiffs never plausibly allege that specific
information was conveyed to Huang14 or others in NVIDIA’s
management team. They apparently attempt to do so by
relying on the account of CW1, but CW1 lacks personal
knowledge of these facts. “CW1 report[s] that the
communications regarding the [chip] problems were
conducted at a high-level [sic]: ‘These were executive level
guys.’” CW1’s “boss was a senior technical employee, a
Director of Quality.” “‘This guy is not going to be
communicating with guys that are at manager level or
engineering level.’” From this statement, it is evident that
CW1 does not actually know whom from NVIDIA his/her
boss communicated with regarding the chip failures. CW1
surmises that, based on his/her boss’s status in HP’s corporate
hierarchy, he was communicating with executive level
personnel from NVIDIA. Accordingly, CW1 does not appear
to have the requisite personal knowledge to assert that
NVIDIA’s management team received specific information
regarding the defective products.
14

Plaintiffs do allege that Huang was heavily involved in the design of
its chips. They support this allegation with the account of Confidential
Witness No. 17 (“CW17”). Nevertheless, CW17 merely indicates that
Huang “did not want an undue number of [solder] bumps” on the chips.
This description of Huang’s involvement does not indicate that he was
heavily involved in the design of the flawed GPUs and MCPs or in the
decision to use high-lead solder.

Case: 11-17708

38

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Even assuming CW1 did have personal knowledge that
his/her boss communicated with NVIDIA’s executive-level
personnel, the specific information allegedly conveyed
related to chip failures, not the root cause of them or
NVIDIA’s liability. As we discussed above, CW1’s assertion
that HP had determined the root cause in early 2007 is not
plausible.
Again, even assuming HP had determined the root cause
in early 2007, Plaintiffs do not allege any facts to support the
notion that anyone at NVIDIA arrived at the same conclusion
as HP regarding the root cause, or that NVIDIA would be
liable. Nor do Plaintiffs provide a basis to infer that anyone
at NVIDIA was aware that its financial liability would exceed
its normal reserves. Plaintiffs argue that “[k]nowledge of the
financial impact of the chip defect should be presumed when
the nature of the problem concerned [NVIDIA’s] flagship
product and was cause for concern to [NVIDIA’s] two largest
customers.” We do not agree. Without factual allegations
showing that at least someone at NVIDIA had knowledge of
the extent of NVIDIA’s liability, there is no basis to presume
that NVIDIA’s management would have had such
knowledge.
Accordingly, neither the collective scienter doctrine nor
the core operations doctrine alone gives rise to a strong
inference of scienter.
****
Having evaluated Plaintiffs’ allegations individually, we
find that none creates a strong inference of scienter alone.
Evaluating the allegations together, we find that they do not
create a strong inference of scienter collectively. The most

Page: 38 of 41

Case: 11-17708

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Page: 39 of 41

39

that a reasonable factfinder could infer from Plaintiffs’
allegations is that NVIDIA was aware that some versions of
its GPUs and MCPs were experiencing problems sometime
in late 2006 or early 2007. At some point, HP determined the
thermal profile that increased the probability that NVIDIA’s
chips would fail due to cracking at the solder bumps. HP
shared with NVIDIA the thermal profile and other data that
it believed demonstrated NVIDIA’s liability. NVIDIA
evidently reproduced these data and thermal profile yet
contested that it was at fault for the chip failures. While
Plaintiffs infer that NVIDIA was merely delaying disclosure
until it had prepared replacement chips, this is not a cogent
and compelling inference. NVIDIA indicated in May 2008
(and even in July 2008) that it had not yet determined the root
cause of the product failures, although it evidently was
working on a solution. Plaintiffs provide no factual basis to
discount those statements. Moreover, product flaws are very
common in the semiconductor industry, and NVIDIA
regularly takes measures to account for this, as reflected in its
disclosures. It warns investors of this possibility and sets
aside a reserve to account for costs related to those flaws.
Although there is some slight support for an inference that
NVIDIA knew it was responsible for the problem before its
disclosure, and thus acted with intent to deceive at least
customers if not investors, a more compelling inference is
that NVIDIA did not disclose because it was investigating the
extent of the problem, whether it was responsible for it, and
if so, whether it would exhaust the reserve. Accordingly, we
hold that the complaint’s allegations do not give rise to a
strong inference of scienter when considered holistically. See
Tellabs, 551 U.S. at 324.
On appeal, Plaintiffs argue that the Supreme Court has
rejected the “inference that defendants were delaying

Case: 11-17708

40

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

disclosure while ‘investigating the scope of the issue.’” It is
true that, in Matrixx Initiatives, the Supreme Court rejected
the assertion that “‘the most cogent inference regarding [the
defendant’s] state of mind is that it delayed releasing
information regarding [a product defect] in order to provide
itself an opportunity to carefully review all evidence.’”
Matrixx Initiatives, 131 S. Ct. at 1324 n.15. Yet the Court
also “d[id] not doubt that this may be the most cogent
inference in some cases.” Id. In Matrixx Initiatives, the
defendant was a pharmaceutical company that manufactured
and sold Zicam, an over-the-counter remedy for the common
cold. Id. at 1313–14. At some point, a doctor began giving
public presentations indicating that Zicam was a flawed
product and posed dangerous health risks to its users. Id. at
1316. In response to these presentations, the defendant
“issued a press release that suggested that studies had
confirmed” that Zicam was not flawed. Id. at 1316, 1324.
Nevertheless, the defendant’s press release was false, as no
such studies existed. Id. at 1324. Accordingly, the Court
held that “the misleading nature of [the defendant’s] press
release is sufficient to render the inference of scienter at least
as compelling as the inference [that the defendant was
investigating the evidence].” Id. at 1324 n.15.
Here, there are no similar facts. There is no allegation
that the issue of an inherent defect in NVIDIA’s Material Set
was ever publicly raised prior to NVIDIA’s disclosure, nor is
there any allegation that NVIDIA knowingly issued a false
press release, attempting to discount any public discussion
regarding its chips’ defects. As such, we reject Plaintiffs’
assertion that Matrixx Initiatives forecloses the inference that
NVIDIA delayed disclosure while it investigated the cause of
the chip defects and the extent of its liability.

Page: 40 of 41

Case: 11-17708

10/02/2014

ID: 9262516

DktEntry: 44-1

IN RE: NVIDIA CORP. SEC. LITIG.

Page: 41 of 41

41

IV
For the reasons discussed above, we affirm the district
court’s judgment in its entirety.
AFFIRMED.