Akamai Technologies, Inc. v. Limelight Networks, Inc.

United States Court of Appeals
for the Federal Circuit
______________________

AKAMAI TECHNOLOGIES, INC.,
THE MASSACHUSETTS INSTITUTE OF
TECHNOLOGY,
Plaintiffs-Appellants

v.

LIMELIGHT NETWORKS, INC.,
Defendant-Cross-Appellant
______________________

2009-1372, 2009-1380, 2009-1416, 2009-1417
______________________

Appeals from the United States District Court for the
District of Massachusetts in Nos. 06-CV-11585, 06-CV-
11109, Judge Rya W. Zobel.
______________________

Decided: November 16, 2015
______________________

SETH P. WAXMAN, Wilmer Cutler Pickering Hale and
Dorr LLP, Washington, DC, argued for plaintiffs-
appellants. Also represented by THOMAS G. SAUNDERS,
THOMAS G. SPRANKLING; MARK C. FLEMING, ERIC F.
FLETCHER, LAUREN B. FLETCHER, BROOK HOPKINS, Boston,
MA; DAVID H. JUDSON, Law Offices of David H. Judson,
Dallas, TX; DONALD R. DUNNER, ELIZABETH D. FERRILL,
Finnegan, Henderson, Farabow, Garrett & Dunner, LLP,
Washington, DC; JENNIFER S. SWAN, Palo Alto, CA;
ROBERT S. FRANK, JR., G. MARK EDGARTON, CARLOS
2 AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC.

PEREZ-ALBUERNE, Choate, Hall & Stewart, LLP, Boston,
MA.

AARON M. PANNER, Law Office of Aaron M. Panner,
P.L.L.C., Washington, DC, argued for defendant-cross-
appellant. Also represented by JOHN CHRISTOPHER
ROZENDAAL, MICHAEL E. JOFFRE, Kellogg, Huber, Hansen,
Todd, Evans & Figel, P.L.L.C., Washington, DC; MICHAEL
W. DE VRIES, ALLISON W. BUCHNER, Kirkland & Ellis LLP,
Los Angeles, CA; YOUNG JIN PARK, New York, NY; DION
D. MESSER, Limelight Networks, Inc., Tempe, AZ.

JEFFREY I.D. LEWIS, Fried, Frank, Harris, Shriver &
Jacobson LLP, New York, NY, for amicus curiae Ameri-
can Intellectual Property Law Association. Also repre-
sented by KRISTIN M. WHIDBY, Washington, DC; LISA K.
JORGENSON, American Intellectual Property Law Associa-
tion, Arlington, VA.

SCOTT A.M. CHAMBERS, Porzio, Bromberg & Newman,
P.C., Washington, DC, for amicus curiae Biotechnology
Industry Organization. Also represented by CAROLINE
COOK MAXWELL; HANSJORG SAUER, Biotechnology Indus-
try Organization, Washington, DC.

CHARLES R. MACEDO, Amster Rothstein & Ebenstein
LLP, New York, NY, for amicus curiae Broadband iTV,
Inc. Also represented by JESSICA CAPASSO.

PAUL H. BERGHOFF, McDonnell, Boehnen, Hulbert &
Berghoff, LLP, Chicago, IL, for amicus curiae Intellectual
Property Owners Association. Also represented by PHILIP
S. JOHNSON, Johnson & Johnson, New Brunswick, NJ;
KEVIN H. RHODES, 3M Innovative Properties Co., St. Paul,
MN; HERBERT C. WAMSLEY, Intellectual Property Owners
Association, Washington, DC.
AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC. 3

CARTER G. PHILLIPS, Sidley Austin LLP, Washington,
DC, for amicus curiae Pharmaceutical Research and
Manufacturers of America. Also represented by JEFFREY
P. KUSHAN, RYAN C. MORRIS; DAVID E. KORN, Pharmaceu-
tical Research and Manufacturers of America, Washing-
ton, DC; DAVID R. MARSH, LISA A. ADELSON, Arnold &
Porter, LLP, Washington, DC; ROBERT P. TAYLOR, MONTY
AGARWAL, San Francisco, CA.

DEMETRIUS TENNELL LOCKETT, Townsend & Lockett,
LLC, Atlanta, GA, for amici curiae Nokia Technologies
Oy, Nokia USA Inc.

DONALD R. WARE, Foley Hoag LLP, Boston, MA, for
amicus curiae The Coalition for 21st Century Medicine.
Also represented by MARCO J. QUINA, SARAH S. BURG.

______________________

Before PROST, Chief Judge, LINN and MOORE, Circuit
Judges.
LINN, Circuit Judge.
This case first came to this court after, inter alia, a ju-
ry verdict finding Akamai’s U.S. Pat. No. 6,108,703 (“’703
patent”) not invalid and directly infringed by Limelight,
followed by the entry of judgment as a matter of law
(“JMOL”) overturning the jury’s infringement verdict on
the basis of divided infringement. Akamai Techs., Inc. v.
Limelight Networks, Inc. (Akamai II), 614 F. Supp. 2d 90
(D. Mass. 2009). After several rounds of appeals and
remands, culminating with the en banc court’s reversal of
the district court’s JMOL determination on the divided
infringement issue, the case returns to this panel, which
is tasked with resolving “all residual issues” in the appeal
and cross-appeal. Akamai Techs., Inc. v. Limelight Net-
works, Inc. (Akamai IV), 797 F.3d 1020, 1025 (Fed. Cir.
2015) (en banc).
4 AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC.

On this record, the only issues remaining stem from
Limelight’s cross-appeal, which argued alternative
grounds for overturning the jury’s verdict of infringement
and challenged the damages award. Specifically, three
issues remain to be adjudicated. First, whether the
district court erred in construing the claim term “tag-
ging.” 1 Second, whether the district court properly con-
structed the term “optimal,” and properly instructed the
jury on the construction. 2 Third, whether the district
court erred in allowing Akamai to present a lost profits
theory based on the testimony of its expert.
Because the district court did not err in its claim con-
structions and appropriately instructed the jury, and
because we find no error in the district court’s allowance
of Akamai’s lost profits expert, we decline Limelight’s
invitation to find an alternate basis to overturn the jury
verdict on infringement and its damages award. Accord-
ingly, we reiterate the en banc court’s reversal of the
district court’s grant of JMOL of non-infringement and
remand with instructions to reinstitute the jury’s original
verdict and damages award. We also confirm our previ-
ously reinstated affirmance of the district court’s judg-
ment of non-infringement of U.S. Patent Nos. 6,553,413
(the “’413 patent”) and 7,103,645 (the “’645 patent”).

1 Limelight argues that the district court erred in
its construction, and that the jury lacked sufficient evi-
dence to find infringement in light of the correct construc-
tion.
2 Limelight argues both that the claim construction
was erroneous, and that the subsequent jury instruction
improperly left claim construction to the jury.
AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC. 5

I. BACKGROUND
A. The Technology and the Nature of the Dispute
A detailed description of the technology and the
claims at issue in this case is set forth in the prior report-
ed opinions of this court and the Supreme Court and will
not be repeated except to the extent germane hereto. See
Akamai IV, 797 F.3d 1020; Limelight Networks, Inc. v.
Akamai Techs., Inc., 134 S. Ct. 2111 (2014); Akamai
Techs., Inc. v. Limelight Networks, Inc. (Akamai III), 629
F.3d 1311 (Fed. Cir. 2010).
B. Prior Proceedings
In 2006, Akamai sued Limelight in the United States
District Court for the District of Massachusetts asserting
infringement of claims 19–21 and 34 of the ’703 patent,
along with certain claims of the ’413 and ’645 patents.
After the district court’s first claim construction order,
Akamai Techs., Inc. v. Limelight Networks, Inc., 494 F.
Supp. 2d 34 (D. Mass. 2007), Akamai stipulated that it
could not prove infringement of the ’645 patent under the
district court’s construction. The district court thus
entered judgment of non-infringement. The district court
subsequently entered summary judgment of non-
infringement of the asserted claims of the ’413 patent.
As relates to the ’703 patent, the parties stipulated to
a construction of “tagging” in claims 17, 19, and 34 of
the ’703 patent as “providing a ‘pointer’ or ‘hook’ so that
the object resolves to a domain other than the content
provider domain.” Akamai Techs., Inc. v. Limelight
Networks, Inc. (Akamai I), No. 06-11109, 2008 WL
697707, at *1 (D. Mass. Feb. 8, 2008). The meaning of
this term was not disputed until Limelight requested a
jury instruction explaining that tagging could only be
accomplished by “either prepending or inserting a virtual
server hostname into the URL,” and filed Rule 50 motions
for judgment of non-infringement because the accused
6 AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC.

products did not tag in this way. The district court denied
the requested jury instruction and the Rule 50 motions.
The parties also stipulated that “to resolve to a do-
main other than the content provider domain” in claims
17, 19, and 34 of the ’703 patent should be construed as
“to specify a particular group of computers that does not
include the content provider from which an optimal server
is to be selected.” Akamai I, 2008 WL 697707 at *1 (em-
phasis added). However, the parties disagreed on the
meaning of the word “optimal” in the construction, with
Limelight arguing that a single optimal server must be
selected, and Akamai arguing that several servers could
be “optimal” if they each met some criteria. Id. The
district court construed “optimal server” as “requir[ing]
the selection of a content server that is better than other
possible choices in terms of the criteria established by the
specification.” Id. at *3.
Akamai’s claim that Limelight infringed the ’703 pa-
tent proceeded to a jury trial. The district court instruct-
ed the jury on “tagging” per the stipulation discussed
above, and added the following gloss for “an optimal
server”:
one or more content servers that are better than
other possible choices considering some or all of
the following criteria: (1) being close to end users;
(2) not overloaded; (3) tailored to viewers in a par-
ticular location; (4) most likely to already have a
current version of the required file; and (5) de-
pendent on network conditions.
To prove damages, Akamai relied heavily on the tes-
timony of its expert, Dr. Keith Ugone’s calculation of
Akamai’s lost-profits. Dr. Ugone considered the elasticity
of the market for content delivery network services, the
competition between Akamai and Limelight, and the price
disparity between Akamai’s and Limelight’s products.
Ultimately, Dr. Ugone concluded that but-for Limelight’s
AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC. 7

infringement, Akamai would have collected about $74
million.
The jury returned a verdict of infringement and
awarded Akamai approximately $40 million in lost prof-
its, $1.4 million in reasonable royalty damages, and $4
million in price erosion damages. As noted, supra, the
district court did not let the verdict stand and, instead,
granted JMOL of no infringement. Akamai II, 614 F.
Supp. 2d at 96.
Akamai appealed the district court’s rulings regarding
all three patents-in-suit and Limelight cross-appealed.
This court rejected Akamai’s argument that Limelight’s
cross-appeal was improper, Akamai Techs., Inc. v. Lime-
light Networks, Inc., No. 2009-1372, 2010 WL 331770
(Fed. Cir. Jan. 27, 2010) (Order), and subsequently af-
firmed the district court’s rulings regarding the ’413
and ’645 patents. Akamai III, 629 F.3d at 1322–31. The
portion of this court’s Akamai III opinion dealing with
the ’645 and ’413 patents, though initially vacated upon
grant of en banc rehearing, Akamai Techs., Inc. v. Lime-
light Networks, Inc., 419 F. Appx 989 (Fed. Cir. 2011),
was later reinstated, see Akamai Techs., Inc. v. Limelight
Networks, Inc., 786 F.3d 899, 903-904 (Fed. Cir. May 13,
2015) (explaining procedural history), overruled en banc
on other grounds by Akamai IV, 797 F.3d 1020. As noted,
supra, this court reversed the non-infringement judgment
and returned the case to this panel for resolution of all
residual issues. Akamai IV, 797 F.3d at 1025.
This court has jurisdiction pursuant to 28 U.S.C. §
1295(a)(1).
II. DISCUSSION
A. Standard of Review
The “ultimate interpretation” of a claim term, as well
as interpretations of “evidence intrinsic to the patent (the
patent claims and specifications, along with the patent’s
8 AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC.

prosecution history),” are legal conclusions, which this
court reviews de novo. Teva Pharms. USA, Inc. v. Sandoz,
Inc., 135 S. Ct. 831, 841 (2015). Review of the district
court’s interpretation of the parties’ pre-trial stipulations
is “much like” review of any contract interpretation, see
Scanner Techs. Corp. v. ICOS Vision Sys. Corp. N.V., 528
F.3d 1365, 1383 (Fed. Cir. 2008): this court reviews un-
derlying factual findings for clear error and reviews the
ultimate interpretation of the stipulation de novo, see
Teva, 135 S. Ct. at 837–38.
This court reviews challenges to jury instructions,
grants or denials of motions for JMOL, and questions of
judicial estoppel under the law of the regional circuit
where the district court sits. See AbbVie Deutschland
GmbH & Co., KG v. Janssen Biotech, Inc., 759 F.3d 1285,
1295 (Fed. Cir. 2014); Source Search Techs., LLC v.
LendingTree, LLC, 588 F.3d 1063, 1071 (Fed. Cir. 2009).
“When examining preserved claims of instructional error,
[the First Circuit] afford[s] de novo review to questions as
to whether jury instructions capture the essence of the
applicable law, while reviewing for abuse of discretion the
court’s choice of phraseology.” Ira Green, Inc. v. Military
Sales & Servs. Co., 775 F.3d 12, 18 (1st Cir. 2015) (cita-
tions omitted). The First Circuit “review[s] the district
court’s grant or denial of judgment as a matter of law de
novo . . . viewing the evidence in the light most favorable
to the verdict-winner, and vacating the jury verdict only if
it lacks a sufficient evidentiary basis.” Kennedy v. Town
of Billerica, 617 F.3d 520, 537 (1st Cir. 2010). The First
Circuit “review[s] the district court’s decision not to
invoke judicial estoppel for abuse of discretion . . . ac-
cept[ing] the trial court’s findings of fact unless they are
clearly erroneous, and evaluat[ing] its answers to abstract
questions of law de novo.” Knowlton v. Shaw, 704 F.3d 1,
9–10 (1st Cir. 2013) (citations omitted).
“Whether lost profits are legally compensable in a
particular situation is a question of law that we review de
AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC. 9

novo.” Siemens Med. Solutions USA, Inc. v. Saint-Gobain
Ceramics & Plastics, Inc., 637 F.3d 1269, 1287 (Fed. Cir.
2011).
B. Claim Construction
1. “[T]agging”
In a prior litigation relating to the ’703 patent, the
district court construed “tagging” as “providing a ‘pointer’
or ‘hook’ so that the object resolves to a domain other than
the content provider domain.” Akamai Techs., Inc. v.
Digital Island, No. 00-11851-RWZ, 2001 WL 36172136, at
*1 (D. Mass. Nov. 8, 2001). The district court defined “to
resolve to a domain other than the content provider
domain” as “to specify a particular group of computers
that does not include the content provider from which an
optimal server is to be selected.” Id. (emphasis added).
The parties accepted these constructions by stipulation in
the instant case. Akamai I, 2008 WL 697707 at *1. This
construction was not disputed during the Markman
hearing, and was first challenged by Limelight in at-
tempting to re-craft the construction for the jury instruc-
tions.
Limelight argues that: 1) in the context of the ’703 pa-
tent, “tagging” is necessarily limited to using a “pointer”
or “hook” that either prepends or inserts a virtual server
hostname into the URL because the ’703 patent discloses
no other way to “tag” to achieve the goals of the invention;
and 2) that “alphanumeric string” as used in the ’645
patent (and which this court has construed to necessarily
include prepending or inserting a virtual server hostname
in the URL) is the product of tagging in the ’703 patent,
which necessarily means that the ’703 patent incorporates
the same limitations as the ’645 patent. Akamai counters
that: 1) Limelight waived the argument by failing to
assert it during Markman and again failing to assert it
after the jury instructions were read; 2) the stipulation to
which Limelight agreed was made without further limita-
10 AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC.

tion of the types of “hook” or “pointer” to use, and is thus
binding on Limelight; 3) “tagging” in the ’703 patent is not
equivalent to “alphanumeric string” in the ’645 patent;
4) certain claims in the ’703 patent specifically require
prepending while others don’t, and claim differentiation
requires that the broader term “tagging” thus not be
limited to prepending; 5) prepending is merely a preferred
embodiment and Limelight is improperly attempting to
limit the claim scope to a preferred embodiment; and
6) Limelight argued that the asserted claims lacked
written description because the specification taught that
the only way to tag was to prepend a virtual hostname
into an existing URL – but the jury rejected this argu-
ment.
Limelight’s attempt to import a “prepending” limita-
tion into the claims fails. “[O]ur cases recognize that the
specification may reveal a special definition given to a
claim term by the patentee that differs from the meaning
it would otherwise possess. In such cases, the inventor’s
lexicography governs.” Phillips v. AWH Corp., 415 F.3d
1303, 1316 (Fed. Cir. 2005) (en banc). However, a claim
term is only given a special definition different from the
term’s plain and ordinary meaning if the “patentee . . .
clearly set[s] forth a definition of the disputed claim term
other than its plain and ordinary meaning.” Thorner v.
Sony Comput. Entm’t Am. LLC, 669 F.3d 1362, 1365 (Fed.
Cir. 2012) (citations omitted). A patentee can also disa-
vow claim scope, but the standard “is similarly exacting.”
Id. at 1366. “[C]laims are not necessarily and not usually
limited in scope simply to the preferred embodiment.” RF
Del. v. Pac. Keystone Techs., Inc., 326 F.3d 1255, 1263
(Fed. Cir. 2003).
The ’703 patent describes prepending as a “pre-
fer[ence].” ’703 patent, col.4 ll.2–3. Figure 4 describes
“prepend[ing a] virtual server host name,” but the patent
likewise describes Figure 4 as showing the “preferred”
method. Id. at col.6 ll.44–45. The patent’s reference to
AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC. 11

preferred embodiments where the virtual server host-
name is prepended does not provide the clarity necessary
to find that the patentees intended to limit the term
tagging to the preferred embodiment. Moreover, claim 17
of the ’703 patent expressly recites “tagging . . . by pre-
pending,” suggesting that the term “tagging”—without
modification and as recited in the asserted claims—is not
so limited. See Ancora Techs., Inc. v. Apple, Inc., 744 F.3d
732, 735 (Fed. Cir. 2014) (explaining that using the
phrase “application software program” in one claim, and
“program” alone in another “tends to reinforce . . . adop-
tion of the broad ordinary meaning of ‘program’ by itself”).
The prosecution history cited by Limelight also fails to
provide the necessary clarity to limit “tagging” to the
preferred embodiment. During prosecution, Akamai
amended what is now claim 17 to require tagging “by
prepending” and amended claim 19 to require that the
content provider “serv[e] the given page” and that the
Content Delivery Network serve the embedded image. In
their remarks, the applicants stated that “the embedded
object URL is modified . . . to prepend given data to the
domain name and path normally used to retrieve the
embedded object.” In view of the amendment now requir-
ing claim 17 to tag “by prepending,” a person of skill in
the art could reasonably understand the applicants’
description of prepending the data as referring only to
claim 17. This statement therefore does not provide the
necessary clarity required for disavowal in claim 19.
Limelight claims that the only method of tagging de-
scribed in the ’703 patent involves prepending a virtual
server hostname. However, as this court has held, “even
where a patent describes only a single embodiment,
claims will not be read restrictively unless the patentee
has demonstrated a clear intention to limit the claim
scope using words or expressions of manifest exclusion or
restriction.” Innova/Pure Water, Inc. v. Safari Water
Filtration Sys., 381 F.3d 1111, 1117 (Fed. Cir. 2004)
12 AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC.

(internal citations omitted). As explained above, no such
indication of exclusion appears in the patent specification
or prosecution history.
We note that the district court read to the jury the
construction of “tagging” to which Limelight stipulated.
Though Limelight points to O2 Micro International Ltd. v.
Beyond Innovation Technology Co., 521 F.3d 1351, 1361
(Fed. Cir. 2008), for the proposition that its stipulation
did not “give up any right to argue that further construc-
tion or interpretation of tagging would be needed,” that
case is inapposite. In O2 Micro, the Court was clearly
aware of the parties’ disagreement about the claim term
“only if,” and the Court refused to construe it beyond its
ordinary meaning. Id. at 1357 (“The parties agreed, for
the most part, that a previously issued claim construction
order . . . controlled in this case. . . . However, the parties
presented a handful of additional terms for the court to
construe [of which “only if” was one].”); id. at 1361 (“The
parties presented a dispute to the district court regarding
the scope of the asserted claims.”). See also id. at 1361
(“[T]he parties disputed not the meaning of the words
themselves, but the scope that should be encompassed by
this claim language.” (emphasis in original)). Here, the
parties agreed in the stipulation as to both the meaning
and the scope of the term during claim construction:
“tagging” means “providing a ‘pointer’ or ‘hook’ so that the
object resolves to a domain other than the content provid-
er domain.” This meaning was agreed-upon with no
further limitations. The lack of further limitations was
itself a characteristic of the construction to which both
parties agreed. Limelight cannot argue at the jury in-
struction stage – after the bulk of the trial was framed
and directed by the Markman construction to which it
agreed – that the construction was somehow too broad.
Limelight stipulated to a construction of “tagging,” and it
is bound by that stipulation.
AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC. 13

We find no error in the district court’s claim construc-
tion of “tagging” or the jury instruction pursuant thereto.
The parties do not assert that there is any remaining
issue of fact as to whether Limelight performs “tagging”
(apart from the “optimal server” issue addressed below).
2. “[A]n optimal server”
The second remaining dispute is whether “an optimal
server” is necessarily limited to a single “best” server, or
can refer to several potentially optimal servers from
which content is retrieved.
The phrase “optimal server” does not appear in the
patent. Instead, it is nested within the parties’ stipulated
claim constructions as follows. “Tagging” was stipulated
to mean “providing a ‘pointer’ or ‘hook’ so that the object
resolves to a domain other than the content provider
domain.” The phrase “to resolve to a domain other than
the content provider domain” was stipulated to mean “to
specify a particular group of computers that does not
include the content provider from which an optimal server
is to be selected.” Substituting the stipulated construc-
tions into claim 19 results in the following, with emphasis
added:
19 [substituted]. A content delivery service,
comprising . . . .
for a given page normally served from the
content provider domain, providing a
‘pointer’ or ‘hook’ to embedded objects
on the page so that requests for those
objects specify a particular group of
computers that does not include the
content provider from which an opti-
mal server is to be selected
....
14 AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC.

serving at least one embedded object of
the given page from a given content
server in the domain instead of from
the content provider domain.
During the Markman hearing, the parties disputed
the meaning of “an optimal server.” The district court
construed claim 19 to require “the content delivery system
to serve an embedded object from one or more content
servers which are ‘[m]ost favorable or desirable,’ that is,
servers which meet some or all of the criteria described in
the specification.” In the jury instruction, the district
court elaborated on the criteria, explaining that “an
optimal server” was: “one or more content servers that are
better than other possible choices considering some or all
of the following criteria: (1) being close to end users; (2)
not overloaded; (3) tailored to viewers in a particular
location; (4) most likely to already have a current version
of the required file; and (5) depend[e]nt on network condi-
tions.”
Limelight argues that: 1) the unambiguous meaning
of “optimal” is necessarily restricted to a single aggregate
“best” server; 2) the court’s ambiguous construction im-
properly left a claim construction issue for the jury; and
3) Akamai is judicially estopped from arguing that “opti-
mal” does not require a single “best” server by its state-
ments equating “optimal” to “best.”
Limelight’s arguments are unconvincing. First, Lime-
light fails to appreciate the context of the selection of “an
optimal server” in the claim. The selection of “an optimal
server” describes the functionality enabled by the neces-
sary “tagging.” In other words, the embedded objects are
tagged such that a group of computers is identified, and
from which an optimal server is chosen. The ’703 patent
is replete with examples in which conditions or circum-
stances independent of the tag influence which server
ultimately serves the embedded object. The tagging
AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC. 15

described in the ’703 patent thus allows for the tag to
ultimately lead to service from more than a single possi-
ble server. When the browser makes a request for an
object, then the software on the ghost does the following:
“If a copy of the file is already stored on the ghost, then
the data is returned immediately. If, however, no copy of
the data on the ghost exists, a copy is retrieved from the
original server or another ghost server.” ’703 patent,
col.12, ll.31-35. Similarly, the specification explains that
the tagging allows “a ghost server [to] redirect the user to
a closer server (or to another virtual address that is likely
to be resolved to a server that is closer to the client).” Id.
at col.12, ll.44-47. And again, “[p]erformance for long
downloads can also be improved by dynamically changing
the server to which a client is connected based on chang-
ing network conditions.” Id. at col.12, ll.53-55.
These examples undermine Limelight’s position in
two ways. First, the tagging of the embedded objects
provides the capability to select a server, and then select a
different server – in other words, tagging enables the
selection of one of several servers. Second, the criteria for
server selection are not aggregated during tagging,
wherein the system only allows serving from the single
server that is the “winner” of the aggregated criteria.
Instead, in one instance, a server may be chosen because
it is closest to the user; in another instance, because
another server does not have the file; and in yet another
instance, because of overload of the server or network
conditions. Choosing based on any of these criteria is
indicated as a capability of the claimed tagging system –
not merely choosing a single “aggregate best” server.
Nothing in the patent limits the functionality of the tag to
selecting an “aggregate best” – indeed, which criteria is
ultimately decisive is not a function of the tag, but occurs
while the objects are served.
This reading is confirmed by dependent claims 21 and
22, which further limit the serving step in claim 19 to
16 AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC.

“resolving a request to the domain as a function of a
requesting user’s location,” ’703 patent, cl. 21, or “resolv-
ing a request to the domain as a function of a requesting
user’s location and then-current Internet traffic condi-
tions,” id. at cl. 22. In other words, determining the
ultimate server from which the embedded object will be
served using one particular criteria, or two criteria.
Nothing in the specification or the claims implies that
these two functionalities would necessarily return the
aggregate best server, or that the two rules would return
the same server. Limelight argues that this identification
of “one or more content servers” is an additional step
identifying the list of all possible content servers from
which the optimum server is selected. Limelight Supp.
Opening Br. at 4. Limelight ignores that claim 20 is
limiting “the serving step,” which occurs after the tagging
step.
Limelight’s argument that the “unambiguous” mean-
ing of “optimal” is a single “best” is also unconvincing. As
discussed above, the intrinsic evidence supports the
district court’s construction. Moreover, neither the plain
meaning of “optimal” nor the plain meaning of “best” is as
limited as Limelight suggests to an “aggregate best” or
“aggregate optimal.”
The district court’s construction did not improperly
leave a claim construction issue for the jury by not con-
struing a disputed term. The district court construed
optimal server during the Markman hearing as described
above, and elaborated during jury instructions that “an
optimal server” was: “one or more content servers that are
better than other possible choices considering some or all
of the following criteria: (1) being close to end users; (2)
not overloaded; (3) tailored to viewers in a particular
location; (4) most likely to already have a current version
of the required file; and (5) depend[e]nt on network condi-
tions.” Nothing in the construction or the jury instruc-
tions requires the jury to construe the term. Limelight
AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC. 17

merely disagrees with the construction the district court
adopted.
Finally, Akamai’s use of “optimal” and “best” in the
prior litigation does not estop Akamai from arguing that
“optimal” allows for serving from other than a single
composite best server because the point at issue in the
discussions cited was distinct from the issue here. Lime-
light points to a colloquy wherein the district court ques-
tioned Akamai’s counsel about the functionality and
sequencing of the tagging step, and Akamai’s counsel
stated: “at some time during the serving of that object,
picking the best computer to serve that object, that’s
during the serving step, identifying the best computer,”
and also agreed with the district court’s categorization
that the process “is two steps. It tags to find the best
domain and then also identifies the best computer or
server within that domain.” Limelight’s reliance on this
colloquy is misplaced. The discussion in that case was
about the role of tagging, and Akamai’s attorney ex-
plained that the timing of the tagging step occurs with the
selection of a domain, but that the selection of the “best
computer” occurs during the object serving step. The
issue of whether tagging enables serving from only a
single “optimal server” or from a server which performs
better than others within a particular criteria was never
addressed.
For these reasons, there is no error in the district
court’s construction of “an optimal server,” nor in the jury
instruction.
C. Damages
To collect lost profits, a “patentee must show ‘a rea-
sonable probability that ‘but for’ the infringing activity,
the patentee would have made the infringer’s sales.”
Ericsson, Inc. v. Harris Corp., 352 F.3d 1369, 1377 (Fed.
Cir. 2004) (citations omitted). This is done by determin-
ing what profits the patentee would have made absent the
18 AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC.

infringing product. Id. This analysis must be supported
by “sound economic proof of the nature of the market and
likely outcomes with infringement factored out of the
economic picture.” Id. (citing Grain Processing Corp. v.
Am. Maize-Prods. Co., 185 F.3d 1341, 1350 (Fed. Cir.
1999)).
Limelight argues the district court committed legal
error in allowing lost profits as a measure of damages
because Akamai failed to show a causal connection be-
tween Limelight’s infringement and Akamai’s lost profits.
Limelight argues that Dr. Ugone’s calculation of the share
of Limelight’s customers that would have gone to Akamai
absent Limelight’s infringement was arbitrary and not
based in sound economic theory. The underlying basis for
this argument is the price disparity between Limelight’s
and Akamai’s products, which Limelight says Dr. Ugone
either failed to incorporate into his analysis, or incorpo-
rated arbitrarily. Limelight’s arguments are inapposite.
Limelight originally sold a different, non-infringing
service than the one at issue in this case. Limelight’s
infringing service was released in April of 2005. Dr.
Ugone testified that in 2005 Akamai had a market share
of 79.8% and Limelight had a market share of 5% and in
2006 Akamai had a market share of 74.7% and Limelight
had a market share of 10.7%. Dr. Ugone then calculated
an adjusted market share 3 for the years when Limelight’s
infringing service was on the market and concluded that,
assuming Limelight only sold its earlier software, Aka-
mai’s market share would have been 81% in 2005 and
79.9% in 2006. Because he did not have sufficient data to
determine the market share for 2007, he assumed it
would be the same as the market share for 2006. For the

3 An adjusted market share is the calculated mar-
ket share Akamai would have had absent infringement.
AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC. 19

sake of “conservatism,” Dr. Ugone reduced Akamai’s
share by 3% and excluded the lowest earning 25% of
Limelight’s customers who he categorized as particularly
price sensitive consumers, who may be unlikely to pur-
chase a higher-priced alternative without Limelight’s
infringing products in the market. Subject to these as-
sumptions and modifications, Dr. Ugone opined that
Limelight’s infringing sales totaled approximately $87.5
million.
The lost profit analysis was complicated by the fact
that Limelight sold its product for half the price of Aka-
mai’s. This affected Dr. Ugone’s calculations in two ways.
First, he assumed that in the but-for world where Lime-
light did not sell an infringing product, Akamai would sell
its product to some of those customers for twice as much
as Limelight had. Second, because of the difference in
price between Akamai’s product and Limelight’s product,
Dr. Ugone assumed that the demand for Akamai’s prod-
uct would be 25% less than the demand for Limelight’s
infringing products. Dr. Ugone explained that, in econom-
ics, how a change in price affects a change in demand is
described as “elasticity.” The more elastic the demand,
the more sensitive it is to change. A demand is described
as “inelastic” if, when the price changes by a certain
percentage, the demand changes by a smaller percentage.
As Dr. Ugone explained, “if you change prices by 10
percent and quantity demanded changes by only 5 per-
cent . . . that’s an example of something we call inelastic.”
Dr. Ugone opined that the demand for Akamai’s prod-
ucts was relatively inelastic (i.e. relatively price-
insensitive) and provided two justifications for calculating
that 75% of Limelight’s sales would potentially have been
made by Akamai. First, because Akamai’s costs were
“revenue-generating costs,” customers would be more
willing to expend money to buy Akamai’s product. Sec-
ond, though there would be some “price sensitivity” such
that some of Limelight’s customers would not purchase
20 AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC.

the higher-priced Akamai product, the demand was
relatively inelastic – meaning the quantity demanded
would not change as much as the price changed. The
relative inelasticity of demand was supported by Aka-
mai’s evidence that Akamai and Limelight were direct
competitors, including statements by Limelight that 1)
Akamai was its largest competitor; 2) “Limelight and
Akamai are, from a scale and quality standpoint, head
and shoulders above the rest of . . . Limelight’s competi-
tion”; 3) demand was driven by end-users not customers;
and 4) Akamai maintained a dominant market share
despite Limelight’s infringing service and lower price. Dr.
Ugone conceded that in picking 75% he “had to make a
judgment call based on the attributes and come to a
conclusion what the adjustments would be.” Based on his
assumptions, Dr. Ugone determined that Akamai’s lost
profits were about $74 million.
The considerations outlined above sufficiently support
the district court’s decision to allow Dr. Ugone’s adjusted
lost-profits analysis. This court has repeatedly approved
similar adjusted market share analyses for estimating
lost profits. See, e.g. Ericsson, 352 F.3d at 1377–80.
There is no basis for Limelight’s claim that such an anal-
ysis here is legally unavailable.
Limelight’s argument appears to be that the price
disparity between Akamai’s and Limelight’s prices neces-
sarily created a market segmentation in which Akamai
was separate from Limelight. Limelight’s argument rests
on BIC Leisure Prods., Inc. v. Windsurfing International,
Inc., 1 F.3d 1214 (Fed. Cir. 1993), where this court deter-
mined that lost profits were unavailable because the
accused infringer and the patentee serviced different
markets based on a 60–80% price disparity. Limelight
argues that in the face of a 100% price disparity, lost
profits are legally unavailable. However, this court’s
decision in BIC did not rest solely on the price disparity of
the two companies. For one, the court noted that “[the
AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC. 21

patentee] Windsurfing concentrated on the One Design
class hull form and BIC [the infringer] did not. Windsurf-
ing’s boards differed fundamentally from BIC’s boards.”
Id. at 1218. Moreover, the court explained that “at least
fourteen competitors vied for sales in the sailboard mar-
ket.” Id. In the instant case, in contrast, Akamai pre-
sented evidence that Akamai and Limelight were direct
competitors, and the two leaders in the field, with capabil-
ities and infrastructure beyond those of its competitors.
Next, the court in BIC explained that the “record contains
uncontradicted evidence that demand for sailboards is
relatively elastic.” Id. Again, the instant case is different
- Dr. Ugone explained that the market was relatively
inelastic, and set forth a number of reasons, discussed
above, for this conclusion. Furthermore, the court noted
that Windsurfing had licensed its patent to two competi-
tors, both selling Boards similar to the patentee’s Boards
at significantly lower prices. Id. Finally, there was
evidence that Windsurfing’s “sales continued to decline
after the district court enjoined BIC’s infringement,” but
that the market share indeed went to one of the patent-
ee’s licensees. Id. No such evidence exists here.
In conclusion, Dr. Ugone’s 25% adjustment for market
elasticity was sufficiently grounded in economic principles
for the district court to allow it. Though Limelight is
correct that its customers expressed a clear preference for
lower-priced products — as evidenced by their buying
Limelight’s significantly cheaper product — and therefore
would have been less likely to buy Akamai’s products
than the average consumer, Dr. Ugone’s testimony took
this consideration into account both in excluding the
lowest 25% of Limelight’s customers from his lost profits
analysis, and for discounting the potential award for price
elasticity. Whether this discount was sufficient is not a
legal challenge to the availability of lost profits, but as to
the amount of lost profits, which Limelight failed to
address in its panel briefing.
22 AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC.

For the first time in supplemental briefing, Limelight
attempts to challenge the evidentiary basis for the
amount of the jury’s award, as distinct from the legal
challenge to the availability of lost profits. Also for the
first time in the supplemental briefing, Limelight argues
that Dr. Ugone’s 3% downward adjustment of Akamai’s
adjusted market share indicates an internal inconsisten-
cy, because the reduction was not carried through into the
lost profits calculation. Because the supplemental brief-
ing was limited to arguments contained in the panel
briefing, Akamai Techs., Inc. v. Limelight Networks, Inc.,
2009-1372 (Fed. Cir. Aug. 19, 2015) (non-precedential
order) (“[T]he parties are requested to file letter briefs
supplementing the original briefing of Limelight’s cross-
appeal and limited to the issues raised therein.” (emphasis
added)), we need not consider this argument. Even if we
were to consider it, we find it unconvincing.
Limelight also appears to argue that the jury’s dam-
ages award “was unfairly tainted by the district court’s
refusal to instruct the jury that it should reject a los[t]
profits claim based on speculative evidence.” This argu-
ment is without merit. The district court did instruct the
jury that “[t]he amount of lost profits must be proved with
reasonable certainty and cannot be left simply to specula-
tion.” The district court’s instructions captured the
applicable law.
III. CONCLUSION
The en banc court reversed the district court’s grant of
Limelight’s motion for JMOL of non-infringement of
the ’703 patent. For the foregoing reasons, we conclude
that the outstanding arguments in Limelight’s cross-
appeal have no merit. Thus, this case is remanded with
instructions to reinstate the jury verdict and the jury’s
damages award. This court’s previously reinstated affir-
mance of the district court’s judgment of non-
AKAMAI TECHNOLOGIES, INC. v. LIMELIGHT NETWORKS, INC. 23

infringement of the ’413 and ’645 patents is also re-
confirmed.
AFFIRMED-IN-PART, REVERSED-IN-PART, AND
REMANDED
COSTS
Each party shall bear its own costs.