Brooktree Corporation brought suit against Advanced Micro Devices, Inc. (herein AMD) for patent infringement, 35 U.S.C. § 271, and infringement of mask work registrations, 17 U.S.C. § 910, in connection with certain semiconductor chips used in color video displays. The United States District Court for the Southern District of California entered judgment that the patents were valid and infringed and that the registered mask works were infringed, assessing damages.1

The principal issues on appeal arise under the Patent Act, of which the Federal Circuit has exclusive appellate jurisdiction, 28 U.S.C. § 1295(a)(1), and the Semiconductor Chip Protection Act, of which this court's appellate jurisdiction is pendent. Thus for issues of fact and law under the Semiconductor Chip Protection Act we apply the discernable law of the Ninth Circuit, in accordance with the principles set forth in Atari, Inc. v. JS & A Group, Inc., 747 F.2d 1422, 1438-40, 223 U.S.P.Q. (BNA) 1074, 1086-87 (Fed.Cir.1984) (en banc ) (applying copyright law of the circuit in which the case was tried, thus avoiding creating new opportunities for forum shopping). Judicial consideration of the Semiconductor Chip Protection Act has thus far been sparse, and we have given particular attention to the statute and its history, for the parties dispute significant aspects of statutory interpretation.

This case occasioned a lengthy trial over the course of seven weeks before the jury, in consecutive determinations of liability and damages. The jury verdicts were the subject of duly filed motions for judgment notwithstanding the verdict and for a new trial, which motions were denied by the district court. AMD charges error on issues of mask work infringement and damages, and also on issues of patent validity, infringement, and willfulness. Brooktree cross-appeals certain damages rulings, and the denial of attorney fees under both the Patent Act and the Semiconductor Chip Protection Act.

* MASK WORKS

The Semiconductor Chip Protection Act of 1984, Pub.L. 98-620, Title III, 98 Stat. 3347, codified at 17 U.S.C. §§ 901-914, arose from concerns that existing intellectual property laws did not provide adequate protection of proprietary rights in semiconductor chips that had been designed to perform a particular function. The Act, enacted after extensive congressional consideration and hearings over several years, adopted relevant aspects of existing intellectual property law, but for the most part created a new law, specifically adapted to the protection of design layouts of semiconductor chips.

Chip design layouts embody the selection and configuration of electrical components and connections in order to achieve the desired electronic functions. The electrical elements are configured in three dimensions, and are built up in layers by means of a series of "masks" whereby, using photographic depositing and etching techniques, layers of metallic, insulating, and semiconductor material are deposited in the desired pattern on a wafer of silicon. This set of masks is called a "mask work", and is part of the semiconductor chip product. The statute defines a mask work as:

a series of related images, however fixed or encoded

(A) having or representing the predetermined, three dimensional pattern of metallic, insulating, or semiconductor material present or removed from the layers of a semiconductor chip product; and

(B) in which series the relation of the images to one another is that each image has the pattern of the surface of one form of a semiconductor chip product.

17 U.S.C. § 901(a)(2). The semiconductor chip product in turn is defined as:

the final or intermediate form of any product--

(A) having two or more layers of metallic, insulating, or semiconductor material, deposited or otherwise placed on, or etched away or otherwise removed from, a piece of semiconductor material in accordance with a predetermined pattern; and

(B) intended to perform electronic circuitry functions.

17 U.S.C. § 901(a)(1).

The design of a satisfactory chip layout may require extensive effort and be extremely time consuming, particularly as new and improved electronic capabilities are sought to be created. A new semiconductor chip may incur large research and development costs, yet after the layout is imprinted in the mask work and the chip is available in commerce, it can be copied at a fraction of the cost to the originator. Thus there was concern that widespread copying of new chip layouts would have adverse effects on innovative advances in semiconductor technology, as stated in the Senate Report:

In the semiconductor industry, innovation is indispensable; research breakthroughs are essential to the life and health of the industry. But research and innovation in the design of semiconductor chips are threatened by the inadequacies of existing legal protection against piracy and unauthorized copying. This problem, which is so critical to this essential sector of the American economy, is addressed by the Semiconductor Chip Protection Act of 1984.

* * * * * *

The Semiconductor Chip Protection Act of 1984, ... would prohibit "chip piracy"--the unauthorized copying and distribution of semiconductor chip products copied from the original creators of such works.

S.Rep. No. 425, 98th Cong., 2d Sess., 1 (1984) (hereinafter Senate Report ).

In the evolution of the Semiconductor Chip Protection Act it was first proposed simply to amend the Copyright Act, 17 U.S.C. § 101 et seq., to include semiconductor chip products and mask works as subject of copyright. See H.R. 1028, 98th Cong., 1st Sess. (1983). However, although some courts had interpreted copyright law as applicable to computer software imbedded in a semiconductor chip, see Apple Computer, Inc. v. Franklin Computer Corp., 714 F.2d 1240, 1249, 219 U.S.P.Q. (BNA) 113, 121 (3d Cir.1983), cert. dismissed, 464 U.S. 1033, 104 S. Ct. 690, 79 L. Ed. 2d 158 (1984), it was uncertain whether the copyright law could protect against copying of the pattern on the chip itself, if the pattern was deemed inseparable from the utilitarian function of the chip. Indeed, the Copyright Office had refused to register patterns on printed circuit boards and semiconductor chips because no separate artistic aspects had been demonstrated. Copyright Protection for Semiconductor Chips: Hearings on H.R. 1028 Before the Subcomm. on Courts, Civil Liberties, and the Administration of Justice of the House Comm. on the Judiciary, 98th Cong., 1st Sess., 77 (1983) (hereinafter 1983 House Hearings ) (statement of Dorothy Schrader, Associate Register of Copyrights for Legal Affairs). Concern was also expressed that extension of the copyright law to accommodate the problems of mask works would distort certain settled copyright doctrines, such as fair use. 1983 House Hearings at 16-17 (statement of Jon A. Baumgarten, Copyright Counsel, Association of American Publishers, Inc.)

The patent system alone was deemed not to provide the desired scope of protection of mask works. Although electronic circuitry and electronic components are within the statutory subject matter of patentable invention, see 35 U.S.C. § 101, and some original circuitry may be patentable if it also meets the requirements of the Patent Act, as is illustrated in this case, Congress sought more expeditious protection against copying of original circuit layouts, whether or not they met the criteria of patentable invention. Senate Report at 8.

The Semiconductor Chip Protection Act of 1984 was an innovative solution to this new problem of technology-based industry. While some copyright principles underlie the law, as do some attributes of patent law, the Act was uniquely adapted to semiconductor mask works, in order to achieve appropriate protection for original designs while meeting the competitive needs of the industry and serving the public interest.

The Semiconductor Chip Protection Act provides for the grant of certain exclusive rights to owners of registered mask works, including the exclusive right "to reproduce the mask work by optical, electronic, or any other means", and the exclusive right "to import or distribute a semiconductor chip product in which the mask work is embodied". 17 U.S.C. § 905. Mask works that are not "original", or that consist of "designs that are staple, commonplace, or familiar in the semiconductor industry, or variations of such designs, combined in a way that, considered as a whole, is not original", are excluded from protection. 17 U.S.C. § 902(b). Protection is also not extended to any "idea, procedure, process, system, method of operation, concept, principle, or discovery, regardless of the form in which it is described, explained, illustrated or embodied" in the mask work. 17 U.S.C. § 902(c).

The sponsors and supporters of this legislation foresaw that there would be areas of uncertainty in application of this new law to particular situations, and referred to "gray areas" wherein factual situations could arise that would not have easy answers. Those areas are emphasized by both parties in the assignments of error on this appeal.

Brooktree was granted mask work registration MW 2873 on August 6, 1987, and registration MW 3838 on July 6, 1988, for its chips identified as Bt451 and Bt458. These Brooktree chips embody a circuit design that combines the functions of a static random access memory (SRAM) and a digital to analog converter (DAC). This circuitry, sometimes referred to as RAMDAC, acts as a "color palette", producing the colors in color video displays having high speed and enhanced picture resolution. A Brooktree witness described these chips as a technological breakthrough, exceeding limits in speed and performance that had been believed impossible to exceed. Brooktree stated that a single Bt458 chip replaced a previously used set of 36 chips (an AMD product) and offered many advantages. A Brooktree witness testified that these chips were extremely successful commercially, and were soon incorporated into new designs for video display systems made by several large manufacturers.

A critical component of the Brooktree chips is the core cell, a ten-transistor SRAM cell which is repeated over six thousand times in an array covering about eighty percent of the chip area. Each core cell consists of ten transistors and metal conductors electrically connecting the transistors throughout the three dimensions of the multilayered cell. Brooktree charged that this core cell was copied by AMD, thus infringing Brooktree's mask work registrations.

AMD does not challenge the validity of these mask work registrations, or dispute Brooktree's position that its chips are protected under the Semiconductor Chip Protection Act. AMD does, however, assert that its accused chips are not infringements, for reasons we shall discuss.

The Semiconductor Chip Protection Act defines an "infringing semiconductor chip product" as one which is "made, imported, or distributed in violation of the exclusive rights" of the mask work owner. 17 U.S.C. § 901(a)(9). The text of the Semiconductor Chip Protection Act sets forth the subject matter of protection in terms of certain exclusive rights, including, inter alia, the exclusive right to "reproduce the mask work", 17 U.S.C. § 905. This usage mirrors the words of the Copyright Act, which states the exclusive rights of copyright owners "to reproduce the copyrighted work". 17 U.S.C. § 106. Although the Semiconductor Chip Protection Act does not use the word "copy" to describe infringement, the parallel language reflects the incorporation of the well-explicated copyright principle of substantial similarity into the Semiconductor Chip Protection Act, as discussed infra.

The jury instruction on the criteria for establishing infringement included the instruction that infringement requires substantial similarity to a material portion of the registered mask work:

To establish infringement, Brooktree must show that A.M.D.'s mask works are substantially similar to a material portion of the mask works in Brooktree's chips covered by Brooktree's mask work registration. No hard and fast rule or percentage governs what constitutes a, quote, "substantial similarity." Substantial similarity may exist where an important part of the mask work is copied, even though the percentage of the entire chip which is copied may be relatively small. It is not required that A.M.D. make a copy of the entire mask work embodied in the Brooktree chip.

AMD states that it does not on appeal challenge this jury instruction. Instead, AMD argues that because the non-SRAM portion of its accused chip was not copied, the chips are not "substantially similar", whatever the materiality of the SRAM cell to the total mask work. It was undisputed that there was not duplication of the entire chip. AMD states that the Semiconductor Chip Protection Act requires copying of the entire chip, and therefore that it was entitled to judgment in its favor as a matter of law, or at least to a new trial on the issue.

The principle of substantial similarity recognizes that the existence of differences between an accused and copyrighted work may not negate infringement if a material portion of the copyrighted work is appropriated. Shaw v. Lindheim, 919 F.2d 1353, 1362, 15 USPQ2d 1516, 1523 (9th Cir.1990). If the copied portion is qualitatively important, the finder of fact may properly find substantial similarity under copyright law, Baxter v. MCA, Inc., 812 F.2d 421, 425, 2 USPQ2d 1059, 1063 (9th Cir.), cert. denied, 484 U.S. 954, 108 S. Ct. 346, 98 L. Ed. 2d 372 (1987), and under the Semiconductor Chip Protection Act.

Brooktree agrees that the SRAM portion of the accused chips covers only eighty percent of the chip area, and that the remaining circuitry was not copied by AMD. Infringement under the statute does not require that all parts of the accused chip be copied. The district court's explanation to the jury was in full accord with the statutory grant of exclusive rights to reproduce the mask work. The statutory interpretation now pressed by AMD, viz., that the entire chip must have been copied, is unsupported. Indeed, the House Report states that it was contemplated that the cell layout alone could be misappropriated:

Mask works sometimes contain substantial areas of (so-called "cells") whose layouts involve creativity and are commercially valuable. In appropriate fact settings, the misappropriation of such a cell--assuming it meets the original standards of this chapter--could be the basis for an infringement action under this chapter.

H.R.Rep. No. 781, 98th Cong., 2d Sess., 26, 27 (1984), reprinted in 1984 U.S.Code Cong. & Admin.News 5708, 5750, 5775, 5776 (hereinafter House Report ) (footnote omitted). As explained in the Explanatory Memorandum--Mathias-Leahy Amendment to S. 1201, 130 Cong.Rec. S12,916 (daily ed. Oct. 3, 1984) (hereinafter Mathias-Leahy Memorandum ), written to explain the House and Senate bills as they were enacted, id. at S12,923, the Congressional intention was to protect against "piecemeal copying" and "copying of a material portion" of a chip:

... the amendment, like both bills, incorporates the familiar copyright principle of substantial similarity. Although as a practical matter, copying of an insubstantial portion of a chip and independent design of the remainder is not likely, copying of a material portion nevertheless constitutes infringement. This concept is particularly important in the semiconductor industry, where it may be economical, for example, to copy 75% of a mask work from one chip and combine that with 25% of another mask work, if the copies are transferable modules, such as units from a cell library.

As the Senate report notes, no hard and fast percentages govern what constitutes a "substantial" copying because substantial similarity may exist where an important part of a mask work is copied even though the percentage copied may be relatively small. Nonetheless, mask work owners are protected not only from wholesale copying but also against piecemeal copying of substantial or material portions of one or more mask works.

Mathias-Leahy Memorandum at S12,917.

Whether an appropriation of a cell layout constitutes infringement in a particular case is for the trier of fact, and can not be decided as a matter of law. See Sid & Marty Krofft Television Productions, Inc. v. McDonald's Corp., 562 F.2d 1157, 1164, 196 U.S.P.Q. (BNA) 97, 102-03 (9th Cir.1977) (question of substantial similarity under copyright law is for the trier of fact). We shall discuss this question post, for it was litigated in the context of AMD's defense of reverse engineering.

AMD's position at trial, and on appeal, was that its core cell was the product of reverse engineering of the Brooktree chip, and therefore does not constitute infringement under the Semiconductor Chip Protection Act. Reverse engineering is a statutory defense, included in the Act upon extensive congressional attention to the workings of the semiconductor chip industry.

The statute provides that it is not an infringement of a registered mask work for

(1) a person to reproduce the mask work solely for the purpose of teaching, analyzing, or evaluating the concepts or techniques embodied in the mask work or the circuitry, logic flow, or organization of components used in the mask work; or

(2) a person who performs the analysis or evaluation described in paragraph (1) to incorporate the results of such conduct in an original mask work which is made to be distributed.

17 U.S.C. § 906(a). The statute thus provides that one engaged in reverse engineering shall not be liable for infringement when the end product is itself original. In performing reverse engineering a person may disassemble, study, and analyze an existing chip in order to understand it. This knowledge may be used to create an original chip having a different design layout, but which performs the same or equivalent function as the existing chip, without penalty or prohibition. Congress was told by industry representatives that reverse engineering was an accepted and fair practice, and leads to improved chips having "form, fit, and function" compatibility with the existing chip, thereby serving competition while advancing the state of technology. Senate Report at 21.

Much attention was given by Congress and by witnesses to the question of how to determine whether a chip layout was born of legitimate reverse engineering or of copying. It was foreseen that there would be a "gray area" wherein the rights of the parties, on the facts of a particular chip design, would require resolution on a fact-dependent, case-by-case basis. The following colloquy illustrates concerns raised at the hearings:

Rep. EDWARDS: ... Is the chief reservation here the idea that reverse engineering, which all the witnesses agree is appropriate, might be confused with pirating and that any kind of reverse engineering might be interpreted under this law as pirating?

Mr. MacPHERSON (of Fairchild Camera & Instrument Corp.): I think that's one of the very strong concerns that we have, yes. There is a very gray area here in the very nature of reverse engineering, which would leave an individual engaged in that practice uncertain what his ultimate rights would be should he use that particular result in another product.

Copyright Protection for Imprinted Design Patterns on Semiconductor Chips: Hearings Before the Subcomm. on Courts, Civil Liberties, and the Administration of Justice of the House Comm. on the Judiciary, 96th Cong., 1st Sess., 66 (1979).

This aspect was explored over the several years of legislative gestation. The reverse engineering procedure was described by witnesses, and distinguished in purpose and mechanism from the copying against which the Semiconductor Chip Protection Act was intended to guard. It was explained that a person engaged in reverse engineering seeks to understand the design of the original chip with the object of improving the circuitry, the chip layout, or both. The presence of innovation and improvement was stressed as the hallmark of an original layout. A witness explained the difference as determining "was anything innovative done in the process or was it simply a reproduction of what was already there?" The Semiconductor Chip Protection Act of 1983: Hearings on S. 1201 Before the Subcomm. on Patents, Copyrights, and Trademarks of the Senate Comm. on the Judiciary, 98th Cong., 1st Sess., 84 (1983) (testimony of Stanley C. Corwin). Another witness explained that reverse engineering generally produces a "paper trail" recording the engineer's efforts to understand the original chip and to design a different version after reverse engineering:

"Whenever there is a true case of reverse engineering, the second firm will have prepared a great deal of paper--logic and circuit diagrams, trial layouts, computer simulations of the chip, and the like; it will also have invested thousands of hours of work. All of these can be documented by reference to the firm's ordinary business records. A pirate has no such papers, for the pirate does none of this work. Therefore, whether there has been a true reverse engineering job or just a job of copying can be shown by looking at the defendant's records. The paper trail of a chip tells a discerning observer whether the chip is a copy or embodies the effort of reverse engineering."

Senate Report at 22 (quoting statement of Leslie L. Valdasz, Senior Vice President, Intel Corporation). The Committee reports and the statements of the Semiconductor Chip Protection Act's supporters show the belief that evidence of the presence or absence of such a paper trail would significantly reduce the gray area between legitimate and illegitimate behavior. See Mathias-Leahy Memorandum at S12,917; House Report at 21.

Senators Mathias and Leahy explained that § 906(a) includes a provision

to clarify the intent of both chambers that competitors are permitted not only to study the protected mask works, but also to use the results of that study to design, distribute and import semiconductor chip products embodying their own original mask works.

* * * * * *

The end product of the reverse engineering process is not an infringement, and itself qualifies for protection under the Act, if it is an original mask work as contrasted with a substantial copy. If the resulting semiconductor chip product is not substantially identical to the original, and its design involved significant toil and investment, so that it is not mere plagiarism, it does not infringe the original chip, even if the layout of the two chips is, in substantial part, similar.

Mathias-Leahy Memorandum at S12,917.

In illuminating the meaning of "original" in the context of reverse engineering, Senators Mathias and Leahy distinguished between a substantial copy, on one hand, and the product of reverse engineering which might be similar to the original, but if not a substantial copy would not be an infringement. For the latter, the "paper trail" was expected to document efforts in "analyzing, or evaluating the concepts or techniques embodied in the mask work or the circuitry, logic flow, or organization of components used in the mask work", as the effort required would be reflected in the documents. Id.

AMD's defense was that its chips were independently designed after the Brooktree chips were subjected to reverse engineering to learn the Brooktree design. The question of whether AMD's activities were acceptable reverse engineering, or unacceptable copying, was explained to the jury as follows:

Reverse engineering is permitted and is authorized by the Chip Protection Act. It is not infringement of an owner's exclusive right and protected mask work for another person, through reverse-engineering, to photograph and to study the mask work for the purpose of analyzing its circuitry--correction--the circuitry, logic flow and organization of the components used in the mask work and to incorporate such analysis into an original mask work.

The end product of the reverse-engineering process may be an original mask work, and therefore not an infringing mask work, if the resulting semiconductor chip product is not substantially identical to the protected mask work and its design involved significant toil and investment so that it is not mere plagiarism.

You should place great weight on the existence of reverse paperwork trail in determining whether the defendant's mask work is an original mask work from reverse-engineering.

A.M.D. mask work constitutes an original mask work if A.M.D.'s mask work incorporates its own new design elements which offered improvements over or an alternative to Brooktree's mask work.

These instructions focus the jury on whether AMD produced an original mask work, as the statute requires. The instructions were not challenged by AMD on its motion for new trial or on appeal, and were adapted from AMD's proposed instructions. Brooktree calls the instructions "too lenient". Whether or not too lenient, they were not objected to by AMD at trial, and are not now criticized by AMD as incorrect. They are the law applied in this case. See Fed.R.Civ.P. 51; Herrington v. County of Sonoma, 834 F.2d 1488, 1500 n. 12 (9th Cir.1987), cert. denied, 489 U.S. 1090, 109 S. Ct. 1557, 103 L. Ed. 2d 860 (1989) (failure to object to a jury instruction precludes appellate review).

The factual premises of the issues of infringement and AMD's reverse engineering defense were extensively explored at trial, through examination and cross-examination of witnesses, exhibits, displays, and attorney argument. To summarize, AMD argued at trial, and repeats on this appeal, that it did not intend to copy Brooktree's layout, and did not do so. AMD pointed to its "paper trail" of its two and a half years of effort at a cost in excess of three million dollars. AMD stated that if its intent had been to copy the Brooktree layout, it would simply have directed duplication of the circuit layouts, requiring a matter of months, not years. AMD stressed differences between its chips and those of Brooktree, and pointed out its aversion to piracy. According to AMD, this case is not in the gray area where reasonable minds could differ over whether there was reverse engineering or copying. AMD argues that Congress could not have intended that mask work infringement be found in the circumstances of this case, and that this court should hold, as a matter of law, that AMD did not infringe Brooktree's mask work registrations.

Brooktree, on its part, argued that AMD's cell layout is not original, but was directly copied from Brooktree's SRAM core cell, and repeated 6,000 times. Brooktree stressed AMD's lengthy and expensive failures at designing a layout. Brooktree observed that AMD had incorrectly analyzed Brooktree's chip during its attempts at reverse engineering, and that throughout this entire period of attempted duplication of function, AMD was unable to come close to Brooktree's results. Brooktree pointed to the rapidity with which AMD changed to Brooktree's layout when the error in analysis was discovered, AMD immediately producing, without further experimentation, a substantially identical SRAM cell.

At trial there was extensive evidence of AMD's design efforts, including its full paper trail. Evidence included the following: William Plants, AMD's assigned designer of the SRAM core cell, testified that AMD initiated a project to design a color palette chip in January, 1986, based on Brooktree's announcement of such a chip the previous November and introduction of its Bt451 chip in February 1986. In April AMD obtained several Brooktree Bt451 chips. Plants opened up one of the chips and had it photographed, including blow-ups of the SRAM portion, which he said was of particular interest to him. After studying the chip he prepared a "reverse engineering report" in June, 1986. Plants incorrectly concluded that the Brooktree chip had eight transistors in each SRAM cell rather than ten. Plants then abandoned his efforts to design an SRAM cell based on six transistors, and attempted, over the next six months, to create an SRAM core cell design based on eight transistors. These attempts were not successful.

There was evidence of increasing pressure to complete the design, including suggestions by AMD supervisors that Plants reexamine the Brooktree chip because the project had gotten "bogged down". Plants did so some time in late January or early February. Plants testified that it was toward the end of January 1987 when he changed to a ten-transistor design in the same arrangement as Brooktree's design. Plants admitted that he did "solve the problem in a week" by adopting the ten-transistor design. Plants denied, however, that he learned of the ten-transistor design from his reexamination of the Brooktree chip, but explained that the design had been suggested to him by a job candidate he interviewed during that period. That candidate did not testify, and an AMD witness said he was dead.

There was extensive exploration at trial of the events during this period, of Plants' design processes, and of the timing and other details of the successful AMD ten-transistor design. For instance, Plants testified that his reexamination of the Brooktree chip was prompted not by pressure from his supervisors, but by a memory flash, like a "bolt from the blue", of something he had seen previously in the Brooktree chip.

In addition, there was extensive testimony and demonstrative and documentary evidence that the similarities in the circuit designs led to the substantial identity between the layouts of the Brooktree and AMD core cells. For example, Plants testified that in laying out the ten-transistor SRAM cell for fabrication in silicon, he never changed the locations of the ten transistors in his design, and never considered alternatives to Brooktree's transistor arrangement. Michael Brunolli, the engineer responsible for the layout of the Brooktree SRAM cell, testified that, in contrast, he had gone through about six changes of transistor layout before he achieved the final design.

Brooktree's expert witness Richard Crisp testified that the layout of the Brooktree SRAM cell was unique and original, and was not a commonplace or staple design. Crisp stated that the most important factor in layout of the SRAM cell is the location of the transistors. Crisp showed the jury exhibits picturing alternative layouts of electrically identical ten-transistor SRAM core cells in other chips, illustrating different layouts of the same circuitry. Plants had agreed that "you could conceivably come up with an infinite number of layouts" for the ten-transistor cell.

Plants admitted that the metal lines connecting the transistors in one layer of his layout were in the same sequence or order as the corresponding lines in the Brooktree chip, although his design had an extra metal line in it. Crisp testified that the extra metal line in AMD's layout did not relate to any function of the SRAM core cell itself. Plants also admitted that he originally included 45-degree angled portions in the polysilicon layer of his initial layout, a feature found in the Brooktree layout. Plants said he later removed the angles at the request of AMD's production engineers because it was incompatible with AMD's process methodology. Brunolli testified that the 45-degree angled portions arose in his design as a result of design rules used at Brooktree, and that he chose the angles to make his cell layout more compact.

Crisp testified that the transistors, which "form the substance of the cell" were "grouped together in the same way" in the AMD and Brooktree layouts. Crisp showed a videotape of the Brooktree and AMD SRAM cells side-by-side and then overlaid. (This tape was shown four times, including once in the jury room.) He stopped the tape at various points to illustrate the similarities in the layout of the two chips, and explained that the differences were primarily the result of AMD's smaller technology and removal of the 45-degree angles. Crisp testified that the differences between the cells were "very minor, trivial, insignificant", and "insignificant electrically", and that "the substance of the cells is the same." Plants had testified that minor differences in design layout would arise because the differing technologies of AMD and Brooktree resulted in each firm having different standard design rules for layout of semiconductor chips.

The standard of review is whether there was legally sufficient evidence (often called "substantial evidence") whereby a reasonable jury could have reached the verdict reached by this jury. Transgo, Inc. v. Ajac Transmission Parts Corp., 768 F.2d 1001, 1014, 227 U.S.P.Q. (BNA) 598, 602 (9th Cir.1985), cert. denied, 474 U.S. 1059, 106 S. Ct. 802, 88 L. Ed. 2d 778 (1986). Rule 50 as amended states that judgment as a matter of law2 shall be granted when "a party has been fully heard with respect to an issue and there is no legally sufficient evidentiary basis for a reasonable jury to have found for that party with respect to that issue". Fed.R.Civ.P. 50(a)(1) (1992). This standard posits that a reasonable jury would have assessed the credibility of witnesses, considered and weighed the evidence presented by both sides, and applied the law in accordance with the court's instructions.

A reasonable jury could have placed weight on the evidence that AMD did not use an alternative transistor configuration but used Brooktree's configuration. A reasonable jury could have inferred that AMD's paper trail, insofar as it related to the SRAM cell, related entirely to AMD's failures, and that as soon as the Brooktree chip was correctly deciphered by reverse engineering, AMD did not create its own design but copied the Brooktree design. A reasonable jury could have resolved credibility questions in favor of Brooktree. Issues of credibility of witnesses are for the jury, and are not amenable to appellate review. Nicholson v. Rushen, 767 F.2d 1426, 1427 (9th Cir.1985).

AMD argues on appeal that the sheer volume of paper established the reverse engineering defense "as a matter of law". AMD appears to misperceive the process, whereby the jury is instructed on the law, and applies the law to the facts and inferences reasonably drawn therefrom, based on the evidence adduced at trial. When there was sufficient evidence to support the verdict in light of the entire record, and upon correct or unobjected instructions of law, the jury's verdict must stand. Oltz v. St. Peter's Community Hospital, 861 F.2d 1440, 1450 (9th Cir.1988); Transgo, 768 F.2d at 1014, 227 USPQ at 602. Accord, e.g., Orthokinetics, Inc. v. Safety Travel Chairs, Inc., 806 F.2d 1565, 1571, 1 USPQ2d 1081, 1085 (Fed.Cir.1986); Connell v. Sears, Roebuck & Co., 722 F.2d 1542, 1546, 220 U.S.P.Q. (BNA) 193, 197 (Fed.Cir.1983). Reversal is proper only if the evidence permits only one reasonable conclusion. Venegas v. Wagner, 831 F.2d 1514, 1517 (9th Cir.1987). See generally M.B. Louis, Allocating Adjudicative Decision Making Authority Between the Trial and Appellate Levels: A Unified View of The Scope of Review, The Judge/Jury Question, and Procedural Discretion, 64 N.C.L.Rev. 993 (1986).

The jury was instructed on the law of the reverse engineering defense, and was told to place "great weight" on the paper trail. The correctness of a jury instruction is a question of law and, if objection to the instruction was timely made at trial, is reviewed on appeal to determine whether, on the whole, the jury instructions were adequate to ensure that the jury fully understood the legal issues for each element of the case. Benigni v. City of Hemet, 879 F.2d 473, 479 (9th Cir.1988); Los Angeles Memorial Coliseum Commission v. National Football League, 726 F.2d 1381, 1398 (9th Cir.), cert. denied, 469 U.S. 990, 105 S. Ct. 397, 83 L. Ed. 2d 331 (1984).

AMD's proposed interpretation of the reverse engineering defense is not supported in the statute. 17 U.S.C. § 906(a) requires, as an element of the defense of reverse engineering, that the product be original. The Senate Report described reverse engineering as activity that "spurs innovation and technological progress, as competitors seek to develop ever faster or more efficient chips, to perform similar or related functions". Senate Report at 21. The statute does not reflect an intent to excuse copying, as a matter of law, if the copier had first tried and failed to do the job without copying. The paper trail is evidence of independent effort, but it is not conclusive or incontrovertible proof of either originality or the absence of copying.

The questions of originality and copying are heavily fact-dependent, not absolutes of law. Where the evidence is such that reasonable minds could draw different conclusions, such questions are the province of the jury, as illustrated in analogous copyright cases, e.g., Twentieth Century-Fox Film Corp. v. MCA, Inc., 715 F.2d 1327, 1329, 217 U.S.P.Q. (BNA) 611, 612 (9th Cir.1983) (question of whether there was copying was a "close enough question" to require trial); Dezendorf v. Twentieth Century-Fox Film Corp., 99 F.2d 850, 851 (9th Cir.1938) (issue of originality is for the jury, not for the court).

AMD points out that there was conflicting evidence as to whether the AMD chips were improved. For example, AMD points to evidence that its accused chips were smaller, and as a result faster, than those of Brooktree, while Brooktree points to evidence that this was because AMD used 1.6 micron technology as opposed to Brooktree's older 2 micron technology. (The micron size refers to the length across a transistor gate.)

There was also extensive evidence of differences in the circuit layouts, and their significance. For example, AMD's expert, Dominick Richiuso, testified to specific differences between the AMD and Brooktree chips in the layouts of the interconnections between the SRAM cells and other portions of the chips, and in the SRAM cells themselves. These differences were acknowledged by Brooktree's expert Crisp, who testified that the similarities in the chip layouts, including the locations and interconnections of the transistors, were of far greater significance than the differences.

These fact-dependent areas were thoroughly aired at trial. The grant of judgment as a matter of law, by the trial judge or by the appellate court, is appropriate only when the evidence, with inferences drawn favorably to the party with the verdict, could not reasonably support the verdict. Dean v. Trans World Airlines, Inc., 924 F.2d 805, 810 (9th Cir.1991).

We conclude that there was a legally sufficient evidentiary basis whereby a reasonable jury could have found infringement of the mask work registrations. The judgment is affirmed.

The three Brooktree patents in suit were found valid and infringed. Two of the patents relate to portions of the Brooktree chips that were the subject of mask work registrations. As to these two patents, only validity is appealed. As to the third patent, only infringement is appealed.

U.S. Patent No. 4,831,282, issued May 16, 1989 to Joseph H. Colles, (the '282 patent), relates to a portion of the digital to analog conversion circuitry that allows operation at high frequencies without "glitch" errors (discontinuities in the analog output signal), under variable operating conditions such as power and temperature. AMD argues that the '282 patent is invalid under 35 U.S.C. § 101 and § 112, on the basis that the invention is inoperable as described in the specification.

If the claimed subject matter is inoperable, the patent may indeed be invalid for failure to meet the utility requirement of § 101 and the enablement requirement of § 112. Raytheon Co. v. Roper Corp., 724 F.2d 951, 956, 220 U.S.P.Q. (BNA) 592, 596 (Fed.Cir.1983), cert. denied, 469 U.S. 835, 105 S. Ct. 127, 83 L. Ed. 2d 69 (1984). To violate § 101 the claimed device must be totally incapable of achieving a useful result, see Tol-O-Matic, Inc. v. Proma Produkt-Und Marketing G.m.b.H., 945 F.2d 1546, 1552-53, 20 USPQ2d 1332, 1338 (Fed.Cir.1991); Moleculon Research Corp. v. CBS, Inc., 793 F.2d 1261, 1268, 229 U.S.P.Q. (BNA) 805, 811 (Fed.Cir.1986), cert. denied, 479 U.S. 1030, 107 S. Ct. 875, 93 L. Ed. 2d 829 (1987); Envirotech Corp. v. Al George, Inc., 730 F.2d 753, 762, 221 U.S.P.Q. (BNA) 473, 480 (Fed.Cir.1984), a determination of fact, Raytheon, 724 F.2d at 956, 220 USPQ at 596. In this case the questions of utility and enablement turned on the same disputed facts, and were treated similarly at the trial. See generally Raytheon, 724 F.2d at 959-60, 220 USPQ at 599; Moleculon Research, 793 F.2d at 1268, 229 USPQ at 811. We review the proceedings at trial to ascertain whether there was substantial evidence whereby a reasonable jury could have reached the verdict that the patent had not been proved invalid based on the asserted inoperability of the invention. Tol-O-Matic, 945 F.2d at 1552-53, 20 USPQ2d at 1338.

Substantial evidence is such relevant evidence as a reasonable mind could have accepted as adequate to support the conclusion that was reached by the jury, even if it is possible to draw inconsistent conclusions from the evidence. St. Elizabeth Community Hospital v. Heckler, 745 F.2d 587, 592 (9th Cir.1984)3; Landes Construction Co., Inc. v. Royal Bank of Canada, 833 F.2d 1365, 1370-71 (9th Cir.1987); Orthokinetics, 806 F.2d at 1571, 1 USPQ2d at 1085.

Brooktree's witnesses explained to the jury that the '282 invention embodies a three-transistor switch that provides output current to create the color image on the video screen. It was explained that the invention enabled operation at much higher speeds than were previously believed obtainable. The circuit is illustrated in Figure 1 of the patent:NOTE: OPINION CONTAINS TABLE OR OTHER DATA THAT IS NOT VIEWABLE

It was explained at trial that when the circuit operates in the voltage mode, the current mode of the circuit is not operational. The specification states that when the circuit operates in the current mode, switch 22 is connected in the upper position shown in the drawing, closing a circuit including voltage source 18, transistor 16, fixed resistance 17, and ground. This circuit produces a constant current in the transistors 16, 26, 28, 30, and 32. The current flowing through these transistors, in conjunction with a resulting constant voltage on the source of transistor 34, results in a constant voltage bias applied to the gates of the digital-to-analog converting transistors 36, 38, and 40, which receive digital input signals from transistors 42, 44, and 46, respectively. In the current mode of operation, voltage source 12 and operational amplifier 14 are inoperative in the circuit.

When the circuit operates in the voltage mode switch 22 is in the down position, and voltage reference source 12 and amplifier 14 are included in the circuit along with voltage source 18, resistance 17, and transistor 16. In this mode, source 12 supplies constant biasing voltage directly to the converting transistors 36, 38, and 40 by directly establishing constant voltage on the source of transistor 34. The constant voltage bias on transistors 28, 30, and 32 is provided from operational amplifier 14, which provides a stabilizing function, by means of a feedback loop including resistor 17.

AMD argued that its accused chips never contain both a constant current source and a reference voltage source at the same time. AMD argued that all the claims require a "means for introducing a reference voltage ... at first particular times" and "a means ... for providing a substantially constant current at second particular times different from the first particular times" (from claim 1). AMD contended that one of these elements is always missing in either configuration of its device. AMD thus argued that it avoids the following limitation, as represented by this clause of claim 1:

... means responsive to the output of the operational amplifier at the first particular times [voltage mode] for overriding the operation of the last mentioned means to introduce the reference voltage to the transistors in the second plurality to maintain the substantially constant current through such transistors and to obtain the production of the substantially constant voltage from such transistors, ...

AMD's expert, Mr. Camenzind, testified that the AMD circuit performs no overriding because in its circuitry the current source and the voltage source are not present simultaneously. AMD argues that in order for there to be "overriding" the claims must be interpreted as requiring the simultaneous presence of both the current source (resistance 17 in conjunction with voltage 18) and the voltage source 12; otherwise there would be nothing in the circuit to be overridden.

Brooktree's expert, Dr. Martin, testified that according to the '688 patent the circuit is used alternatively with either a current reference or a voltage reference input. He testified that one skilled in the art would understand that the two modes could not be used at the same time. Dr. Martin explained that the Brooktree chips manufactured in accordance with the patent were useful in either mode, depending on between which pins a "jumper", corresponding to switch 22, is connected. Martin also testified that in the accused AMD chips some connection must be made by the user in order to make operative either the voltage or the current mode, and that such connection is achieved by a "jumper". He stated that the electrical functioning of the accused chips was identical to that described in the patent, and that every claimed element in the patent was present in the accused chips. He presented claim charts and diagrams, identifying for each claim element the corresponding element in the accused circuitry. The specification describes the overriding function as follows:

When the reference voltage is introduced to the reference generator 10, the reference generator operates to override the stages producing the substantially constant voltage during the introduction of the substantially constant current.

The jury was instructed on patent infringement as follows:

A product infringes a claim of a patent when it contains each element of the invention as defined by the particular patent claim: If an A.M.D. product contains each element of any single claim of a Brooktree patent, the A.M.D. product ... infringes that claim....

Determining patent infringement requires examination of the patent claims and a comparison of those claims to the alleged infringing product, not a comparison of the accused product and the patentee's product.

To determine infringement, you must first construe the meaning and scope of the claims and, second, compare the properly construed claims to the alleged infringing device. You should refer to extrinsic evidence, including specifications, prosecution history, and other claims to construe the meaning and scope of the claims.

A reasonable jury, applying this law to the evidence, could have found that the specification does not support AMD's contention that "overriding" implies the simultaneous presence of reference voltage and constant current sources. There was substantial evidence whereby the jury, having interpreted the claims to mean that the two different modes (introducing a reference voltage or providing a constant current) operate at different times, could have applied this interpretation to find that the '688 patent was infringed by AMD's accused chips, thus providing a "legally sufficient evidentiary basis" for upholding the verdict. Fed.R.Civ.P. 50(a)(1). See Snellman v. Ricoh Co., Ltd., 862 F.2d 283, 288, 8 USPQ2d 1996, 1999-2000 (Fed.Cir.1988), cert. denied, 491 U.S. 910, 109 S. Ct. 3199, 105 L. Ed. 2d 707 (1989) (substantial evidence supported jury's apparent claim interpretation); Tol-O-Matic, 945 F.2d at 1549-52, 20 USPQ2d at 1335-38 (reviewing jury verdict to ascertain whether reasonable jurors could have interpreted the claim in a way that supports the verdict of infringement); Palumbo v. Don-Joy Co., 762 F.2d 969, 974, 226 U.S.P.Q. (BNA) 5, 8 (Fed.Cir.1985) (when the interpretation of a claim term is disputed, it "should be left to the trier or jury under appropriate instruction").

The judgment of infringement of the '688 patent is affirmed.

The Semiconductor Chip Protection Act defines an infringing product as one "made, imported, or distributed in violation of the exclusive rights of the owner of a mask work". 17 U.S.C. § 901(a)(9). In accordance with 17 U.S.C. § 911(b), damages are measured as the "actual damages suffered by [the owner of the mask work] as a result of the infringement" plus "the infringer's profits that are attributable to the infringement and are not taken into account in computing the award of actual damages."

The jury was not asked to separate the damages for mask work infringement and for patent infringement, and no issue is raised as to any difference in the measurement of damages. The parties do not disagree that the appropriate measure is Brooktree's lost profits due to the infringement. The disagreement relates to how Brooktree's lost profits should have been measured in this case.

AMD argued that it was incorrect to consider price reductions made by Brooktree after AMD announced its chips but before they were sold by AMD, and other competitive price reductions made by Brooktree after AMD entered the market. AMD contended, and Brooktree readily agreed, that Brooktree reduced its price following announcements by AMD that it would market competing chips at a lower price than Brooktree's then-current price. AMD asserted that these were voluntary, tactical actions by Brooktree starting in April 1988, and that AMD did not begin commercial sales of its chips until August 1988. Brooktree responded with evidence that AMD was providing samples and quoting prices at least as early as February, 1988.

AMD argues on appeal that the Semiconductor Chip Protection Act provides only for damages "as a result of the infringement", 17 U.S.C. § 911(b), and therefore that the damage calculations were incorrect as a matter of law, since they included events occurring before any infringement began. AMD argues that it was improper to permit the jury to decide whether the Brooktree price reductions were "a result of the infringement", and that the district court should have held as a matter of law that Brooktree's price reductions, made before there were infringing sales, should not be included. Brooktree responds that the measurement of damages is a classical question of fact; that the issue was properly presented to the jury; that it was the subject of extensive evidence at trial; and that AMD did not object to the jury instructions as to the legal standard to be applied.

As in damage determinations in general, the measurement of actual damages is a question of fact, Bubar v. Ampco Foods, Inc., 752 F.2d 445, 449 (9th Cir.), cert. denied, 472 U.S. 1018, 105 S. Ct. 3481, 87 L. Ed. 2d 616 (1985), including the question of whether AMD's premarketing activities caused financial losses to Brooktree. Causation is a classical jury question. See Union Insurance Co. v. Smith, 124 U.S. 405, 423, 8 S. Ct. 534, 544, 31 L. Ed. 497 (1888); Lies v. Farrell Lines, Inc., 641 F.2d 765, 770 (9th Cir.1981).

The jury was instructed that actual damages could be based on finding that Brooktree could have charged higher prices but for AMD's infringing activities:

Brooktree's actual damages for mask work infringement may be its lost profits caused by A.M.D.'s infringement. Brooktree may establish lost profits from A.M.D.'s mask work infringement by establishing, by a preponderance of the evidence, that--and there are two elements: one, A.M.D.'s infringement of Brooktree's mask work reduced A.M.D.'s time to market and allowed A.M.D. to make sales earlier than it otherwise would have made, had it not infringed Brooktree's mask work; and, two, Brooktree would have made additional sales or charged higher prices had A.M.D. not made these earlier sales.

To recover lost profits for mask work infringement, Brooktree must establish with reasonable probability that, but for A.M.D.'s infringement, Brooktree would have made additional profits.

AMD did not object to this instruction, which accommodates consideration of the price reductions made by Brooktree when AMD announced its competing products.

In patent cases, as in other commercial torts, damages are measured by inquiring: had the tortfeasor not committed the wrong, what would have been the financial position of the person wronged? See Aro Manufacturing Co. v. Convertible Top Replacement Co., 377 U.S. 476, 507, 84 S. Ct. 1526, 1543, 12 L. Ed. 2d 457, 141 U.S.P.Q. (BNA) 681, 694 (1964) ("had the infringer not infringed, what would [the patentee] have made?"). In TWM Manufacturing Co., Inc. v. Dura Corp., 789 F.2d 895, 902, 229 U.S.P.Q. (BNA) 525, 529 (Fed.Cir.), cert. denied, 479 U.S. 852, 107 S. Ct. 183, 93 L. Ed. 2d 117 (1986), this court sustained a damages award that included recovery for price reductions occasioned by infringing activities. In Lam, Inc. v. Johns-Manville Corp., 718 F.2d 1056, 1067, 219 U.S.P.Q. (BNA) 670, 676 (Fed.Cir.1983), we sustained a damages award that included recovery for losses due to reduced prices, based on evidence that the patentee reduced its prices to meet the infringer's competition. See also Kalman v. Berlyn Corp., 914 F.2d 1473, 1485, 16 USPQ2d 1093, 1102 (Fed.Cir.1990) (affirming damages award including recovery for depressed prices due to infringement). In a common-law copyright action, Meta-Film Associates, Inc. v. MCA, Inc., 586 F. Supp. 1346, 1360-61, 222 U.S.P.Q. (BNA) 211, 222 (C.D.Cal.1984), the court held that damages for copyright infringement could be recovered for losses suffered before the publication of the infringing film, when the loss was caused by the known pending production of the film.

Applying the general principle that the measure of recompense is the actual loss due to the infringement, then losses incurred upon announcement by AMD of the infringing activity may be included, when the losses are found to be reasonably related to the infringing activity. We take note that for products of high technology, the innovator's opportunity to recover the investment made in research and development is often concentrated in the few years immediately following introduction of a new product. Thus we benefit from the guidance of patent precedent,4 in this combined patent/mask work damages calculation. We also note the analogy of this portion of the Semiconductor Chip Protection Act to copyright law, wherein actual damages may be measured in terms of the diminished market value of the copyrighted work. Frank Music Corp. v. Metro-Goldwyn-Mayer, Inc., 772 F.2d 505, 512, 227 U.S.P.Q. (BNA) 687, 690-91 (9th Cir.1985), citing 3 M.B. Nimmer, Nimmer on Copyright § 14.02, at 14-6 (1985).

Brooktree presented evidence that it was forced to reduce its prices when AMD announced its chips at lower prices, and that but for the infringement Brooktree would have continued to sell its chips at the prices that had already been established, and at which Brooktree had been making substantial sales since 1986. A witness for Brooktree projected price reductions of ten percent per year in the absence of infringement, and compared this projection with Brooktree's actual price reduction of thirty percent during the first year of AMD's entry.

There was conflicting evidence of the nature and timing of AMD's activities in the marketplace. Brooktree's sales manager, Mr. Van Eynde, testified that AMD engaged in "pre-selling" its chips in early 1988. Van Eynde testified that AMD promised customers who had bought Brooktree's chips that AMD had a product that would soon be available, and quoted prices lower than those of Brooktree. Van Eynde testified that Brooktree was forced to lower its prices at that time or lose business that it had already developed. In evidence was an internal AMD document of October 1987 entitled "Product Strategy", stating AMD's strategy to "identify users of the Bt458 and steal sockets [into which the chips are placed] based on price and delivery." A memorandum of February 11, 1988 from AMD to IBM quoted prices on a color palette chip and offered to provide samples. An AMD status report dated April 12, 1988 states that the AMD chip was successfully demonstrated at the National Computer Graphics Conference.

AMD argues that even if a cause-effect relationship were found, infringement damages can not include these initial price reductions, as a matter of law, because at the time of Brooktree's initial price reductions in response to AMD's sales activities, Brooktree could not have known that the AMD products would infringe Brooktree's mask work registrations and Brooktree's later-issued patents. AMD states that Brooktree would have lowered its prices to meet AMD's competition regardless of whether AMD's products were an infringement of the mask work registrations or patents.

Brooktree states that since the products did indeed infringe, AMD was incurring liability for infringement whether or not Brooktree knew that AMD was infringing. Brooktree states that this liability arose out of AMD's activities, and is not limited to when Brooktree confirmed the structure of AMD's chips and the infringing nature of AMD's activities.

Liability for actual damages may accrue before the infringement has been verified or proved in court, as discussed supra. Whether competitive price reductions are properly included as lost profits is a question of fact, which turns on the relationship between AMD's actions and Brooktree's price reductions. The district court, in declining to grant judgment n.o.v., stated:

AMD's contentions are based on one theory as to when its activities first impacted on Brooktree's sales. Brooktree's contentions are based upon another theory. Both theories find support in the evidence at trial. AMD had the opportunity at trial to make its arguments as to the proper measure of lost profit damages during trial. As the verdict now stands, the jury's determination finds support in the evidence presented.

Brooktree Corp., 757 F. Supp. at 1095, 18 USPQ2d at 1699.

Applying the standard of appellate review of a jury award of damages, the jury's finding must be upheld unless the amount is "grossly excessive or monstrous", clearly not supported by the evidence, or based only on speculation or guesswork. Los Angeles Memorial Coliseum Commission v. National Football League, 791 F.2d 1356, 1360 (9th Cir.1986), cert. denied, 484 U.S. 826, 108 S. Ct. 92, 98 L. Ed. 2d 53 (1987). See also Weinar v. Rollform, Inc., 744 F.2d 797, 808, 223 U.S.P.Q. (BNA) 369, 375 (Fed.Cir.1984), cert. denied, 470 U.S. 1084, 105 S. Ct. 1844, 85 L. Ed. 2d 143 (1985) (affirming jury damage award where "the jury had every right to believe the evidence presented to it respecting a proper damage award sufficient to compensate ... for the injury"). To the extent that there were conflicts in the evidence, neither the trial court upon motion for judgment n.o.v. nor the appellate court may substitute its choice of result for that of the jury. Rinker v. County of Napa, 831 F.2d 829, 831 (9th Cir.1987); DMI, Inc. v. Deere & Co., 802 F.2d 421, 427, 231 U.S.P.Q. (BNA) 276, 280 (Fed.Cir.1986).

We conclude that there was a legally sufficient evidentiary basis in the record from which a reasonable jury could have concluded that Brooktree's price reductions were made as a result of AMD's actual and announced marketing of the infringing chips, and, accordingly, included these price reductions in the calculation of damages.

Brooktree argues, on cross-appeal, that it is also entitled to damages for future lost profits for a limited period (two years was proposed) after cessation of infringement due to the price erosion that occurred as a result of infringement, and that Brooktree asserted could not be recovered as a matter of practical marketing. The district court held that determination of such future estimated damages was speculative, and declined to present the question to the jury.

Although projected future losses may be recovered when sufficiently supported, see Lam, 718 F.2d at 1068, 219 USPQ at 678, the amount of lost profits awarded must not be speculative, Bio-Rad Laboratories, Inc. v. Nicolet Instrument Corp., 739 F.2d 604, 616, 222 U.S.P.Q. (BNA) 654, 664 (Fed.Cir.), cert. denied, 469 U.S. 1038, 105 S. Ct. 516, 83 L. Ed. 2d 405 (1984). The burden of proving future injury is commensurately greater than that for damages already incurred, for the future always harbors unknowns. We take note of the discussion before the trial judge of the uncertainties of future pricing, future competition, and future markets, in this fast-moving field, as well as the requirements of proof of future losses. Brooktree has not shown that the district court erred in determining that the evidence was too speculative to meet the threshold requirements for a sustainable jury verdict.

The damages award is affirmed. Brooktree's cross-appeal on this issue is denied.

The Semiconductor Chip Protection Act, like the Copyright Act, contains no provision authorizing or prohibiting the enhancement of damages. The patent statute, in contrast, authorizes the court to increase damages up to three times the amount assessed as actual damages. 35 U.S.C. § 284. Brooktree appeals the district court's refusal to award enhanced damages, and states that the jury's finding that AMD's patent infringement was willful should have led to enhancement of damages.

A finding of willful patent infringement supports, but does not compel, enhancement of damages. S.C. Johnson & Son, Inc. v. Carter-Wallace, Inc., 781 F.2d 198, 201, 228 U.S.P.Q. (BNA) 367, 369 (Fed.Cir.1986). The increase of damages based on willfulness of the infringement is within the discretionary authority of the trial court, "informed by the court's familiarity with the matter in litigation and the interest of justice." Id.; Rite-Hite Corp. v. Kelley Co., 819 F.2d 1120, 1126, 2 USPQ2d 1915, 1919 (Fed.Cir.1987).

In exercising its discretion the trial court may consider the weight of the evidence of willfulness. The court thus considers the culpability of the infringer with due attention to any circumstances in mitigation, the closeness of the question, and any other factors pertinent to the award, in that case, of exemplary damages. In Modine Manufacturing Co. v. Allen Group, Inc., 917 F.2d 538, 543, 16 USPQ2d 1622, 1626 (Fed.Cir.1990), cert. denied, --- U.S. ----, 111 S. Ct. 2017, 114 L. Ed. 2d 103 (1991), this court affirmed the district court's refusal to award enhanced damages, where the question of willfulness was "sufficiently close on the evidence". See also S.C. Johnson & Son, Inc. v. Carter-Wallace, Inc., 231 U.S.P.Q. (BNA) 668, 669, 1986 WL 494 (S.D.N.Y.1986) (refusing to increase damages because the case was "close").

Brooktree's charge of willfulness was based on evidence that AMD continued to produce and sell its infringing products after being notified of the issuance of each of Brooktree's patents, and the absence of any evidence of a good faith belief by AMD of patent invalidity or non-infringement. No written opinion of counsel on either invalidity or noninfringement of any of the three Brooktree patents was offered by AMD. In response, AMD pointed out that none of the three patents had issued at the time that AMD's challenged activities began, or at the time the lawsuit was filed by Brooktree. AMD states that it was unaware of the issuance of these patents until it was charged with infringement thereof, by amended complaint.

The district court observed that although there was sufficient evidence of willfulness to sustain the jury verdict, "the evidence supporting the jury's finding was not as strong as it could have been, and is not of the weight and strength that would support the imposition of enhanced damages." Brooktree Corp., 757 F. Supp. at 1097, 18 USPQ2d at 1700.

Viewing the arguments on both sides, and with appropriate deference to the trial court's close familiarity with the matter, on this question of punitive enhancement of damages we discern no abuse of discretion in the court's ruling. The denial of enhanced damages is affirmed.

In view of our affirmance of the denial of enhanced damages (and of attorney fees, infra ), we need not reach AMD's appeal of the verdict of willful infringement.

The patent statute authorizes the award of attorney fees to the prevailing party "in exceptional cases". 35 U.S.C. § 285. The award of attorney fees is within the informed discretion of the trial court. Del Mar Avionics, Inc. v. Quinton Instrument Co., 836 F.2d 1320, 1329, 5 USPQ2d 1255, 1262 (Fed.Cir.1987). The statutory purpose of such award is to reach cases where the interest of justice warrants fee-shifting. Thus the trial court has broad discretion in the criteria by which it determines whether to award attorney fees. See generally Rohm & Haas Co. v. Crystal Chemical Co., 736 F.2d 688, 690-91, 222 U.S.P.Q. (BNA) 97, 99 (Fed.Cir.1984) (attorney fees awarded to prevent gross injustice). We have held that a finding of willful infringement meets the criteria of "exceptional case", Del Mar Avionics, 836 F.2d at 1329, 5 USPQ2d at 1262, but that fee-shifting does not necessarily follow. S.C. Johnson, 781 F.2d at 201, 228 USPQ at 369. All of the circumstances must be considered.

The district court, explaining its denial of attorney fees, remarked on the closeness of the question of willfulness, and declined to find this an "exceptional case", the sine qua non of this fee-shifting statute. For fee-shifting issues particular deference is due to the trial judge, who had the opportunity to observe those intangibles missing from the appellate record. Id. The trial court is in the best position to evaluate the conduct of the parties throughout the period of patent infringement, and to allocate the burdens of litigation commensurate with the interest of justice.

We affirm the denial of an award of attorney fees under the patent statute.

Costs and Attorney Fees Under the Semiconductor Chip Protection Act

The Semiconductor Chip Protection Act provides that in

any civil action arising under this chapter, the court in its discretion may allow the recovery of full costs, including reasonable attorneys' fees, to the prevailing party.

17 U.S.C. § 911(f). This provision is commensurate with 17 U.S.C. § 505 of the copyright statute. Both parties argued to the district court that copyright precedent applies to § 911(f), and the district court relied on Ninth Circuit copyright precedent in its decision. On this appeal both sides rely on Federal Circuit precedent in patent cases.

Many aspects of the Semiconductor Chip Protection Act are based on copyright concepts. The House Report states that this section "provides for counsel fees, similar to 17 U.S.C. § 505." House Report at 28, 1984 U.S.Code Cong. & Admin.News 5777. In contrast with the attorney fee provision of the patent law, 35 U.S.C. § 285, there is no requirement for a threshold finding of "exceptional case". However, the award under the Semiconductor Chip Protection Act continues to be consigned to the court's discretion. We conclude that copyright law principles are more aptly applied to the application of 17 U.S.C. § 911(f).

The Ninth Circuit has held that attorney fees are generally awarded to a prevailing plaintiff under § 505 of the Copyright Act, explaining that the fee-shifting provision is intended to encourage assertion of "colorable" copyright claims, to deter infringement, and to make the plaintiff whole. McCulloch v. Albert E. Price, Inc., 823 F.2d 316, 323, 3 USPQ2d 1503, 1508 (9th Cir.1987). However, attorney fees are not awarded in every case, id., and fees may be denied in various situations, including:

(1) the presence of a complex or novel issue of law that the defendant litigates vigorously and in good faith, (2) the defendant's status as innocent, rather than willful or knowing, infringer, (3) the plaintiff's prosecution of the case in bad faith, and (4) the defendant's good faith attempt to avoid infringement.

Id. (citations omitted). The district court relied on McCulloch, and in particular the first factor, remarking that this was a case of first impression under the Semiconductor Chip Protection Act, and that it involved complex technology as well as novel issues of law.

Attorney fee decisions under the Copyright Act are reviewed under the abuse of discretion standard. Transgo Inc., 768 F.2d at 1026-27, 227 USPQ at 612. The denial of attorney fees under 17 U.S.C. § 911(f) was not, in these circumstances, an abuse of the court's discretionary authority, and is sustained.

The judgment entered on the jury verdicts, and the district court's rulings, are

AFFIRMED.