Marvell Technology Group Ltd.
Frequently Asked Questions Concerning the CMU Litigation
(Revised on January 28, 2013)
As disclosed by Marvell Technology Group Ltd. (“Marvell”) in a press release dated December 27, 2012, on December 26, 2012, a jury in Pittsburgh delivered a verdict in a lawsuit brought by Carnegie Mellon University (“CMU”) against Marvell in the United States District Court for the Western District of Pennsylvania. The jury found that the two CMU patents at issue were literally and willfully infringed and valid, and awarded damages in the amount of $1.17 billion. As stated in Marvell’s December 27 press release, Marvell believes that the evidence and the law do not support the jury’s findings and the award of damages and will seek to overturn the verdict in post-trial motions before the District Court and, if necessary, to appeal to the U.S. Court of Appeals for the Federal Circuit in Washington, D.C.
Marvell is providing the following FAQs as of January 7, 2013 to provide additional information to Marvell stakeholders and partners regarding the CMU litigation and to elaborate on Marvell’s positions described in the December 27 press release. Marvell has compiled the following from publicly available sources including the proceedings of the litigation. Marvell believes that additional details regarding Marvell’s position regarding the jury verdict and the litigation may further clarify the status of the CMU litigation.
1. What is Marvell’s non-infringement position?
Marvell’s chips address media noise in a fundamentally different way than the claims in the two CMU patents at issue, U.S. Patent Nos. 6,201,839 and 6,438,180. Marvell’s Media Noise Processor (“MNP”) and Non-Linear Viterbi Detector (“NLD”) features use a simple Viterbi detector, along with either pre- or post-processing. While CMU’s patents claim a theoretical technique, this technique is so complex that it cannot be implemented in real-world silicon chips. Marvell developed its own unique approach, which could be implemented in actual chips. More specifically, Marvell’s chips do not, as required by CMU’s patent claims, determine branch metric values in a trellis by selecting a branch metric function from a set of functions and by applying the functions to a plurality of signal samples.
Marvell’s simulation computer programs for its MNP and NLD features do not infringe for the same reasons. Moreover, a separate computer program designed by Marvell to simulate the theoretical performance of CMU's algorithm was used for evaluation purposes and as a benchmark tool and was not incorporated into any Marvell chips. None of these simulation computer programs infringe as they are not detectors that process signal samples, as required by CMU’s patent claims, but rather are merely computer software programs that process data from text files. [January 7, 2013]
2. Does Marvell have its own patents that cover its MNP feature and its NLD feature?
Marvell has sought and been awarded more than 100 patents on its read channel technologies, including U.S. Patent Nos. 6,931,585 and 7,155,660, which are directed to its media noise post-processor. For example, the '585 patent covering Marvell’s MNP post-processor was jointly developed in-house at Marvell by co-inventors Dr. Zining Wu and Mr. Gregory Burd. Their work was based on an earlier Marvell post-processor architecture developed in 1998 and patented by Marvell’s Dr. Andrei Vityaev. In addition, Marvell was awarded a patent on its NLD feature, U.S. Patent No. 8,160,181. The ‘181 patent was invented by Dr. Wu, Hongxin Song, Seo-How Low, and Panu Chaichanavong. [January 7, 2013]
3. Why are the CMU patents invalid, in Marvell’s view?
CMU’s patents are anticipated by the prior work done by Seagate, including the work reflected in U.S. Patent No. 6,282,251. This prior art patent was invented by Mr. Glen Worstell, a Seagate engineer. The CMU patents stated that the difference between the prior art and the CMU patents was that the prior art methods took into account signal dependent noise in the Viterbi detector, but failed to take into consideration correlated noise. See Col. 1:57-67 of U.S. Patent No. 6,201,839 . This was incorrect. In fact, the Seagate patent took into consideration the correlation between noise samples in the read back signal. This is plain from the very title of Seagate’s patent 6,282,251: “Modified Viterbi Detector Which Accounts For Correlated Noise.” See also, for example, Column 2, lines 3-7 of the Seagate patent. The Seagate patent was filed on March 21, 1995, approximately 14 months before the CMU inventors, Drs. Kavcic and Moura, even conceived of their alleged invention.
What’s more, Mr. Worstell informed CMU that he himself had done work on a “Viterbi detector modification to account for noise correlation.” But CMU never informed the U.S. Patent Office about this communication from Seagate or about the work at Seagate. As a result, the U.S. Patent Office did not consider the work of Seagate in its examination of the CMU patent applications. Indeed, throughout the prosecution of both patents, CMU never cited a single prior art patent reference to the U.S. Patent Office, other than cross-referencing in the ‘180 patent the prior art patents already cited by the U.S. Patent Office in the ‘839 patent. [January 7, 2013]
4. If the Seagate patent invalidates that CMU patent claims, why did the Seagate inventor say in an email that the CMU invention goes beyond his work and is probably more interesting?
The Seagate inventor reviewed an early invention disclosure of the CMU patent. The disclosure did not include any patent claims, and only included equations using covariance matrices. The Court granted Marvell’s motion for summary judgment of non-infringement of several CMU patent claims that require the use of covariance matrices as Marvell's chips do not use such matrices. Even if the use of covariance matrices went beyond the Seagate patent or was probably more interesting, it has no bearing on the claims CMU was asserting at trial, which do not require the use of covariance matrices. [January 7, 2013]
5. Why did CMU sue Marvell only and not any other silicon providers?
CMU’s attorneys have not indicated why they have not sued any other companies. Marvell is the market leader in HDD silicon and therefore an attractive target. [January 7, 2013]
6. How did the jury derive the $1.17B damage award?
CMU’s attorneys sought $1.17B in damages, based on 50 cents for every chip Marvell has sold worldwide since March 6, 2003. [January 7, 2013]
7. Why is 50 cents per chip not a reasonable royalty?
Marvell believes the 50 cents per chip and 1.17 billion dollar damages sought by CMU led the jury to an erroneous result for several reasons.
First, CMU’s damages expert ignored the most relevant evidence regarding the form and amount of a reasonable royalty payment in this case – i.e., the real world licenses, offers, and projections involving the patents-in-suit. For example, CMU previously offered licenses to the patents in suit for as little as a one-time flat fee of $200,000. CMU’s damages expert effectively disregarded this and other CMU licensing documents. But when properly considered, Marvell believes this evidence proves that even if any royalties are due to CMU (under the assumption Marvell used CMU technology), these royalties should be for a very nominal amount.
Second, CMU’s damages expert lacked the technical and industry expertise to reliably conduct the “price premium” analysis used to arrive at the 50 cent number. For example, there is no dispute that Marvell’s MNP feature was one of many improvements in Marvell’s accused chips – yet CMU’s damages expert had no basis, and in fact was not qualified to determine the value attributable to that functionality, as opposed to other improvements. Further, the 50 cent per chip was derived from information relating to only one historical data point, which was for the sale of a small quantity of sample chips sold to one of Marvell’s smallest customers. Indeed, a royalty rate of 50 cents per chip yields a royalty that as a percentage of the average sales price of a chip is far in excess (perhaps an order of magnitude greater or more) of typical industry rates for much more significant technologies.
Third, CMU’s expert also assumed that all of the success of Marvell’s products was due to the allegedly infringing MNP feature, and failed to account for the fact that more than 80 additional features were included in Marvell’s chips at the same time the MNP feature was added. This failure to properly apportion damages between the allegedly infringing feature and other non infringing features runs afoul of the law, including the recent decision by the Federal Circuit in Uniloc USA, Inc. v. Microsoft Corp., 632 F.3d 1292, 1318 (Fed. Cir. 2011).
Fourth, CMU improperly sought damages for alleged U.S. patent infringement based on Marvell’s worldwide sales, as discussed in more detail below. [January 7, 2013]
8. Why shouldn’t CMU obtain royalties attributable to products made, used, and sold outside the U.S.?
The U.S. patent laws, like the laws of other countries, are geographically limited in scope. As a result, use of a method outside the U.S. does not infringe a U.S. method patent. The vast majority of Marvell’s chips are sold overseas, and most of these chips never enter the United States. Based on legal precedent, we believe the law does not allow for damages against products which are exclusively made, used, and sold overseas, and which never enter the U.S. Such “extraterritorial” conduct is simply beyond the scope of U.S. patent laws and U.S. courts, and Marvell believes it was erroneous for CMU’s damages expert to assess damages against the entirety of Marvell’s overseas chip sales. Nonetheless, this fundamental legal error pervaded virtually every aspect of CMU’s damages expert’s analysis. The sales of non infringing chips overseas accounts for nearly 80 percent (or over $935 million) of CMU’s total damages figure. [January 7, 2013]
9. Will this lawsuit be expanded to Marvell customers?
CMU’s attorneys have not indicated any intention to pursue Marvell’s customers, and Marvell does not expect CMU to do so. Based on the Supreme Court's decision in the Quanta Computer, Inc. v. LG Electronics, Inc., 553 U.S. 617, 128 S.Ct. 2109, 170 L.Ed.2d 996 (2008) case, CMU cannot seek royalties from both Marvell and Marvell’s customers for the use of the same chips. [January 7, 2013]
10. What Marvell products were accused?
CMU's allegations only related to Marvell's read channel and SOC HDD chips that include the MNP or NLD features. No other Marvell chips or product lines were involved, including Marvell's other storage-related products, such as Marvell's SSD products. [January 28, 2013]
11. Why does Marvell believe that CMU failed to meet its burden on the objective prong of the willfulness analysis?
Although the jury found willful infringement, it is the Court’s ultimate decision whether Marvell acted despite an “objectively high likelihood that its actions constituted infringement of a valid patent.” In this case, Marvell believes that there was no infringement and that the patents are invalid. Marvell also believes that its actions were objectively reasonable for a number of reasons.
First, CMU’s inventor, Dr. Kavcic, explained in an October 2001 email to Seagate that he had not invented a Viterbi-like detector that accounted for data dependency in a “post-processor” (as do Marvell’s MNP chips). Rather, Dr. Kavcic believed his claims were limited to a modified Viterbi trellis. Dr. Kavcic even described Marvell’s patented MNP technology as “novel” in an article that he co-authored. Kavcic, Aleksandar, and Ara Patapoutian. “The read channel.” Proceedings of the IEEE 96, no. 11 (2008): 1761-1774. As the Federal Circuit recently explained, “[i]f the accused infringer’s position is susceptible to a reasonable conclusion of no infringement, the first prong of [the test for willfulness] cannot be met.” Uniloc USA, Inc. v. Microsoft Corp., 632 F.3d 1292, 1310 (Fed. Cir. 2011) (emphasis added).
Second, Marvell demonstrated a strong invalidity case. In fact, the Court characterized Marvell’s motion for summary judgment of invalidity as a “close call.”
Third, Marvell freely and voluntarily disclosed the CMU patents to the U.S. Patent Office when Marvell was pursuing patents on its own technology. In deciding to grant Marvell’s patents, the U.S. Patent Office implicitly agreed with Marvell that the Marvell technology was patent ably distinct from the CMU patents.
Finally, the fact that CMU did nothing for six years after its solicitation letters to the industry failed to yield a single license suggests CMU’s technology was not commercially feasible – not willfully infringed. Moreover, at no time prior to filing its lawsuit in March 2009 did CMU communicate that it believed Marvell was infringing on its patents. [January 7, 2013]
12. Why could no reasonable jury find Marvell had the required subjective intent for willful infringement?
As even CMU acknowledges, Marvell is a technology innovator. Marvell developed and patented its own solutions to the signal dependent noise problem, which it then incorporated into its chips. At trial, Marvell’s inventors explained that they did not copy CMU’s patents and detailed how they came up with their own independent solutions. In fact, CMU failed to produce evidence that Marvell copied any part of the asserted patent claims. [January 7, 2013]
13. Does the verdict affect Marvell’s current shipment of products?
There will be no impact on shipment of products unless and until the Court issues an injunction. If CMU files a post-trial motion for an injunction, under the Court’s current schedule, the issue would likely not be resolved before a hearing scheduled for May 1-2, 2013. Marvell has strong grounds to oppose any request for an injunction. Specifically, since CMU does not make any actual products that compete with Marvell’s products, CMU is not suffering any “irreparable harm,” which is a prerequisite for an injunction. [January 7, 2013]
14. Does Marvell anticipate that an injunction may be granted?
No. Marvell has strong grounds to oppose any request for an injunction. For example, money damages are adequate to compensate CMU for any alleged harm. Further, the lack of irreparable harm to CMU in the absence of an injunction supports Marvell. In addition, the balance of hardships favors Marvell, particularly since the parties are not competitors and there is no evidence that any other company has ever used the patented technology at issue in the case. [January 7, 2013]
15. What are the grounds for appeal?
Marvell intends to file a number of post-trial motions, including a renewed motion for a mistrial, a motion based on laches, and motions for judgment as a matter of law as to non-infringement, invalidity, non-willfulness, and as to the alleged damages. If Marvell’s post-trial motions are successful, an appeal to the U.S. Court of Appeals for the Federal Circuit may become unnecessary.
If necessary, however, Marvell has strong grounds for appeal. For example, CMU's substantive claims are barred by the doctrine of laches; under the correct claim construction, the asserted CMU patent claims are invalid and not infringed; and even if infringement were found, any damages should have been commensurate with the nominal license fees that CMU previously obtained from others, and further, damages should be limited to allegedly infringing use of Marvell chips in the United States. [January 7, 2013]
16. What are the immediate next steps in the case?
Marvell and CMU will each submit post-trial motions to the court, concerning a variety of contested issues relating to the trial. The court has set a briefing schedule for these motions, and a hearing on these motions is scheduled to be held on May 1-2, 2013. The court will likely issue its ruling on these motions some time after the hearing. At that time, if the district court’s ruling is adverse to Marvell, Marvell will file an appeal to the Federal Circuit in Washington, D.C. The Federal Circuit is a specialized appellate court that has jurisdiction over all appeals in patent cases. The timing of an ultimate ruling from the Federal Circuit may vary depending on such factors as the status of the court's docket and the time sensitivity of the issues that are being appealed. [January 7, 2013]
These FAQs contain forward-looking statements that involve risks and uncertainties, including statements regarding the complex nature of the patents at issue in the CMU litigation; Marvell’s non-infringement position; Marvell’s own patents; the CMU patents at issue as to invalidity and infringement; reasonableness of the assessed damages; findings of the CMU damages expert; CMU’s failure to meet the burden concerning the willfulness analysis and the jury’s finding Marvell willfully infringed; Marvell’s expectations concerning disruptions to its business or customers; matters related to the possibility of an injunction; statements about the nature and grounds for an appeal by Marvell; and statements about post-trial actions including motions and appeals processes. The forward-looking statements contained in this report are subject to risks and uncertainties, which may cause the actual outcomes or results to vary from those indicated by the forward-looking statements. These risks and uncertainties include any adverse outcomes of any motions or appeals against Marvell that might result in enforcement of the existing verdict unchanged or with enhancements that CMU may seek in post-trial motions and other risks and uncertainties, including those more fully described in Marvell’s latest Quarterly Report on Form 10-Q for the quarter ended October 27, 2012, and other factors detailed from time to time in Marvell’s filings with the SEC. Facts and circumstances referenced and asserted by Marvell are subject to change and Marvell undertakes no obligation to revise or update any of this information in respect of future events.
Originally Posted: January 7, 2013