Seeking to preserve the huge payout is Carnegie Mellon University (CMU), which sued Marvell in 2009. Lawyers for CMU say the technology described in two CMU patents, which relate to chips that reduce "media noise" from hard disk drives, did nothing less than save Marvell's business. The jury's original verdict of $1.17 billion, later enhanced for willfulness and interest, is based on a royalty of 50¢ per chip. CMU says that's a reasonable rate given the more than $5 billion in profit that Marvell earned on those chips.
A panel of three Federal Circuit judges heard oral arguments in the case earlier this week. In their appeal brief (PDF), Marvell lawyers attack CMU's two patents, numbered 6,201,839 and 6,438,180, as market failures, "incapable of commercial implementation." In 2005, the company offered to license one of the two patents to Intel as part of a larger group for $200,000, but Intel declined (and was not sued). Marvell, though, was slapped with a running royalty of 50¢ per chip on the more than two billion chips it sold worldwide over nearly a decade.
Whether or not Marvell's worldwide sales can be brought to bear on a US patent case may well end up being the most important issue in this case. It's also the issue that has led a group of large technology companies, including Broadcom, Dell, Google, HP, and Microsoft, to support Marvell (PDF) in its appeal. Arguing the other side of the issue, six universities have filed an amicus brief (PDF) supporting CMU.
US patent, worldwide sales
Marvell argues that a per-unit royalty is a ridiculous calculation in the first place, cooked up by CMU's damages expert Catherine Lawton, a "career litigation consultant with no background in economics, accounting, microchips, or intellectual-property licensing." The two patents in this case were licensed to IBM, Seagate, and 3M for payments of $250,000—flat "membership" fees, which included licenses to all CMU patents conceived during their membership terms. Lawton got to her 50¢ royalty by relying on a faulty "excess profits" analysis that overvalues the patented feature, Marvell argues.
At trial, Lawton explained CMU's six-figure deals as being special membership deals that predated the "hypothetical negotiation," which determines patent damages. CMU argued that those agreements were "radically different," and the judge ultimately agreed. The jury clearly found Lawton convincing as well, awarding the full 50¢ per chip royalty she suggested.
Marvell argues that applying that royalty to foreign chips amounts to an "extraterritorial" application of US patent law, with dramatic policy implications—a point that also concerns the tech companies that are supporting it with friend-of-the-court briefs.
CMU says all the infringing activity actually took place in the US, and there are no damages being awarded for "foreign sales." Marvell sells its chips to hard-drive manufacturers, who then install them in the hard drives, most of which are sold outside the US.
The way CMU sees it, the "relevant sales" are the "design wins" in which manufacturers of hard drives chose to use Marvell chips, not the actual hard drives that were sold to customers. Those "design wins" all took place in California, "where Marvell and its customers were so intensely collaborating."
The jury used worldwide sales as a way to value the domestic use of CMU's patents, which was "the only sensible way to calculate damages" in this case, CMU argues.
CMU: Marvell bet on a “coffee warmer” chip, then copied
Marvell's and CMU's views of the invention at issue remain far apart. Throughout the 1990s, many engineers were working on the problem of how to resolve "media noise" that causes errors in reading signals from magnetic storage, such as hard drives. An early solution was called a "Viterbi detector," named after Qualcomm co-founder Andrew Viterbi.
Marvell characterizes the developments by a CMU grad student, Aleksandar Kavcic, and his professor, Dr. Jose Moura, as being admittedly "embryonic" and needing "substantial work to bring to market."
The company also says the patents filed by Kavcic and Moura are invalidated by an earlier filing by Seagate engineer Glen Worstell. When Worstell was shown the Kavcic/Moura work in 1997, he e-mailed CMU, saying he'd done work on similar problems. Problematically for Marvell, he added that "this invention is related but goes beyond my work and is probably more interesting."
Marvell notes that its engineer Greg Burd said in 2001 that Marvell "can not implement" the Kavcic detector, because it's "too large." The Kavcic solution was "optimal" but theoretical, requiring too much computing power to be practicable. Instead, it created a different solution, a "media-noise post processor." However, its engineers named its detector "KavcicPP" (Kavcic post processor) "in homage to Kavcic and his optimal detector." Later, the company changed the name to "media-noise post processor," or MNP.
CMU lawyers deny that Kavcic's invention was too complicated to implement, saying that it was "simply years ahead of its time." At the time of invention, "no one had packed data so densely that the invention was needed to read it." But storage capacities were doubling every year or two, and overcoming the "signal to noise ratio" became a business imperative.
The real story, says CMU, is that Marvell made a "near-fatal bet" on "iterative decoding" and fell behind. Its chip "consumed so much power that it melted," write CMU lawyers. "Gallows humor set in. Marvell executives dubbed it the 'coffee warmer'... Marvell's solution: Copy CMU."
The Marvell circuits look like a "cut and paste" of a figure from one of the patents. CMU, too, quotes Marvell engineer Greg Burd, but to different effect: CMU lawyers say Burd "warned Marvell executives twice" that Kavcic's noise reduction scheme was patented. "But Marvell had to use CMU's invention or risk going out of business," write CMU lawyers. "So it ignored Burd's warnings."
In a stinging opening to its response brief, CMU writes:
Marvell protests the sheer magnitude of the judgment, wondering, “How did this happen?” Simple. A billion-dollar verdict is what happens when your constant exploitation of a patented method trillions of times a day with millions of chips over the course of a decade saves your business from certain ruin, propels you to market leadership, and, in the process, generates over $10.3 billion in revenue and over $5 billion in operating profit.
Marvell says CMU tells a "colorful story" about a "heroic innovation" while neglecting to mention that its "supposedly revolutionary technology never resulted in a single sale to a single customer, was never commercialized, and was never licensed except as part of a research-center membership package."
Whatever happens, the outcome of this appeal will be closely watched by both universities, which have become increasingly aggressive about asserting their patents in recent years, and technology companies, which fear that universities with patents could lead to the kind of massive verdicts they've managed to (mostly) stave off in the case of "patent trolls."
Because of the parties involved and the size of the case, the decision in CMU v. Marvell will have some political fallout as well. University groups like the American Association of Universities played an important role in last year's debate over patent reform and have made their continued opposition clear. Some university officials believe that the Innovation Act goes too far and that in seeking to harm trolls, it will chill legitimate patent enforcement efforts. University opposition provides important political cover for politicians wanting to cast a "no" vote, since educational institutions are much more popular than other powerful groups that have opposed patent reform, which include pharmaceutical companies and lawyers' groups.
No comments:
Post a Comment