Guest Post by Alan Cox: The Damages Testimony in VLSI Technologies v. Intel

Alan Cox is an economist with over thirty years’ experience testifying in Antitrust and Intellectual Property matters. He holds a Ph.D. in Economic Analysis and Policy from the Haas School of Business at the University of California at Berkeley.  He is an Affiliated Consultant at NERA Economic Consulting.  In this guest post he provides his observations of the damages testimony in VLSI Technologies v. Intel. As noted in the letter from Professor Chao that Dennis posted earlier, key filings in this case are not publicly accessible (nor are the transcripts). – Jason


VLSI Technologies sued Intel for the alleged infringement of several patents.  At trial VLSI asserted two patents: 7,523,373 (“the ‘373 Patent”), and 7,725,759 (“the ‘759 Patent”).  VLSI asserted that the technology described in these patents was incorporated into 987 million Intel microprocessors of various models.  On March 2, a jury in the Western District of Texas awarded VLSI $2.18 billion in damages for Intel’s pre-trial use of the patents.

The court allowed the public to listen to public portions of the trial by telephone.  This article describes aspects of the testimony and attorney statements on damages as logged in contemporaneous notes.

The damages testimony in this trial is interesting for several reasons, including Plaintiff’s presentation of a regression analysis as the basis of its damages claim.  Regression analysis is a sophisticated quantitative technique that is widely used in the academic literature, in making business decisions and in many areas of litigation, though it is seldom presented in patent trials.  The data necessary to undertake regression analysis is generally unavailable in patent matters.  As is often the case in the application of complicated statistical methods, the results can vary with inputs, underlying assumptions, data availability and other factors.  While Intel did address possible problems or partiality in the application of the methodology, the defendant’s core attack on the regression analysis was its repeated assertion that the method has never been undertaken in the real-world process of licensing patents.

By way of additional background, it appears that ownership of the patents had changed hands at least twice as the result either of mergers or as part of a sale of the patents themselves, including the sale to VLSI.  These transactions indicated a much lower value for the patents than was claimed by VLSI and awarded by the jury.  The patented technology had been originally assigned to Freescale, an integrated circuit manufacturer spun off from Motorola, later purchased by another semiconductor manufacturer, NXP.  Freescale was the owner of the patents at the time that Intel first allegedly infringed them.  Finally, as Intel pointed out at trial, VLSI is a non-practicing entity that undertakes no research nor production and has two employees.

VLSI’s Regression Analysis

A regression is a statistical analysis that is used to estimate the impact of several variables on another variable.  One use of the method, hedonic regression, measures the impact of product attributes on the price of that product.  For example, hedonic regressions have been used to determine how prices of automobiles vary with changes in horsepower, gas mileage, brand, and other attributes.  Hedonic regressions provide ratios that measure the amount by which prices will change with a change in one of these attributes.  For example, such a ratio may be the percentage by which the price of a car increases with a percentage increase in horsepower.

In this case, VLSI’s affirmative damages case was presented by a Ph.D. economist, Ryan Sullivan.  He applied regression analysis to estimate the impact of different attributes of Intel microprocessors on the price of those microprocessors.  Most important was his estimate of the response of price to changes in microprocessor speed.  His regression result indicated that a one percent increase in microprocessor speed increased the price of an Intel microprocessor by 0.764 percent, holding all other attributes of the microprocessor constant.

There was little discussion of the regression analysis at trial since VLSI clearly wanted to focus on demonstrating how the regression resulted in a large damage number and Intel did little to challenge the implementation of the methodology.  In cross examining Dr. Sullivan, Intel’s attorneys did, however, challenge VLSI’s expert about his inclusion of some potential explanatory variables and the exclusion of others.  Some of VLSI’s expert’s responses indicated that Intel may have had some effective points of attack of VLSI’s implementation of the regression methodology had it chosen to pursue them.

Calculating Damages from Regression Results

Next, VLSI’s damages expert used his regression results to calculate the total value that Intel purportedly received for the use of the two patents.  For this he relied on VLSI’s technical experts who had testified that the ‘373 patent decreased microprocessor power consumption by 5.45 percent and that the ‘759 patent increased microprocessor performance by 1.11 percent.  He also relied on a crucial assertion by VLSI’s technology experts that a one percent improvement in power consumption or a one percent increase in performance could be “valued as” a one percent increase in speed.  Using those asserted equivalencies in the value of microprocessor speed, power consumption and performance, VLSI’s damages expert calculated the value of the benefit of the patented technology by multiplying these percentage improvements in power consumption and performance by the percentage increases in price attributable to increases in speed.  Finally, he multiplied the resulting percentage increase in revenue by the total revenue that Intel had earned on the accused products.

This calculation can be restated concisely in summary mathematical form.  VLSI’s damages expert undertook the following calculations:

Intel’s Increase in Revenue due to its use of the ’373 Patent =

Price-Speed Ratio X % Reduced Power Use X $Accused Product Sales =

0.764 X 5.45% X $Accused Product Sales


Intel’s Increase in Revenue due to its use of the ’759 Patent =

Price-Speed Ratio X % Increased Performance X $Accused Product Sales =

0.764 X 1.11% X $Accused Product Sales

VLSI’s damages expert testified that there would be no increase in direct manufacturing costs from utilizing the patented technology and only a very small increase in selling and marketing costs.  Since there were no incremental costs to subtract from the additional revenue allegedly earned from using the patented technology, the increase in profits was equal to the claimed increase in revenues.

The Outcome of the Hypothetical Negotiation

The next step was to determine how Intel and VLSI would share these benefits in a hypothetical negotiation, the fictional event that takes place on the eve of first infringement during which the litigants are imagined to reach agreement on a royalty.  It is often customary to discuss the negotiating process over the sharing of benefits as being shaped by the relative bargaining power of the parties at the hypothetical negotiation.  VLSI’s damages expert instead undertook an unusual calculation.  He first calculated what he claimed was Intel’s contribution to the commercial success of its microprocessors.  He then subtracted the amount of Intel’s contribution from his estimate of Intel’s increase in profits due to its use of VLSI’s patents.

To implement this division, VLSI’s damages expert started by claiming that Intel’s contribution to its success in generating the patent-related additional profit could be measured as the sum of certain costs divided by its total revenue.  The reasoning that justified this approach was that these costs were incurred in the development and marketing of Intel chips.  More specifically, he took the sum of Intel’s Selling and Marketing Costs (S&M); its General and Administrative Costs (G&A); and its Research and Development Costs (R&D) and then divided that sum by Intel’s total revenue.  That is:

Intel’s Proportional Contribution = (S&M + G&A + R&D) / Intel’s Total Revenue

The result, 20.7 percent, was deemed to be Intel’s share of the estimated increase in profits that it earned from incorporating the patented technology into its microprocessors.  The remaining 79.3 percent was, therefore, attributable to VLSI.

The asserted outcome of the hypothetical negotiation was then simply the result of multiplying 79.3 percent by the increase in Intel profits, described above.  VLSI’s damages expert estimated the amount of money that a reasonable royalty would have generated before trial, a figure presumably close to the over $2 billion awarded by the jury.  VLSI’s damages expert claimed that Intel would have been willing to pay such a royalty because it faced strong competition and needed the competitive edge that the patented technology provided to maintain its market position.

The presentation of damages as a single payment for past infringement did create a potential problem for VLSI.  During the trial, Intel asserted that VLSI’s expert had not presented a per unit royalty in his pre-trial disclosure and should not be allowed to present a per unit royalty in court.  A ruling in Intel’s favor may have made future royalties unavailable.  As it happened the court ruled in VLSI’s favor on this point.

VLSI’s Anticipation of Defendant’s Critiques

In its direct testimony on damages, VLSI addressed several additional critiques that had been raised by Intel in pretrial filings, reports or deposition testimony.  First, VLSI’s damages expert was asked in his direct testimony about Intel’s claim that VLSI and its experts had not identified any comparable real-world licenses that resulted in payments as high as those that he was claiming to be reasonable; that is, in the hundreds of millions or billions of dollars.  They also asked him about the much lower patent values implied from the purchase of Freescale and the sale of the patents to VLSI, mentioned above.

VLSI’s damages expert responded that none of those real-world licenses and sales were analogous to the hypothetical license in this case.  He asserted that licensors in real-world licensing negotiations do not have comprehensive knowledge of the importance of the patents that are under negotiation.  Similarly, purchasers of the ‘373 and the ‘759 patents would not have known of the value of the patents to Intel or even the extent of the use of the patented technologies.  Any of the earlier transactions involving the patents would not, he claimed, have incorporated Intel’s expectations and experience of the benefits of using the technology.  VLSI frequently claimed through this and other witnesses that only Intel would have understood the value of the patented technology that it was exploiting.

VLSI’s damages expert characterized the hypothetical negotiation, by contrast, as one in which both side’s “cards were facing up.”   Consequently, the licensor, Freescale, would have been fully informed of everything that Intel knew about the usefulness and value of the licensor’s technology.  VLSI appeared to claim that the information that was available to Freescale in the up-facing cards included information about sales, profits and other outcomes after the date of the hypothetical negotiation.  VLSI’s damages expert indicated that the royalty rates found in Intel’s purported real-world comparable licenses were analogous to what Tom Brady was paid as a rookie before he had demonstrated his abilities as a quarterback in a professional football league.  According to VLSI, what was relevant at the time of the hypothetical negotiation was equivalent to what Tom Brady was paid during the peak of his career, not on what was known about him before his rookie year.  The use of the analogy appears to indicate that VLSI viewed the hypothetical negotiation as being informed by events and outcomes that took place after Intel’s earliest infringement, sometimes rationalized by reference to the “Book of Wisdom.”

He was also asked about the evidence that the technology was not used by any other company other than Intel, not even by Freescale and NXP, the semiconductor manufacturers who had previously owned the patents.  He described that point as merely a “red herring” since use by the patent owner was not the only determinant of the value of the patents.  As a basis for this point, he cited the Georgia-Pacific factors that call for consideration of the extent to which the infringer has made profitable use of the invention.

Damages-Related Testimony from Other VLSI Witnesses

Before the Plaintiff’s damages expert was called, a VLSI technical witness presented the results of a “patent scoring” exercise undertaken by Innography, an IP management firm.  According to Innography’s scoring, the two patents at trial were in the top 10 percent of patent value.  VLSI’s introduction of this result may contradict its key point that only Intel had sufficient information to have some understanding of the value of the technology since Innography could presumably only make such a claim from public data.

Defense Response on Damages

Intel had apparently provided expert reports on damages from at least three expert witnesses.  These included: a Professor of Operations, Information, and Decisions at Wharton with a Ph.D. in Management from MIT; a CPA with extensive experience as an IP damages expert; and an electrical engineer with 20 years of experience as a licensing executive with IBM.  VLSI moved to have the CPA’s “cumulative” opinions excluded as being redundant and “an improper attempt to bolster” the opinions of Intel’s other damages experts.  Whatever the court’s ruling on this and other motions to exclude, Intel called only the former licensing executive, Hanc Huston, to rebut VLSI’s damages expert.

Most of Intel’s damages testimony was not available on the public telephone feed because it involved confidential information primarily related to the comparable licenses that he had identified.  Intel’s damages expert brought up the point that the patented technology had not been used by any of the assignees of the patents and that VLSI had not identified anyone else who had used the technology.  He asserted that, in his experience, such lack of use of a patent was indicative of its low value and would have driven down the royalty paid on a licensed patent.  He also testified that he had never heard of hedonic regression being used in patent licensing negotiations.

He also disagreed with VLSI’s assertion that only Intel had the information necessary to comprehend the extremely high value of the patented technology.  He pointed out that the patent itself was public knowledge and that VLSI and others could have reverse engineered Intel to understand the extent that its technology was being used.

After discussing 18 licenses that he had determined were comparable, VLSI’s damages expert concluded that a reasonable royalty for Intel’s use of the technology described in the two patents totaled $2.2 million.

On cross-examination, Intel’s damages expert was asked about a policy at IBM to license patents for a one percent royalty per patent up to a maximum royalty of five percent.  VLSI pointed out, in cross and in closing, that a one percent per patent royalty was more than the amount that VLSI was asking in damages.  He was also asked about specific licenses and settlements between IBM that involved payments in the hundreds of millions of dollars.

The Testimony of VLSI’s Rebuttal Damages Expert

The final witness on damages was Mark Chandler, the president of another IP management firm, who was called to rebut Intel’s damages testimony.  He criticized Intel’s damages expert as relying on the “indirect evidence” of royalty rates from purportedly comparable licenses as opposed to the evidence of Intel’s success with the accused products.  He also repeated the claim made by VLSI’s affirmative damages expert that none of the previous transactions were indicative of the outcome of the hypothetical negotiation.  Those transactions, he asserted, could not have taken into account the value of the technology to Intel.  He further criticized Intel’s expert’s process for identifying ostensibly comparable licenses, indicating that his method was based on technical considerations without consideration of the commercial value of the licensed patents or any other economic considerations.

This expert also indicated that Intel’s damages expert had calculated per patent royalty rates by merely dividing the reported royalty in his purportedly comparable licenses by the number of patents being licensed.  Such an average, he testified, fails to take into account the range of values of the licensed patents.  He further testified that Intel’s expert had “cherry-picked” from among over 300 licenses that were part of the record in this case, ignoring licenses that resulted in payments of hundreds of millions of dollars with at least one being over a billion dollars.

This expert also claimed that Intel had ignored “direct evidence” of the value of the patents, the value of the use that Intel had made of the patents and the additional profits allegedly derived therefrom.  He repeated the claim that, unlike in the real world, in the hypothetical negotiation the parties would be “playing with their cards up” and have full knowledge of the value of the patents to each other.


The trial in VLSI Technologies provided not only a dramatic outcome but useful insights in the presentation and rebuttal of a damages case.  Interested observers can look forward to further motions, orders and appeals that should provide additional information about the technology and the appropriateness of the award.

Disclosure: I have no involvement with any of the parties or the law firms at this time. I have had no involvement in this litigation. I have never worked with VLSI. I have worked on at least one litigated matter with Wilmer Hale. I have consulted for Intel twice and testified for them once over my entire career but, to repeat, have no involvement with Intel at this time.

11 thoughts on “Guest Post by Alan Cox: The Damages Testimony in VLSI Technologies v. Intel

  1. 7

    BTW, I thought the main reason Intel was deprived of the usual defendant’s statutory right to a limited pre-trial PTO test of validity [if they have good enough prior art patents or publications to make that cut] was that they waited until ten months after they were sued to Petition for an IPR. But, probably not anyway, since one comment has noted that: “ In one recent analysis, 39 out of 41 denials based on Fintiv were on patents related to litigation in the Western or Eastern District of Texas. IPRs relating to cases filed in those two districts were more likely to be denied than instituted. In contrast, 95% of IPRs relating to cases filed anywhere else in the country were instituted.” [The ides of Texas are upon you?]

  2. 6

    Speaking of patent infringement damages [and Alice 101?], press reports today announce a jury verdict Friday in Marshall Texas [EDTX] holding that Apple must pay around $308.5 million to a “closely held” ‘local licensing firm” [guess what that is] for infringing a digital rights management patent asserted to be used in Apple products. Some reports refer to “running royalties.”

  3. 5

    Martin this is carried over from last thread, I’m at my cap apparently:

    ” That (housing costs high, goods cost low) may change back, in time, because freedom.”

    Like what in like 50 years? What about the people alive now that basically can’t start a life? And basically can’t be doing the whole making babies thing?

    Further I don’t think it’s going to “change back” anytime soon. As to housing the issues as to why its expensive are clear, zoning, NIMBY, land being limited in “in demand” locations and nearby, foreigners and speculators buying up of land/housing not to live in, and etc. etc. none of which are likely to change all that much. Though yes we may have another housing bust/crisis but it’ll go right back to sky high after.

    “Our findings suggest that inflation reduces the share of national income held by the top one percent.”

    Um yeah no duh, but it doesn’t actually affect the 1%’s life much at all other than then them having one less house or whatever among their 3x+ houses. The poors either lose their house, go bankrupt etc. etc.

    “In other words, it is getting more and more expensive to maintain a rich lifestyle, but not much more expensive to support a working-class lifestyle. ”

    Again probably true, but rich people not being able to maintain a rich people lifestyle doesn’t actually affect them all that much, they just become upper-middle class. Whooptie dooo. Here I’m talking about actual effects on your life. You go bankrupt, you lose house, you can’t go to college, etc. etc.

    “The assumption that wages won’t rise in an inflationary environment is based on what?”

    It’s not an herp derp “assumption”, it’s what has happened. Real wages have barely budged in over 30 years brosefulous. And the min wage, which plenty of the pop still makes hasn’t moved in like 20 years and barely moved the last time.

    “Why wouldn’t they rise?”

    There’s many reasons as to why they didn’t rise. Trade agreements (making americans compete against ROW workers), mass immigration, destruction of labor orgs, etc. etc. the list is long and mostly irrelevant because we’re discussing what already happened not some hypothetical that might happen in the future.

    “Housing, outside of certain markets,”

    Right but those “certain markets” are making the avg go through the roof, and a whole lot of people need to be in those areas for one reason or other (work/fam). Obviously land/housing prices in bum f nowhere didn’t rise that much. Same with costs of college except they are high increases almost across the board.

    “Do you want to address the proposition that the danger of deflation is much more fearsome ?”

    No, it is not the subject of discussion, and I don’t know why you’re so obsessed with it. It likely is the larger “muh danger” depending on how much deflation and how much inflation we’re comparing in terms of nums, but that is 100% irrelevant to this topic. And, just fyi, it supposedly isn’t the no. 1 reason why the fed reserve tries to keep the reported inflation where it is. The reason of course is supposedly to “promote near full employment” (aka keeping low unemployment). They could avoid deflation at 1% inflation or .5% just fine and dandy and easily, but they shoot for 2-2.5% for other reasons, mainly the one noted above. But that is also irrelevant to this discussion. This discussion is about whether the “real inflation rate” is in fact, through the roof and whether that is a primary thing keeping disadvantaged groups down.

    1. 5.1

      “It’s not an herp derp “assumption”, it’s what has happened. Real wages have barely budged in over 30 years brosefulous.”

      That’s another way of saying that the change in wages has closely matched inflation.

      1. 5.1.1

        “That’s another way of saying that the change in wages has closely matched inflation.”

        Not if we’re talking about the “true inflation rate” measure that is the topic of discussion derp. Instead the change in wages mehish matches, for avg workers (not the lowest workers who are at min wage etc), the broadcast inflation numbers which are much lower than the “true inflation rate” being discussed. That is to say the wages (note real wages) are increasing for the avg worker like 1-2-3% or so with the mainstream inflation nums at like 2.5-3%. However, the “true inflation rate” is said to be closer to 7-10% per year which means the real wages would be falling way behind.

  4. 4

    Fascinating — thanks. Looking forward to further case developments.

    A 1000x difference in damages estimates (2.2M v 2.18B) — startling.

  5. 3

    It does not seen like the damage calculation took into consideration the availability of non-infringing alternatives to reduce power consumption/increase speed…

  6. 2

    “His regression result indicated that a one percent increase in microprocessor speed increased the price of an Intel microprocessor by 0.764 percent, holding all other attributes of the microprocessor constant.”
    The price-speed ratio is o.oo764, not 0.764.

  7. 1

    Thank you both, for the article and for its prompt posting. This should make for an interesting appeal decision on infringement damages calculations if not decided on other issues.

    1. 1.1

      Re: “Intel did little to challenge the implementation of the methodology. In cross examining Dr. Sullivan, Intel’s attorneys did, however, challenge VLSI’s expert about his inclusion of some potential explanatory variables and the exclusion of others.”
      Is not that addressing the principal defect in regression analysis?

Comments are closed.