Last week, the Court of Appeal for England and Wales published its decision in Optis Cellular Tech. LLC v. Apple Retail UK Ltd., [2025] EWCA Civ 552. The court effectively overrules Mr. Justice Marcus Smith’s 2023 decision awarding Optis global FRAND royalties, over an eleven-year period, of $56.43 million (see my blog post here). The Court of Appeal ups the award to $502 million, on top of which there will an award of compound interest to be redetermined. The opinion on FRAND royalties is by Lord Justice Birss, joined by Lord Justices Arnold and Newey. Lord Justice Arnold writes an opinion, joined by the other two justices, holding that the lower court also erred when, in setting the other terms of the global FRAND license, it effectively ordered Optis to drop parallel U.S. litigation relating to five of Optis’ U.S. SEPs.
The decision comes to just 56 pages, which is relatively short for FRAND cases, but it is quite dense. The bulk of Lord Justice Birss’ opinion relates to calculation of the FRAND royalty, and as before many of the input numbers are redacted. I will not attempt to summarize the calculation issues in depth, but rather to give a more bird’s-eye view of what the relevant issues are, starting with a brief recap of the lower court decision.
Essentially, Mr. Justice Marcus Smith rejected the analysis provided by both sides’ experts, and went on to calculate the FRAND royalty himself. In doing so, however, he relied on Apple’s proffered comparables, despite some issues he had with these, in preference to Optis’, which at one point he described as “worse than useless.” Assuming an overall royalty stack of 22,000 SEPs, he derived for each relevant Apple license the lump-sum royalty that would be implied for Apple to acquire 100% of the patent stack. (For readers not familiar with the terminology, in para. 19 of the EWCA decision Lord Justice Birss defines the “patent stack” as “all the patents essential to the standard,” and the “royalty stack” as “the amount to be paid for licensing all” of the patents in the stack.) Mr. Justice Marcus Smith then calculated an adjusted average to account for past release and forward licensing, and divided this in two to arrive at an annual rate for the stack. Estimating Optis’ share of the stack as 0.38%, he came up with an annual lump-sum royalty for the Optis SEPs of $5.13 million.
Lord Justice Birss finds much to critique in this methodology. Let me start with some of his introductory comments, however, which define certain terms used throughout the opinion, and which may not be familiar to people coming at this for the first time:
12. . . . In some of the disclosed licences the licence fee was expressed as a per unit royalty. Two common ways of expressing a per unit royalty are as a percentage of the selling price of the unit or as a fixed sum per unit. In this case these two kinds are called ad valorem and DPU respectively. In other licences the agreement consisted of or included a lump sum payment. Both sides advanced cases which involved working out what useful information could be derived from the licences. For the licences in which the licensee had paid a lump sum that lump sum was “unpacked” into an estimate of the per unit royalty which that lump sum had, in effect, capitalised. This is an estimate and involves taking various factors into account, but as I say both sides took this approach. Three obvious factors which have to be taken into account when moving from lump sums to per unit royalties are some estimate of the number of units the parties had in mind, whether the agreement is a cross-licence and whether the per unit royalties are in ad valorem or DPU terms. Other factors include discounts for early receipt, discounts on rates e.g. bulk discounts, rates applicable to specific regions, discounts for the past (see next paragraph) and interest.
13. Another aspect of unpacking is the need to take into account the fact that some contracts not only operate to license future activity, they also operate as a release relating to unlicensed past acts. One approach was to unpack on a “free release” basis assuming the release was free (i.e. zero value) and attributing all the value to the future. Another method was the so-called “simple” basis mentioned already. This assumes the same rate applies in the past but may need to make an assumption about how far into the past the release applies. There are variants on these methods too, e.g. assuming the rate for the past is at a 50% discount. . . .
After summarizing how the lower court reached its FRAND royalty (paras. 55-75), Lord Justice Birss begins addressing the various grounds for appeal, starting with the lower court’s rejection of the experts’ opinions (paras. 86-97). On this issue, the lower court applied too strict a standard:
88. It was not put to either expert, either in cross-examination nor by the judge, that they had strayed outside their zone of expertise ([312]) or had provided no independent judgment ([315]). For that reason alone the judge’s conclusions to that effect were unfair and unjustified. . . .
89. I believe the judge may have been distracted by the way the expertise of both was characterised when they were described below as “valuation” experts. That could be taken to imply that the whole matter of valuation was within the scope of their expertise. That never was so and I have referred to them as accountancy experts for that reason. . . . Their skill is in grappling with the financial details of the kinds of licences in issue in order to extract what in their expert opinion will be useful information for the court. This is the task of unpacking. Neither expressed a view on whether ad valorem or DPU was the appropriate FRAND royalty rate and since they were not asked why not, we do not know. However I suspect a simple explanation is that neither expert thought that the choice was a matter for them or within their expertise. In my judgment either approach to a rate is manifestly capable of being FRAND . . .
90. The other dimension to this issue is a distinction between the comparability of a licence and the reliability of the evidence arising from it. On a number of occasions the judge referred to the task of unpacking, which the experts undertook, as being to render licences comparable. Of course in a very general sense that is true, but this case has drawn attention to a useful distinction which as far as I am aware has not been identified with clarity before. Comparability in the sense of a licence being a good parallel, as referred to by Lloyd LJ in Cimetidine, is primarily concerned with the situation of the parties and the subject matter being licensed. So in a case like this about a FRAND licence for Apple handsets, a licence by the SEP holder of the same portfolio albeit to a different handset manufacturer is likely to be the place to start to identify the best comparable. The extent to which licences are affected by hold up or hold out will also have an impact on their comparability. However this kind of comparability has nothing to do with whether the licence terms involve lump sums or ad valorem or DPU royalties. Thus it can be seen that the accountancy experts are not best placed to express a view about this kind of comparability. Comparability itself will be a matter for the judge to decide based on evaluation of all the evidence, which may well include fact evidence about the situations of the various parties and so on. Unpacking does not alter that sort of comparability at all.
91. On the other hand reliability is concerned with the quality of the information derived from a given licence, perhaps involving unpacking in various respects. With this distinction in mind, unpacking is not concerned with making a licence more or less comparable in the sense I have described it, it is about improving the reliability of the comparison.
92. As an approach, the “comparables approach” is the one described in Cimetidine. Therefore the fact either expert adopted a comparables approach on instructions is no criticism. It is an appropriate approach in law. As accountancy experts they might well need to be instructed as to which licences they should consider. They were unlikely to be able to comment on comparability of individual licences as I have defined it, but they could and did comment on the quality of the data which can be gleaned from the licences – in other words they commented on reliability and were entitled to do so. . . .
95. Before leaving this topic I will say a bit more about starting from the SEP holder’s own licences. I maintain that this is the place to start but that is all it is. Factors like hold out and hold up may well render licences of the same portfolio less good as comparables. In this case there are also licences to the putative licensee (Apple). Such licences are capable of being useful comparables, again subject to hold up and hold out, but using them also involves a further dimension which is why, although they may well be useful in the end, they are not the best place to start. Their comparability (not reliability) also depends on the relationship between the patent portfolio being licensed and the SEP holder’s portfolio. Not only does one need a view about stack shares, the issue of portfolio quality arises. It is not enough to render them comparable to say that the SEP holder’s portfolio is average. The other licensed portfolios also have to examined.
Proceeding then to the lower court’s calculation (paras. 98-123), the court states that the “essence” of that court’s approach can be found in table 13 of the lower court judgment, in which “19 licences are placed on a common scale normalised by patent stack share. The scale used is the implied total lump sum stack price. The method then takes a simple average of these values to produce a single figure and that figure is then adjusted to try to take into account the past releases in the licences. The stack share of Optis is then applied to the adjusted figure to render an annual sum applicable to Optis” (para. 101). The court perceives various problems with the way in which the lower court used and/or adjusted these data, however, in particular its decision “to take a simple average of the entries in the table. This had no precedent or basis in the evidence . . . nor can it be justified in principle” (para. 106). Elaborating on this point, “the judge’s approach is an example of exactly the technique which Lloyd LJ warned against in Cimetidine because it involves using less good comparables to modify a result from better ones. . . . The right approach was to adopt a comparables based approach . . . in the sense of being one based on identifying the best comparable or comparables, excluding others and working from there” (paras. 114-15). Moving on, “[s]ince the correct way forward is to identify the best comparable(s) the next question to address is Optis’s appeal against the judge’s rejection of its case that the Apple licences were tainted with hold out and not FRAND” (para. 116). On this issue, the court concludes that the licenses were so tainted: “When it can do so, Apple’s significant negotiating strength leads some parties to agree lower rates than would be agreed between a willing licensor/willing licensee. There is a degree of hold out involved. . . . Therefore on this ground, and expressed in deliberately qualitative terms, in my judgment the judge was wrong to place weight on the values derived from the Apple licences as a whole. There is hold out involved, particularly as an explanation for the spread of values of these licences whenever they are put on a common scale by stack share” (paras. 122-23).
The question, then, is how to proceed in view of these errors, and the court concludes that the better option is for it to decide the FRAND royalty based on the evidence in the record, rather than to remit the case for a retrial. In this regard, the court takes note of (1) a license between Optis and Google, and (2) 13 licenses between Apple and other companies, including at the high end Ericsson, InterDigital, Nokia, and Sisvel. The court concludes as follows:
137. The first conclusion I draw from this data relates purely to the Apple licence information. Putting these four Apple licences together shows that a DPU rate for Optis lower than rates implied by the Ericsson/InterDigital/Nokia/Sisvel licences would be too low. The judge’s conclusion expressed as a DPU is of the order of $0.02 or $0.01. There is no justification I can see for a DPU rate that low when those four companies are able to license Apple at an appreciably higher rate.
138. The second conclusion I draw involves taking the Google licence into account. The simple conclusion is that the Google licence as a comparable shows that deriving a rate from Apple licences on their own – either as a whole or just focussing on the four I have identified - would produce a rate which was too low. . . .
139. The third conclusion is the converse of the second. Putting rates from the four higher Apple licences beside Google strongly suggests that simply using the Google rate unmodified would be too high. In other words the FRAND answer is between the two, but where?
140. Before going further, another indication that the Google rate relied on by Optis is too high comes from a top-down cross-check. As mentioned above, the judge rejected a 15% ad valorem stack price as too high . . . whereas . . . ad valorem stack totals of [XG]% (S) or [XH]% (FR) from Apple’s comparables were “at least consistent with a sensible business model” and as I observed above, since those data come from the Apple comparable licences it is hard to see how they could not be sustainable at least for Apple.
141. Using Google’s ASP of $470 . . . one can express a 15% stack price total in DPU and then use that to gauge the impact of a rate based on the Optis stack share. . . . A 0.38% share of $70.5 would be $0.27 (not $[XB]). Conversely starting with a rate of $[XB] DPU, based on a 0.38% stack share, that implies a total stack price in DPU of $[XAA] (=$[XB]x(100/0.38)), which is obviously too much for a device with that $470 ASP.
142. This is clear evidence that the $[XB] rate derived directly from Mr Bezant’s unpacking of Google is far too high and even dropping the rate to $0.27 per unit would still be much too high because that would imply an ad valorem overall stack price for a Google priced phone was 15% (the level the judge rejected). This approach also demonstrates that in a case like the present one this sort of top-down reasoning has utility.
143. The question therefore is where to go down from $0.27 DPU. There are two considerations. The first is that if and to the extent there could be a range of FRAND rates, then the patentee is entitled to the top of that range. The second is that the nature of the evidence here does not justify fine distinctions. The realistic options are a DPU of $0.20, $0.15 or $0.10. The last one ($0.10) is clearly too low because it is too far from the Google licence and [XAB].
144. I wondered about a DPU of $0.20 per unit for Optis. The grossed up stack price this would imply would be just over $50 ($0.20x(100/0.38)=$52.63). To gauge its significance I will compare that implied total stack price not only with the Google ASP of $470 per unit but also a figure of $625, which I take as representative of Apple’s ASP in the earlier part of the relevant period . . . . A total stack price of $50 would be 8% of the price of this earlier relevant representative Apple ASP and 10.6% of the price of a $470 phone. That shows that $0.20 per unit is still too much.
145. In my judgment $0.15 per Apple unit for Optis is FRAND. I reach that conclusion based on using Google and also Ericsson/InterDigital/Nokia/Sisvel as the best comparables, recognising that they are not similarly situated and that there would appear to be degrees of hold up and hold out involved. As a cross-check, grossed up it would imply a total DPU stack of just under $40 ($0.15x(100/0.38)=$39.47) which would make the total stack 6.3% of the earlier relevant representative Apple ASP of $625 (and 3.9% of a more current Apple ASP of $1000). It would be 8.4% of a $470 Google ASP phone. . . .
147. As an indication of the sort of lump sum one reaches from this DPU for the purpose of comparison with the judgment below, at sales of 220m pa that would be an annual capital sum of $33m. For eleven years on the judge’s basis that would be $363m.
Having earlier noted, however, that the judge erred in limiting the award to eleven years—five future years and only six past years, as opposed to all past years of infringement (see para. 80, citing InterDigital v. Lenovo, [2024] EWCA Civ 743)—the court reaches a final decision as follows:
148. In its submissions Optis provided a spreadsheet of DPU data which, amongst other things, will convert a DPU into a lump sum on one of two bases. The total sales volume for the period is produced by adding together past sales and projected future sales from the perspective of 2021 up to 2027 (albeit the licence was settled on the footing it would come into force in January 2023). The total is multiplied by the DPU to produce a lump sum. Thus in the spreadsheet a DPU of $0.15 produces a lump sum of either $516m on one basis or $502m on the other basis. The difference between the two bases is that the first uses a total sales volume of 3,440m units and the second uses 3,347m units. I take it the figure for the first basis is derived by taking all relevant actual sales from 2021 back into the past on the basis appeal ground 10 is decided in favour of Optis – as Apple has conceded at this level. For the sales for 2021 to 2027 inclusive I believe this total simply takes the 2020 figure and multiplies by 7. The second basis works in the same way save that it employs a 10% discount on the projected future sales for the 2021-2027 period. I would use the 10% discounted approach which, assuming I have understood the spreadsheet correctly and assuming the approach based on working from 2021 is common ground, means the lump sum (without interest) for a DPU of $0.15 should be $502m.
Following Lord Justice Birss’ opinion, Lord Justice Arnold’s opinion takes up a couple of other terms of the FRAND license the lower court crafted. First, Lord Justice Arnold agrees with Optis that “Apple should be required to pay interest until the date Optis receives payment,” and not just until January 1, 2023, as the lower court held (paras. 164-67). Second, he concludes that the lower court erred in ordering Optis “not to proceed with or take any steps in relation to litigation and/or not to enforce any order (save as regards any order as to costs made against apple by a competent court in any jurisdiction) in relation to the Portfolio in any jurisdiction in respect of all products, services and components manufactured or supplied, or to be manufactured or suppled, by or for Apple” (para. 222). As it happened, Optis filed its complaint in the U.K. on the same day it filed a complaint in the E.D. Texas (adjusted for time zone differences). Because Apple did not agree to be bound by the terms of a global license as determined by the U.K. courts until September 2023, the U.S. litigation proceeded apace—resulting in an initial judgment in the amount of $506 million for the infringement of five U.S. SEPs, then following the district’s order for a new trial, a second judgment in the amount of $300 million. The U.S. matter is now on appeal to the Federal Circuit. From Lord Justice Arnold’s opinion:
243. Optis’ position is that it accepts (and has always accepted) that it is not entitled to double recovery in respect of the five US Patents asserted in the EDTX proceedings. Optis nevertheless contends that the judge should have made appropriate provision in the FRAND licence to give effect and value to the final US judgment, rather than set it at naught, since that is what a willing licensor in the position of Optis and a willing licensee in the position of Apple would agree.
244. As to how appropriate provision should be made, Optis says that, in theory, the right approach would be for the consideration for the licence determined by the English courts to be adjusted so as to replace the value assessed by the English courts for the US Patents and acts covered by the final US judgment with the US judgment sum. Optis accepts, however, that in practice the valuation methodologies employed by both sides do not permit this to be done. Next, Optis says that a pragmatic solution would be to subtract from the English assessment a sum reflecting the value of the five patent families worldwide. This would be favourable to Apple, and hence FRAND. I do not see, however, that this is any more practical than the first solution, for the same reason. Finally, Optis says that an alternative pragmatic solution would be to treat the final US judgment sum as a floor i.e. Optis retains whatever sum is awarded in the US and, to the extent that the total global royalty payment held to be FRAND in these proceedings exceeds the final US judgment sum, Apple should pay Optis the balance. This would be even more favourable to Apple, and hence (if possible) even more FRAND, than the second solution.
Apple, of course, opposes any of the above, arguing inter alia that the above “arguments are inconsistent with [Optis’] position at trial, which was that the judge determine FRAND terms for a worldwide licence of Optis’ portfolio” (para. 247). Lord Justice Arnold disagrees, stating that Optis reserved its position on the Texas judgment, and that is only when Apple agreed to accept the outcome of the English proceedings in September 2023 “that the potential for inconsistency between the EDTX proceedings and the English proceedings crystallized” (para. 248). Moreover, “for the English courts to make an order requiring a regularly-obtained US judgment to be vacated seems . . . manifestly to give rise to an issue of comity” (para. 256). He concludes:
. . . the decisive factor in the present case is that Apple was responsible for the EDTX proceedings having proceeded to judgment (and pending appeals) prior to their change of stance on 15 September 2023. In my judgment a willing licensee in the position of Apple would therefore recognise that it would not be fair or reasonable for the licence determined by the English courts to require Optis to consent to the EDTX judgment (or any judgment of the CAFC on appeal) being set aside. Still less would it be fair or reasonable for it to require Optis to do so without even being reimbursed for the costs of the US proceedings down to 15 September 2023, which is Apple’s position. In principle, the right answer would be for the English courts’ valuation to be adjusted to take account of the EDTX judgment, but as discussed above the valuation methodologies adopted by the parties make that impractical. In those circumstances, the least-worst solution to the problem which Apple has caused is for the US final judgment to be treated as a floor for the royalties payable by Apple under the licence determined by the English courts in the manner proposed by Optis (para. 257).
Since the most Optis is likely to achieve at the conclusion of the Federal Circuit appeal would be a reinstatement of the first jury verdict from the E.D. Tex., in the amount of $506 million, not including prejudgment interest, it seems unlikely to me that the resolution of this issue ultimately matters all that much in the instant case, but its affirmation of the principle of comity is important.
Among the issues in the lower court decision that were not pursued on appeal are whether the SSPPU is the appropriate royalty base, and whether the SEP owner is entitled to any of the value of the standard. Mr. Justice Marcus Smith thought not on both issues, but neither of these are addressed here.
No comments:
Post a Comment