What will TV cost you? Putting a price on HEVC licences

The compression algorithm is a piece of critical technology that enables users to view videos on a phone or a computer and now there is a new one coming our way: HEVC.

How important is the commercial adoption of this new compression technology? Without compression, the movie Thor: Ragnarok – which is 130 minutes long – would be 11.6 TB. With current compression technology that same movie is about 27 GB; with HEVC it is approximately 14 GB (see Figure 1). Thus, HEVC can help consumers to save limited mobile data and businesses to cut costs on data storage and transmission. All while delivering equal – or better – quality video.

Figure 1. Video compression reduces data requirements by 1,000x

However, there is no point pretending that compression technology is easy. The math makes your head spin, the trade-offs are tortuous and the metrics to decide what looks good – whatever that means – will make you question why you ever asked. The implications for chip designers are staggering. Hundreds of billions of dollars of semiconductors are produced with specially designed circuitry and instructions for optimising video compression and decompression. Those semiconductors are built into billions of devices every year. How much circuitry are we talking? We estimate that for 2018, if you divide up all the transistors manufactured for video compression, there would be 4,000 transistors per ant. (There are about 5 quadrillion ants in the world.)

Broad industry adoption of HEVC was kickstarted by Apple’s July 2017 announcement that its iOS 11 would natively support HEVC. Apple’s membership of AOMedia became public as of the time of writing. It is too early to tell whether this membership will cause Apple, and others, to shift away from HEVC adoption.

There are multiple reasons for slow adoption but a complex and expensive patent licensing landscape with three major licensing groups may be one. Compared to a peak price of $0.20 per handset for an AVC patent pool licence, a consumer electronics manufacturer planning to make a handset that supports HEVC would be facing an estimated $1.60 per handset charge to license HEVC from the three pools. There would also be additional royalties for owners of non-pooled patents, which we estimate would bring the bill to $2.25.

One possible reason for the proliferation of licensing groups is that historically, licensing patents around audio/video compression has generated billions of dollars in revenue. Further, the patent battles are slated to continue with the latest HEVC standard. If your company plans to support HEVC, this will be complicated. Solvable but complicated.

This article aims to provide an understanding of the history of HEVC, video compression standards and the associated patent licensing landscape. Given the complexity of this subject, it focuses on providing a starting point to guide companies through some of the relevant patent licence issues. We are not playing favourites among the pools nor are we criticising any one pool or its policies. Rather we have focused on the perspective of HEVC adopters – the customers of the pools. How will they view the pool’s stated rates and policies? With that, we will look at how the pools, their pricing and the licences might affect adopters’ profits and costs.

Brief history of video compression

While HEVC is the sixth major ITU standard for video compression, it is also the third video compression standard jointly worked on with the MPEG (operating under the ISO and the IEC). Table 1 provides a brief overview of key video compression standards from the ITU and MPEG.

Each of these standards has targeted delivery video at lower bandwidth requirements, generally at significantly higher quality. MPEG-2 was notable for its adoption as the standard format for digital TV broadcasting and in DVDs. HEVC has now been adopted for the next generation of digital TV broadcasting (ATSC 3.0 in the United States). Table 2 highlights several technical improvements between most of the successive video standards discussed in Table 1.

Each of the standards builds heavily on those that came before. Thus, the 2013 HEVC standard does not stand alone; rather, many of its fundamental concepts relate to the approaches selected for H.261 back in 1988 – which provides its own set of patent licence challenges. For example, HEVC builds on the macroblock concepts that date back to the 1988 H.261, while adding new, more refined capabilities for segmenting those macroblocks. For those interested in a more in-depth technical analysis of the standards, the presentations and papers by Gary Sullivan are a good starting point (see Google Scholar: https://goo.gl/QrNzhA).

One further point: standardisation is critical to the technology industry and the video encoding space in particular. By standardising the video encoding stream, more devices can interoperate, which leads to the promised value highlighted by Intel’s former CEO, Craig Barrett: “[w]hen you have common interfaces, common protocols, then everyone can innovate and everyone can interoperate. Companies can build their businesses, consumers can expand their choices, the technology moves forward faster, and users get more benefit” (emphasis added). Contrast this with the problem of incompatible technologies where hardware for one format (eg, Betamax or HD-DVD) could not work with another (eg, VHS or Blu-ray). It is expensive for content providers and others in the ecosystem to support multiple formats. It is generally preferable to have fewer divergent standards, as well as higher-quality standards.

We will provide a brief history of video patent pools before turning to the compression capabilities of each of these video compression standards, as well as known licensing fees of established pools, in the context of the ever-changing consumer electronics landscape.

Video patent pools: history

While HEVC is the sixth major standard from the ITU, it is the third major video coding standard to have a patent pool associated with it. The first – MPEG LA’s MPEG-2 pool – was notable for being widely adopted (it was used in digital TV – including the ATSC standards in the United States – and also used in DVDs). MPEG LA provided a one-stop shop for clearing the overwhelming majority of patent rights for use of MPEG-2. For decoding hardware products (eg, a TV, DVD player or computer), the rates were:

• $4.00 (inception in approximately 1997 to 2002);
• $2.50 (2002 to 2010);
• $2.00 (2010 to 2015); and
• $0.50 or $0.35 (2015 onward).

The rates and pricing remain controversial. For example, in August 2017, Haier – a large consumer electronics maker and an ATSC and MPEG-2 licensee of MPEG LA – filed suit over the rates (see Haier America Trading LLC v Samsung, Case 1:17-cv-921, NY Northern District, August 21 2017 – the suit covers both the ATSC and MPEG-2 patent pools operated by MPEG LA). The Haier suit raises questions about the (lack of) effectiveness of the screening process for including patents, inclusion of non-essential patents, the pricing model (flat fee versus scaled by device cost, especially in the face of declining device costs) and antitrust concerns.

One of the complaints in Haier concerns the addition of patents to the pools over time. This will not surprise astute observers of the video standards world. NTSC encoding for analogue TV signals in the United States was first promulgated in around 1941 (for black and white TVs) and then modified in around 1953 to add colour. Nonetheless, patent licensing for improvements on NTSC had a much longer life than 20 years, even given the addition of colour.

Nonetheless, the fact that there are live patents after more than 20 years should raise questions that require answers from patent owners. Thus, even MPEG-2 – which is now over 20 years old – was not completely unencumbered by patents in 2017. MPEG LA continues to run a licensing pool for MPEG-2 that included seven unexpired patents as of July 2017 and charged $0.50 per device for a patent licence (notably, the remaining patents have early priority dates in this case but greater than their 20-year lives due to country-specific laws in the United States, Malaysia and the Philippines). A lower-priced $0.35 tier is also available for companies that waive early termination rights.

The issue of listing patents with priorities post-dating the standard can be at least partially answered: standards are not fixed in time. Even the MPEG-2 standard has been amended since its 1995 release, with the most recent amendment dating from 2012. Also, the vast majority of patents historically in the MPEG-2 pool had priority dates falling roughly between 1990 and 1995. Thus, any late patents may actually represent innovations related to amendments to the standard.

Turning away from MPEG-2, we arrive at the second widely adopted video coding standard: AVC.

MPEG LA also served as the administrator for the AVC patent pool. The rate is $0.20 per device but there are some volume-based pricing tiers, as well as a cap on total licence fees. Compared to MPEG-2, the AVC patent pool was more affordable. (See “Other issues” boxout for a discussion on whether the AVC pool rate was set artificially low.)

Bear in mind that if your device supports multiple standards, you will need to license each one separately. So for a device that supports MPEG-2, AVC and HEVC together, you would need to pay the fees for each separately to the appropriate administrator(s) or independent patent owners.

Before exploring the HEVC patent licensing landscape in more detail, the backdrop of the changing consumer electronics landscape merits investigation.

Figure 2. Computer prices 2002-2015

Source: US Bureau of Labour Statistics: ”Long-term price trends for computers, TVs, and related items” (The Economics Daily, October 13 2015)

Consumer electronics: pricing and performance under Moore’s Law

Discussing video compression standards without reference to the changing consumer electronics landscape can be challenging. Consumer electronics, particularly computers, exist against the backdrop of Moore’s Law which, loosely speaking, predicts that computing capabilities double in performance every two years.

Figure 3. Computer performance 2002-2015

Source: Millions of instructions per second (MIPS) data from Intel, AMD and Wikipedia

Figure 2 shows trends for computer prices from 2002 to 2015. The downward pressure on personal computer prices – even in the face of inflation – is extreme, with the price of a computer dropping from $1,000 to $277 from AVC’s launch in 2003 to HEVC’s launch in 2013. Significantly, a $277 computer bought in 2013 is significantly more powerful than its more expensive 2003 predecessor. Figure 3 shows the processor performance over the period from 2002 to 2015 in millions of instructions per second (MIPS). Around the time that AVC was launched, processors were clocking it at about 10 MIPS, but by the time of HEVC’s launch 130,000 MIPS processors were readily available. Thus, over the same 10-year period where the computer dropped in price by nearly one-quarter, the processing power available went up by a factor of 10,000.

Figures 2 and 3 together translate the dilemma for consumer electronics manufacturers operating in the environment of Moore’s Law into economic terms: deliver laptops, mobile phones and tablets with twice the computing power for lower prices year after year after year. While we do not show the graphs for this, the storage capacity of computers and network bandwidth across the network has been following a similar curve.

It is worth contrasting this with expectations for higher-resolution video over that same period (eg, from 1080p in 2003 to, say, 8K presently). The 1080p video would be 1,920 x 1,080 pixels, while the 8K video is 7,680 x 4,320 pixels. So that is a factor of 16 times more pixels per videoframe; in fact it is slightly higher due to increased bit depths.

Against this backdrop, what does the patent and licensing landscape for HEVC look like?

HEVC licensing demands

The known public demands (and estimates) for licence fees for AVC and HEVC are summarised in Table 3. We know that some of the public data is inaccurate or out of date (see “Moving target analysis” boxout). Nonetheless, for this analysis we will use the public data as is (data retrieval dates are noted), even though that may skew the estimates slightly. Relatedly, while Velos’s data is not publicly available, we have estimated its holdings and royalties, as discussed further below. Also, given the relative newness of HEVC as a standard, there are still pending patent applications around the world that are not yet reflected in the pool’s lists.

We analyse the rates through several different lenses in Table 3 to provide context on how the different rates, caps and patent holdings interact with the pricing. Thus, compared to a peak price of $0.20 per handset for the sole AVC pool, a consumer electronics manufacturer planning to make a handset that supports HEVC would be facing an estimated $1.60 per handset charge. As explained below, that becomes an estimated $2.25 handset charge once unaffiliated companies are included.

Table 3 uses the maximum rate for handsets. For example, HEVC Advance has a rate structure with tiered regional rates and lower prices if only the main profile for HEVC is implemented (as opposed to also implementing all of the optional extensions). Depending on your specific handset configurations and shipping locations, you might see lower rates than this table suggests. If your company made and sold 100% of its handsets in China and included none of the extensions, the HEVC Advance rate would be $0.20 per handset before discounts. As this brief discussion suggests, it is key to accurately modelling your company’s handset volumes and necessary features.

The third pool from Velos includes several companies with significant patent holdings: Ericsson, Sharp, Sony, Panasonic and Qualcomm. Notably, Velos is more than just a pool since it directly owns, licenses and can directly litigate at least some of the patents (though one of the MPEG LA pool holders appears to be an MPEG LA subsidiary with former Panasonic patents). The USPTO patent assignment database shows assignments of over 100 patent assets into Velos Media LLC from members. We used the midpoint of $0.20 (MPEG LA price) up to $1.30 (twice the HEVC Advance price) to model Velos’s rate in Table 3 at $0.75. We modelled the total Velos controlled/administered patents as follows: starting with the assignments listed in the USPTO database, we performed an International Patent Documentation patent family expansion to identify approximately 800 patents and publications assigned to Velos. We then assumed that the participants retained relevant assets as well, which we would estimate as between three and five times the amount transferred to Velos. Thus, we modelled the pool multiplying by four in Table 3 and estimated 3,200 total patent assets. We acknowledge that this creates a bit of an apples-to-oranges comparison between the Velos pool and the other two pools. However, it is a useful starting point for analysis.

As with AVC, some companies will opt to sit outside of the pools for HEVC. Many have been concerned that Technicolor, a large historical participant in video codec patent pools, has opted to sit outside the three pools at this time. This specific example may be somewhat mitigated by the fact that Dolby (in HEVC Advance) has bought a number of patents from Technicolor. As this article went to press, Technicolour announced a deal with InterDigital, a known patent licensor, to own and manage the Technicolor patents (see this issue’s Insight piece on InterDigital’s acquisition of Technicolor’s licensing arm).

Either way, there will be additional licensing demands from companies outside the pools. By way of example from AVC, Motorola was a notable patent owner which opted to license directly. The litigation stemming from Motorola’s licensing programme includes a case brought by Microsoft asking Motorola to set a rate for AVC (Microsoft Corp v Motorola Inc, 2:10-cv-01823-JLR). In that case, the court used principles of proportionality to set a licence rate for Microsoft that was less than 1% of Motorola’s original ask (2.25% of ASP or $4.5 on a $200 ASP smartphone; the award was $0.00555 per handset or a 0.002% royalty on a $200 ASP smartphone).

Table 4. HEVC adopted contributions

Sources: Velos-provided analysis of HEVC meeting minutes decisions available here combined with ROL Group allocation of companies to pools according to January 2018 data. Totals are for the main profile and all extensions

Table 4 provides a framework for measuring the scale of risk from companies outside the three pools. The court ruled against the contribution approach in TCL Communications v Ericsson (SACV 14-341 JVS and CV 15-2370 JVS; December 2017 finding of facts by a magistrate judge in the District Court for the Central District of California). We are using contributions here primarily as a way of assessing the potential size of portfolios of unaffiliated companies. This table synthesises an analysis that the Velos pool is sharing (upon request) about the adopted contributions by the HEVC technical committee. In brief, Velos manually reviewed the meeting minutes of the HEVC technical committee (over 20 meetings) to analyse the disposition of the thousands of proposed technical contributions for inclusion in HEVC. They counted only those contributions that were adopted and not later removed. A single contribution from multiple parties was counted as a contribution for all of the parties (eg, if companies X, Y and Z contributed proposal #1234, then each of the three companies would be counted as having made one contribution). Contributions are a useful way of thinking about sources of intellectual property in the standard. Contributions are not a guarantee that the contributor is the sole inventor of intellectual property in the contribution but intuitively it should correlate.

We received similar data (but for contributions to the main profile only) from HEVC Advance. The two pools’ contribution counts, while different, were directionally quite similar. HEVC Advance’s data reflected more reassignments from patent transactions showing the purchase of patents from unaffiliated entities primarily by the Dolby entities and GE. Surprisingly, unaffiliated companies represent a large percentage of contributions, with 37% of all adopted contributions when both main and extensions are considered. Looking briefly at the pools, when both main and extensions are considered, Velos has the highest percentage of adopted contributions by far (39%). In contrast, for main-only, the gap was narrower, with the Velos and HEVC Advance pools being nearly equal after account was taken of reassignments of patents from unaffiliated contributors.

Returning to the purpose of contribution analysis: estimating unaffiliated patents. This can be done by using the average number of patents listed (or estimated) per pool to compute a ratio of patents per contribution. The average contribution ratio (20.6 per contribution) across the three pools can be multiplied by the number of unaffiliated but adopted contributions (305) to arrive at an estimated number of unaffiliated patents of 6,295. This would dwarf any single pool taken alone in size.

For reference, companies with five or more contributions that are unaffiliated with the pools are listed in Table 5. These 14 companies represent over 80% of the unaffiliated but adopted contributions. As with our discussion of Table 6 (below), these are companies where a cross-licence may be particularly helpful in reducing patent exposure without signing up for a patent pool.

That in turn gives us the insight to estimate the likely patent demands from the unaffiliated contributors. Given the quantity of contributions and estimated holdings among a group of sophisticated patent licensors, we estimate that the unaffiliated parties, in aggregate, will seek an amount comparable to the HEVC Advance rate ($0.65 per device). Using these assumptions, we estimate that the total bill for HEVC would be $2.25 per handset. On a $200 ASP smartphone, this would be a 1.1% royalty.

Before we discuss how to navigate the landscape, an exploration of the pool’s patent holdings is useful.

HEVC patent landscape

Unlike private patent negotiations, historically the video patent pools have made the patents licensed readily available to the public. In the case of HEVC, both MPEG LA and HEVC Advance have websites that list the patents available for license. A third pool, Velos, does not currently publicly list its patents for license.

The availability of public lists provides an easily accessible avenue for a number of different analyses. Three main areas are particularly valuable:

• patent priority year distribution, to compute remaining life;
• geographic distribution of patent coverage, to check alignment between patent coverage versus your practise of the patents; and
• distribution of ownership, to assess the value of direct licensing instead of pooled licensing.

Each of these will help with risk and cost assessments and will be considered in turn.

Before diving into the analysis, one other point is worth mentioning. Technically, the HEVC standard, like prior video compression standards, only specifies the format and mechanism for decoding video. This means that companies encoding video into HEVC format (or transcoding existing video into HEVC format) may do so in any of a number of ways provided that the result is a compatible decodable HEVC stream. For this reason, this discussion of licensing focuses on the decoding or user device side. While streaming services still face demands for licences for their encoding and transmission, we will not be covering those demands in detail here. Suffice it to say, patent owners and pools claim to have patents on the only, or most desirable, ways to prepare HEVC-compatible encoded video files.

Figure 4 shows the distribution of earliest priority dates for US-issued HEVC patents in public lists of patent pools (eg, MPEG LA and HEVC Advance; Velos does not currently publicly list the patents that it licenses).

Figure 4. Earliest priority year of US patents in MPEG LA and HEVC Advance HEVC patent pools

Note: a third HEVC patent pool from Velos does not provide a patent list and is not analysed

Source: MPEG LA AVC and HEVC and HEVC Advance HEVC patent lists analysed with Derwent Innovation for earliest priority year

The long backwards reach of priority dates (back to around 1996) highlights the dependencies between HEVC and prior standards, as discussed in connection with Tables 1 and 2. Notably, over 10% of the US patents have priority dates dating back to the AVC, or earlier, compression standards. This is particularly important with regard to patent overlap with earlier video standard licences. Pool overlap for the US patents listed for HEVC (MPEG LA and HEVC Advance only) that are already being paid for via the MPEG LA AVC pool list is at over 10%.

The patent licensors would argue that a field-of-use limitation to the prior AVC standard limits your usage and thus there is no double payment. However, as a practical matter (and the difference in costs is discussed more below) this smacks of double payments. Most devices and software will need to support AVC side by side with HEVC for an extended period. In many cases the AVC-duplicative patents are being practised by HEVC to achieve the same purpose as they were in AVC – although they are being used to achieve greater coding efficiency. Either way, from the perspective of an HEVC adopter, having over 10% of your HEVC payments going towards patents that are already covered by an earlier $0.20 payment is problematic.

Second, the priority year distribution is such that 50% of the patents in the pool will not expire until 2029. It is important to check for inclusion of assets with priority years after the standard’s first publication. For example, approximately 10% of US patents in the MPEG LA AVC pool have an earliest priority date after the standard’s first publication. No such assets exist within the HEVC pools yet, but this should be watched. This is not a problem per se due to continuing development of the standards – although it does require explanation by the patent licensors. As we saw in the discussion of the MPEG-2 patent pool history above, fees can become exorbitant later in the lifecycle when few patents remain, particularly when considered on a per-patent basis.

Next, we analyse geographic coverage. As shown in Figure 5, the two pools with public asset lists have very different geographic coverage. Consider whether your business is conducted mainly outside the countries where the pools have meaningful patent coverage and whether the requested rates line up with your use of the patents. Both pools have the greatest concentration of patents in China, Europe, Japan and the United States. However, neither pool currently has meaningful patent holdings in emerging markets, such as India or Brazil.

Figure 5. Geographic distribution of assets in MPEG LA and HEVC Advance HEVC patent pools

Prior video codec pools imposed licensing fees based on either the country of manufacture or the country of use. However, with more functionality being supported primarily through software implementations with hardware optimisations, how justifiable is such an approach? For example, if the devices will be nearly exclusively used in countries where there is no or minimal patent protection, is a full rate fair? Also, when looking at Figure 5, bear in mind that this is a snapshot in time. The pipeline of pending applications may fill in more of the gaps in the coverage provided by both pools.

The next thing to consider is the concentration of ownership of patents in the three pools. This is relevant because MPEG LA and HEVC Advance both allow for (and discount) direct licences. Table 6 shows that the patent owners in each pool have at least 50 assets (for Velos, all five licensors are listed as well as Velos itself, which is a patent owner). Notably, while the pools are quite large, there is a high concentration of ownership among approximately 25 companies. In fact, for each of these pools, the patent owners in Table 6 account for 90% of the listed patents. If your company enters a direct patent cross-licence with these companies, you may be able to drastically manage your cost – and risk – from the HEVC licensing pools. (See also the discussion of Table 5, with regard to unaffiliated companies and the discussion of Path 1.)

Before turning away from the portfolios, one last point is worth mentioning. This analysis has treated all of the patents as equivalent. However, intuitively we know that this is not the case. For example, one patent might be on an adopted vectorisation contribution covering how to decode video, while another might form the basis for the contribution for improved prediction algorithms. The second patent on the prediction algorithms might be seen as the heart of improvements of HEVC over AVC. If you were able to license and pay for the patents individually, you would not pay the same amount for these two examples. The pools might do well to distinguish their contents to emphasise a value beyond cost per number of patents licensed.

A ranking of SEPs has been attempted in some court cases. For example, the December 2017 finding of facts in TCL Communications v Ericsson grappled with expert rankings for SEPs in the LTE arena. Ultimately, the court was not persuaded to use the rankings in that case. However, both pools and potential adopters could do more work in this area to demonstrate value and model fees.

FRAND rates

Setting the correct rate – more formally, the FRAND rate – for a standard is challenging. Too low and innovators are undercompensated; too high and manufacturers avoid the standard. As such, it may be helpful to consider some general guiding principles:

• Manufacturers need a way to price the cost of using the standards into their product development.
• Licensors for HEVC need to recognise that it is one of many included standards in the stack (eg, that same phone will face patent licensing demands for LTE, WiFi and others).
• Ultimately, the manufacturers need to be able to make a profit.

This leads to the question: how do we use these principles to set a fair price?

Merely contrasting the combined impacts of Moore’s Law and falling prices with the more modest compression improvements in HEVC (over AVC) risks missing improvements that have accrued to manufacturers due to greater manufacturing efficiencies (eg, lower-cost supplies, improved automation, increased number and types of electronics sold).

However, even ignoring Moore’s Law and declining system prices, there are other factors to consider.

There are more alternatives to HEVC today than there were when $0.20 was set as the pool rate for AVC (including AVC itself). In addition, HEVC patent licensors are able to seek royalties on more types of devices compared to when $0.20 was set as the pool rate for AVC – devices which are shipped in vastly higher quantities.

Finally, there are several real advantages to avoiding litigation (eg, a shorter time to close licences and the fact that litigation is slow and expensive).

Further, because the HEVC pool efforts have fragmented, there is a less transparent, multi-stop licensing process. Contrasting HEVC with AVC (or even MPEG-2), there is no longer a one-stop shop at which one can buy a licence for the overwhelming majority of patents for the compression technology. Undoubtedly, many manufacturers may have resented the amount of the royalties for the past compression standards. However, it was simple and transparent. Manufacturers could easily build the cost into the bill of materials for their products. We believe that licensing may be slowed or stalled today because of the complex HEVC licensing environment.

On balance, given the patent overlaps with AVC (Figure 4), the disparate per-patent asks (Table 3), the impact of declining prices and Moore’s Law (Figures 2 and 3) and a general reluctance to take patent licences, it seems likely that many prospective licensees will find the current asks well above their understanding of the proper FRAND rate.

Managing a shifting landscape

We recommend that companies adopt a process aligned with Figure 6 to evaluate HEVC adoption.

Figure 6. HEVC licensing and adoption process

Measure benefits, costs and options

Given the high potential cost of HEVC adoption (estimated at $1.60 per handset for the three pools and $2.25 with unaffiliated companies included), will your devices and services benefit accordingly? First and foremost, it is necessary to determine the benefits of HEVC adoption for your product or service. Second, plan for some costs of using HEVC in your business plans, bill of materials and profit margins. We periodically see companies that had not planned for some licensing payments caught out when licensing demands arrive. Simply planning for some payments to use HEVC is a key step to supporting the business.

Next, plan to build support for multiple codecs to reduce dependency on HEVC and provide negotiating flexibility. One emerging alternative to HEVC – the AV1 standard from AOMedia, which is still under development – has gathered 30 leading companies, including streaming media content companies (Amazon, Netflix and Hulu), and hardware and software companies (Adobe, AMD, Apple, ARM, Broadcom, Cisco, Google, Intel, Microsoft and NVIDIA). AOMedia aims to make AV1 a royalty-free, next-generation video codec with quality and compression capabilities comparable to HEVC. We are not here to advocate for the AV1 standard; we are merely pointing out how a prospective HEVC adopter may view the situation. Similarly, we would be remiss not to point out that the royalty-free pledge is from the standard promulgators, not all patent owners in the world.

Returning to HEVC licensing, consider the licensing posture of the HEVC pools with a company in two scenarios. First where a company goes all in for HEVC and second where it selectively uses HEVC only while continuing to use AVC, AV1 and other formats when possible. The second scenario puts the company in a stronger posture for negotiations.

Performing a detailed analysis at this step enables a company to:

• understand clearly whether it needs to adopt HEVC (eg, does it benefit sufficiently, what are its chief competitors doing, will it be able to deliver the necessary features with or without adoption);
• prepare a clear roadmap so as to have a higher ground in negotiations;
• arm itself with the data to negotiate with the pools for a fair fee structure, including:
  • is it already licensed to large portions of the pools?
  • are the rates fair across all of the pools?
  • are its products made, shipped and sold where the pool has coverage?
• determine whether its suppliers should be covering these licences; and
• compare what the different pools say they cover.

With this analysis complete, the company can turn to the next step: waiting.

Wait for as long as possible

If HEVC support is not a must-have feature, the best advice is to sit on the sidelines. AVC will continue to provide extremely high-quality video, especially for mobile environments, cost effectively for years. While sometimes early adopters can secure more favourable rates and terms, at other times careful waiting allows you to better understand the landscape. For example, if as a prospective adopter you view pool X’s rates as too high, waiting and letting other companies fight the pricing issue may provide beneficial. Once waiting is no longer an option, the process shifts to picking an entry point into the licensing process.

Pick your entry point

Having chosen to adopt HEVC, how you go about the adoption and where you sit in the ecosystem are crucial.

There are four primary paths from which a company is likely to pick:

• determine that it is sufficiently licensed;
• negotiate first;
• litigate first; or
• launch and wait to be sued.

Two types of company are most likely to fall into the first category – already sufficiently licensed companies with an extensive array of patent cross-licences with the major HEVC patent owners and software companies producing applications that run on other companies’ platforms. Additionally, companies with extensive patent portfolios but which currently lack cross-licences with the major HEVC patent owners may be able to navigate this path to some degree as well, although it will be riskier.

For companies with extensive cross-licences, if you have sufficient patent licensing coverage you may assess the percentage of HEVC patents you already have licensed and determine that you have adequate coverage.

For software companies deploying into a mobile platform, you can often leverage operating system-provided HEVC functionality. It may be the case that this is already licensed, although it is important to clarify this with your vendor. Further complicating this analysis is that it may depend on whether the operating system vendor has paid a licence only for its own hardware as opposed to third-party hardware. A potential hazard here is that your own independent implementation or the use of an open-source implementation of HEVC would not be licensed. To clarify, if you believe that Apple already paid for the iPhone to be licensed for the iOS implementation of HEVC, that licence would not cover your independent, or open-source, HEVC decoder.

The next three paths are those available for everyone else. For example, a small mobile phone manufacturer wanting to make a phone with HEVC support would need to pick among them.

In the second path – negotiate first – you as a company proactively seek out and enter licences with all or some of the pools before launching your products with HEVC. We recommend approaching the negotiation with the pools in a manner akin to negotiating the licensing demands from large corporate patent asserters. Each pool claims to be responsible for licensing a large quantity of SEPs and it – like a corporate patent asserter – needs to prove the value. However, this analogy is not perfect. First, the counter-assertion track will not directly work against the pools themselves. Second, pools may claim a need to maintain uniform rates. However, it is early in the adoption cycle and few non-pool participants of significant scale have signed up.

What would approaching the negotiation like a corporate counter-assertion look like? The pools should be responsive to requests to substantiate the value they are delivering. This is especially true given the vast amount of money that each of them wants to collect. Typically for a licensing engagement of this scale, we would expect them to provide the following:

• An overview of the technical features in HEVC that their licensed patents are fundamentally enabling.
• At least 10 or more claim charts for patents in the pool mapped to the standard. The claim charts should:
  • be from a diverse group of patent owners in the pool;
  • cover a diverse set of sections of the standard;
  • explain the contributions of the patent owner to the standardisation process (or be accompanied by an explanation that a third party’s contribution to the standardisation process led to adoption of a technique that the patent owner had invented before the contribution);
  • explain why the patented claims are truly standard essential, mandatory and deployed; and
  • withstand analysis by, and questions from, your team.
• A royalty proposal that reflects and acknowledges the business realities of profit margins, royalty stacking, comparability and proportionality.

Of the three pools, Velos and HEVC Advance are willing to provide claim charts under NDAs. (At the time of writing, we learned that some of HEVC Advance’s claim charts will be available without an NDA.) Although we have not heard of MPEG LA providing claim charts, they may be available.

The third path is to litigate first. It may be that upon reviewing the requested rates and considering the global case law climate, litigation may be desirable with some or all of the pools (including their licensors where necessary).

For example, outside the United States, the UK patent litigation landscape may be attractive following the decision in Unwired Planet International Ltd v Huawei Technologies Co Ltd (2017 EWHC 711 (Pat)). The court in question recognised the need for a worldwide but regionally adjusted set of rates that respected comparability (nearest licensing deals) and proportionality (thinking about what portion of the patents for the standard are being licensed). The proportionality issue aligns with the issue raised in Table 3 and the comparison of the so-called ‘$ per 1,000 patents for handset’ row. Specifically, as Table 3 shows, different owners are valuing their patents at vastly different rates with no clear external-facing explanation.

Even in the United States, you could proactively force the issue into court. It may be possible with impleader or joinder to bring all three pools (including their licensors) into court in one litigation. This approach may be particularly desirable if your primary concern is to set a fair rate as opposed to litigate validity. The recent findings of fact and memorandum of law in TCL Communications v Ericsson from the US District Court in California appear to suggest that parties in FRAND negotiations would do well to consider using US courts to help set the rate.

The concern about setting a fair rate feeds into the question of how many SEPs there are for HEVC. The care taken in the screening process by the pools varies. For example, Haier has raised a perceived low quality of review as an issue in its case against MPEG-2 and ATSC patent licensors. The Unwired Planet court in the United Kingdom faced the challenge of a massive number of declared patents (over 100,000 patents from nearly 20,000 families). Table 7 sets out the largest numbers in Unwired Planet and compares them against those for HEVC. The vast reduction from nearly 20,000 families considered for LTE down to a mere 2,000 at the highest level may overestimate the reduction possible here, but it does highlight the risks of over-declaration. We listed the high-watermark rows in our table for LTE and HEVC separately to show why Table 7 may overestimate the potential reduction possible from this analysis. Specifically, unlike the pure self-declaration found in Unwired Planet, the pools for HEVC do exercise a screening function; thus the 17,000 patents we estimate have a potentially different character from the 141,000 declared to the European Telecommunications Standards Institute for LTE. Notably, the Unwired Planet numbers align fairly well with the expert conclusions and holdings for the number of SEP patents for LTE in TCL Communications. (One note of caution here on US courts: TCL lost a jury trial on a non-SEP Ericsson patent earlier in December. That Texas jury awarded much higher damages against TCL – $75 million for past use of one non-SEP patent – compared with approximately $17 million by the judge in California for past use of all the SEPs.)

The comparison with LTE licensing dovetails nicely into a consideration of how our estimate of $2.25 in HEVC licensing compares with the fees for other standards. For 2015, the patent licence fees for the LTE stack had a cumulative royalty per phone of $7.25 based on an industry wide ASP of $221.80 – approximately 3%. (See the Hoover Institution Working Paper 16011, “A New Dataset on Mobile Phone Patent License Royalties” by Galetovic et al (p 12)). This suggests a significantly lower estimate for per-handset royalties than previous studies but is aligned with public data about licensing revenues for the major LTE patent licensors. Further, it provides some perspective: all of LTE communication – a vastly more complicated standard – is approximately 3% royalty for a $221.80 ASP phone. The estimated rates for HEVC would put the bill around approximately 1% of the same phone. While we recognise that we are comparing HEVC asks with estimates of actual LTE payments, given the way that the pools operate, the pools’ asks and actuals should be the same. The spread suggests an opening for parties who are willing to push this issue. Patent licensors have often been unsuccessful in obtaining their preferred rates.

The fourth and final path is to launch and wait to be sued. For a small company, launching first may allow it sufficient time to assess the market demand, need, value and ability to charge for HEVC. For example, if the company launched a new streaming product and learned that it could not charge sufficiently more for HEVC (over AVC) to cover the expected royalties, it could discontinue the product. This path may mean that the company finds itself a defendant in a patent lawsuit brought by one or more of the patent licensors or pools.

We are not proposing that companies engage in so-called ‘efficient infringement’ and avoid compensating HEVC patent owners or a fix for the overall patent system. Instead, we are acknowledging that some companies will follow this path. To ignore that would be to miss a common strategy that both prospective licensees and pools need to consider.

Path considerations

Given the money and risks involved, we suspect that most companies will either knowingly (or unknowingly) adopt a hybrid approach that combines a mixture of paths two, three and four. For example, a small company might license with only one pool and choose to ignore the others (paths two and four). A larger company, on the other hand, might license with a different pool but then litigate with the other two (paths two and three).

Either way, given the pricing it seems unlikely that companies will run to license with all three pools. Which pool(s) get picked will vary based on the proportion of contributions, the licensing strength of the companies contributing, the ability to enforce, the scope or breadth of the patent claims and the geographies covered. Bear in mind that companies may continue to face significant risk from the patent owners of the pools from which they chose not to take a licence. Additionally, while we mentioned several cases that have set favourable rates for technology users in aggregate, the case law pendulum may swing back in ways that favour patent owners. If that happens, later adopters may end up disadvantaged.

One other option for potential licensees is to become licensors (eg, by buying relevant patents in the space and joining one of the pools). With tens of thousands of patents for sale each year, the patent market is a viable alternative for those who find themselves short of patents. ROL Group has already seen patents with HEVC claim charts being marketed by brokers.

Another consideration for negotiations is to see whether any of the pools will step up to help address overlapping patent claims. For example, will HEVC Advance help me if MPEG LA members come after me? Not broadly, but on a section-by-section basis of the standard. For example, HEVC Advance’s public materials link certain assets to specific sub-sections of the HEVC standard. One could interpret that as an assertion that their licensors are the true inventors of the key technology in those narrow, detailed sections. Does that not also suggest (at least to a degree) that those licensors should stand behind those claims if a third party later claims that it has a patent on that specific section? Or at least rebate or discount a portion of the fees paid by the company? If nothing else, a pool offering such a feature would be better differentiated on features other than price.

Luck favours the prepared

How you watch your 4K or 8K content depends on a highly complex video compression technology. HEVC appears to be a leading contender to enable your next binge watch – it certainly has hundreds of billions of dollars riding on it.

This article provides:

• a data-driven estimate of the likely asks for HEVC licensing across both the patent pools and the unaffiliated patent owners – $2.25;
• a framework for thinking about pricing in the pools – we used Moore’s Law, declining prices for consumer electronics and LTE handset royalty estimates as a backdrop for analysing pool pricing;
• a process for picking your entry point into HEVC adoption and licensing;
• a glimpse into the complexity and the number of parties licensing HEVC patents, meaning that some fights may be inevitable; and
• an introduction to thinking about how a data-driven analysis can help your company shift the costs and profits available to you from HEVC adoption.

We hope that the complexity of the patent landscape for HEVC has not caused you to mentally rewrite the sentence from our opening to read: “HEVC patent licensing is difficult, the math makes your head spin, the trade-offs are torturous and deciding what to do may make you question why you ever asked.”

Rather, we hope to bring to mind the advice of Edna Mode: “Luck favours the prepared” (The Incredibles, 2004). We have reframed the decision to adopt HEVC to one based in analysis and readiness: careful analysis combined with a well-planned strategy and execution will radically alter your cost structure for – and profits from – adopting HEVC.