I am a fan of advanced mathematics and computer models, and I still spend a fair amount of time building mathematical models and wring software. I am going to describe below some more advanced mathematical analysis of patents, looking for correlations between various characteristics of patents.

1. Conventional Patent Benchmarking
Common approaches to patent portfolio benchmarking include determining metrics and analyzing offensive claim coverage of existing and prospective competitor products. The metrics vary from simple (e.g., number of patents per R&D Dollar) to more creative (e.g., metrics that quantify the return on Dollars invested in patents, metrics that track the ability to innovate at the team or individual level mapped across specific technologies, etc.).

Some of these metrics can be difficult to develop and generally require internal and external data collection. Financial data for specific competing products or technologies is often not available in 10K and 10Q filings with the U.S. Securities and Exchange Commission (SEC). Sometimes more specific financial data is available from filings with foreign governmental bodies, especially for companies headquartered or otherwise incorporated abroad. But often, this data must be obtained through financial databases, harnessing the resources of internal marketing and sales groups, or by hiring consulting entities to perform specific financial analysis. Costs and inaccuracy can compound fast.

2. The Patent Input Data Set
I took a set of 500 United States patents that claim various forms of software implementations. All patents issued in 2006. I set my sample size at 500 so that my data set is large enough to ensure convergence of statistical modeling and increase the confidence of statistical correlations. I limited the search to U.S. patents, but this could be easily extended to U.S. patent applications, Pacific Cooperation Treaty (PCT) patent applications, foreign patents, or foreign patent applications. This analysis will apply regardless of the input data set of patents.

The reason I chose patents issued in 2006 was that I wanted newer patents that reflect modern patent prosecution trends and current examination approaches in the United States Patent and Trademark Office (USPTO). There have been some recent changes in the examination of U.S. patents (e.g., for business method patents), but overall the sample size of my input set is large enough to minimize these changes. Running this analysis on a more recent set of patents would account for that if desired.

On the other hand I wanted patents that were at least 2-3 years old so that there would be a chance for additional patents to start citing back to the input set of patents. That way I could obtain some meaningful data on forward references, which will help with the computation of some metrics that look at the relative value of patents.

The patents I chose cover a variety of technologies, including algorithms for computational biotechnology (e.g., protein mapping, DNA pattern analysis, etc.), software for semiconductor manufacturing, algorithms for optimization of foundry equipment utilization, software for system integration and testing in the electronics industry, methods for supply chain optimization and tracking, algorithms for image and other data compression and de-duplication (”dedup” “dedupe” or “De-Dup”), software for digital rights management (DRM), algorithms for data encryption for storage and transmission, algorithms for parallel and grid computing, software for social networking, application programming interfaces (APIs) intended to interface with e-commerce and social networking websites, software for financial asset management and other banking and investment portfolio optimization, and some more traditional business method claims.

Some of the patents in my input data set may very well be invalidated during subsequent litigation or reexaminations for a variety of reasons, from challenges under 35 USC 101 (e.g., depending on how the In re Bilski decision comes back from the Supreme Court and is subsequently interpreted by the USPTO and other case law) to invalidation based on prior art under 35 USC 102 (anticipation) or 35 USC 103 (obviousness). There are some intelligent guesses we could make even about the probability of invalidation using mathematical modeling. But that analysis is for another day.

Finally, the patents I consider here issued to a variety of entities, from large global corporations to mid-size and small companies, to research groups and universities, and to individuals and non-practicing entities. I am not going to mention any specific assignee names for this analysis, whether to praise or criticize patents, since I do not want anyone to call me next week with complaints. That is also why I am not going to disclose the actual patent numbers. But again, this analysis will apply to any set of patents.

One final comment: for this analysis I intentionally chose a set of patents that covers a wide range of technologies and companies. But obviously, when doing benchmarking for a specific company, the input data set could be restricted to specific technologies, specific companies, and/or specific industry segments.

3. The Mathematical Model
I wrote a computer program that imported a variety of attributes of the patents, including filing dates, publication dates, priority dates, number of domestic forward references, number of pages, number of claims, and so on. This allowed me to compute various metrics for benchmarking the efficiency of the corresponding patent prosecution processes, from the duration of pendency, to the delay in filing after the initial priority date was established, to the duration of prosecution and time to issuance post-filing.

Further, I pulled in all independent claims for each patent, then I parsed out the claims and counted the number of words in each claim, and then selected the shortest independent claim for each patent. A shorter independent claim does not conclusively identify a better patent, but positively correlates with a broader coverage scope. It is an art to write a longer claim that will be easier to prosecute yet still provide few actual limitations, and this method of word counting will not account for that. But when working with a sufficiently large data sample, on the average, shorter claims will indicate broader coverage.

Something else to keep in mind about shorter claims is that while they do indeed tend to have fewer limitations, they may also be easier to invalidate during litigation or reexaminations. I spend a lot of time trying to invalidate patents when defending companies against patent assertions or while negotiating patent licenses and looking for arguments to decrease license fees and royalties. In most cases, if non-infringement arguments are not promising, at least I rejoice in having to produce shorter invalidity claim charts for shorter claims. In this model, however, I can adjust for this issue by looking at various other characteristics of the patents, including the number of domestic references which is an indicator of the scope of the presumption of validity over the prior art (i.e., more prior art references considered by the examiner during prosecution indicate a higher probability that the patent will sustain an invalidity challenge).

One final comment: although my input data set has 500 patents and all associated characteristic data described above, for most analysis scenarios I am running some preprocessing to normalize data and filter out unusual data which may be attributed to errors in USPTO records or otherwise looks unusual based on my experience. So for most graphs, the actual number of points shown will be less than 500. Further, when looking at Fourier transforms or other results of data filtering, the number of samples shown will be lower due to data decimation, filtering or other algorithm artifacts.

4. Analysis and Results
Let’s take a look at how long patents take before being filed. This will give us a sense for how promptly the patent prosecution proceeds once a priority date is established.

a. Prefiling Delays
Figure 1 is a scattered data plot of the delay between the priority date and the filing date for each patent in the data set. The Y-axis shows the pre-filing delay in days, and the X-axis shows the patent number (i.e., rank order) of the patent in the data set.

Figure 1

This figure shows a concentration of delays at zero (0) days, and a concentration around 365 days. This makes sense, since many applications are filed with no delay – these are the non-provisionals filed as new applications, without claiming priority to any previous domestic or international application.

To better quantify the distribution of prefiling delays, let’s sort the data above before plotting it, as shown in Figure 2:

Figure 2

From Figure 2, it is obvious that about 40% of the applications were filed without claiming any priority. I think this is the ideal goal, since filing without relying on a provisional ensures that the priority date is not lost due to lack of enablement. When invalidating patents that claim priority to a provisional application, one of the first things I do is to check the claims as issued against the material disclosed in the provisional application to see if I can push the effective priority date later in time, to the actual filing date. More than 75% of the time I am able to do this. Based on this, as a general rule, I don’t like provisional applications if they can be avoided. This being said, provisional applications are still the right choice in a number of cases where filing dates are critical with little preparation, where cost deferrals are important, or where strategy relating to venture capital (VC) funding or merger and acquisition (M&A) events demands it. 

Also from Figure 2, we see that about 20% of the applications were filed with a delay of less than one year. Generally, these are provisional applications being converted to nonprovisionals. A small percentage of these applications are split applications being filed as continuations or continuations in part (CIPs) based on a recent nonprovisional filing. 

Figure 2 shows that about 25% of the applications were filed approximately 12 months after the priority date. Most of these applications are last minute conversions of provisional applications into nonprovisionals. I am going to suggest that these applications will tend to have a lower relative quality than the other applications. This is because on the average, when faced with a statutory bar of 12 months, some of these applications will inevitably be rushed through the prosecution process. I will explore this hypothesis below.

Finally, Figure 2 shows that about 15% of the applications were filed by claiming priority to preexisting applications, so these must be continuations, continuations in part (CIP), divisionals, or applications entering the national phase in the US under the Pacific Cooperation Treaty (PCT).

b. Relative Quality of Last-Minute Provisional Conversions
I suggested above that I expect that applications filed with a delay of approximately 12 months from the claimed priority date will tend to have a relatively lower quality (whether in terms of claim scope or based on other metrics) than the other applications. Let’s explore that hypothesis.

I extracted the patents with prefiling delays close to 12 months from the rest of the data, and we can now compute various metrics for these patents and compare those metrics against the rest of the patents. Let’s call this set of patents the “Last Minute Conversions.”

Figure 3 superimposes scattered data plots for the two classes of patents. The smaller data set shown in blue represents the Last Minute Conversions. The Y-axis shows the length of the shortest independent claim, measured in words.

Figure 3

As expected, it appears that in relative terms, the Last Minute Conversions tend to have longer independent claims. To verify this, I computed the average independent claim length in the two data sets (again, these were preselected as the shortest independent claims in each patent):

Average indpendent claim length for the Last Minute Conversions: 127 words.
Average independent claim length for the other patents: 116 words.

The difference is not large, but on the average it looks like patents that are filed with less time pressure (i.e., not close to the 12-month statutory bar) tend to issue with better claims.

Figure 4 looks at a different metric comparing the two classes of patents: prosecution delay. Again, the smaller data set in blue represents the Last Minute Conversion patents.

Figure 4

From Figure 4, it is obvious that the Last Minute Conversions have a longer pending time. This is confirmed by looking at the averages:

Average prosecution pending time for the Last Minute Conversions: 1,752 days.
Average prosecution pending time for the other patents: 1,483 days.

This suggests that patents filed around the 12 month statutory bar date tend to take about 270 more days to issue, which means that they tend to be more expensive to prosecute and will likely provide a lower return on investment for the patent owner.

A lesson from this analysis is that by waiting to convert provisional patent applications until the end of the 12-month statutory bar dates, on the average, a company will likely spend more for prosecution, will obtain narrower patents, and will derive a lower return on its investments in patent assets.

Figure 5 shows a 3D plot of the prosecution pending time (in days) and the length of the patent (in pages) for the patents that are not Last Minute Conversions.

Figure 5

To better see the spatial distribution of these data points, it is helpful to rotate this graph and view it from a few different directions, something that I can do on my computer. By doing that, it is possible to identify a general volume within which the majority of these data points fall. The mean prosecution pending time for these patents is 1,483 days, as identified above. The mean length of these patents is 26 pages. Benchmarking against this data set, a company should seek to have its patents fall close to these mean values as a first goal, and then refine the prosecution metrics further as described below.

The length of a patent is generally positively correlated with cost. Even for patent firms writing patents on a fixed fee, more pages generally mean a higher level of investment in the specification. The optimal length of a patent is to be long enough to ensure enablement of the claims, but no longer. The exception is if the patent is a strategic filing that will be used as the basis for subsequent continuations, divisionals and continuations in part, in which case investing in a longer patent at the application stage will be depreciated over more than one patent.

Consequently, except for “umbrella” or “omnibus” applications intended to serve as the basis for multiple future filings, to optimize its investment in patent assets a company’s goal should be to place its patents closer to the upper right back edge of this 3D graph (i.e., along the “Pat No” axis), such that the prosecution pending time and the length of the patent are both minimized.

Finally, let’s take a look at how the length of a patent correlates with the number of claims. Figure 6 plots the length of patents (in pages) on the Y-axis, mapped against the number of claims.

Figure 6

From Figure 6, it appears that on the average, regardless of the number of claims, patents tend to be between 10 pages and 30 pages long. Within this range, however, there does not seem to be any specific correlation between the number of claims and the length of the patent.

By applying a 2D Discrete Fourier Transform to this data, we can isolate the frequency spectra of the two data sets, as shown in Figure 7 and look for patterns:

Figure 7

From Figure 7, we see a slight distribution bias in a planar region along the Y = a * X line (where ”a” is close to 1 and indicates the slope), suggesting a slight correlation between the length of the patents and the number of claims, which is something we guessed from Figure 6 for the 10-30 page range.  But other than that, Figure 7 shows an essentially random distribution of points for the 2D frequency spectra of the two data sets.  Other than the Y = a * X line distribution bias, there is no apparent frequency concentration in either dimension, and the data appears largely uncorrelated.  This is what we expected based on Figure 6.

In conclusion, the analysis above provides a few ideas on some more sophisticated metrics that could be developed to automatically benchmark or value a patent portfolio with minimal human and financial resource allocation. There are many other metrics that could be computed based on this or similar data, and there are some other interesting correlations I noticed playing with this data (e.g., the number of claims is positively correlated with prosecution duration (as expected), and is negatively correlated with the length of the claims (as hoped!)), but I have to go deal with my real job now.

As always, send me a private email if you want to discuss further. Let me know if you have a specific set of patents that you are curious about and I would be happy to run it through this model.

This is the first in a series of posts on issues that smaller companies face on a regular basis.  Some of the issues, however, will also apply to large public companies (e.g., commercial engagements and government contracts), but in this series of posts I will look at these issues from the perspective of smaller companies.  I will flip sides another time and look at such issues from a large company perspective (there are some significant differences in what matters most to each class of entities).

So here we go:

It depends.  Overall, having seen a variety of startups going through Initial Public Offerings (IPOs) or acquisition events and experiencing unexpected IP issues at the last minute, I now tend to favor proactive Intellectual Property due diligence.  But this should not be construed as an invariable general rule.

There are a few questions that the executive team should pose before deciding that proactive IP due diligence makes sense:

1. Are there any reasons to be concerned?

Is there anything unusual in the formation or evolution of the company?  For example, are there any early employees or consultants, including the founders, who may have worked for the company without a valid employment or consulting agreement?  Did any significant employees of the company leave prior prominent jobs with other entities and then made immediate and substantial contributions to company’s technology?  Have any other parties made any allegations of trade secret misappropriation against any company employees, whether directly, indirectly, formally or informally?  Did the company founders work on any aspect of the company’s technology while employed by any other entity?  Did the company license in any Intellectual Property from any university, research laboratory, other company, or any other entity or individual?  And so on.

2. Is the timing right?

If the company is firing on all cylinders and all personnel is fully immersed in business or product development, then the additional stress that could be posed by IP due diligence may not be desirable at this time.  While experienced IP counsel could perform much IP due diligence based on written materials, there is always a human component that requires some interactive collaboration with corporate personnel.  If a significant corporate event is not imminent, then IP due diligence could probably be delayed for now, provided that corporate IP management processes are sufficiently rigorous.

3. Are resources available?

If the company has internal counsel available, then IP due diligence could be performed with minimal external resources and limited added cost.  In many cases, however, startups do not have internal counsel at early stages, in which case IP due diligence would have to be done by outside counsel.  And even if internal counsel is available, the company may still choose to use external counsel for the analysis. 

One reason to engage outside counsel even when in-house counsel exists is to avoid distraction of in house attorneys who may be focused on other strategic and critical projects within the company.  Another reason to engage outside counsel is if in house attorneys do not have the requisite level of IP expertise.  For example, patent due diligence requires specific review from qualified patent counsel, depending on the level of analysis desired.  At one end of the spectrum, patent due diligence could be limited to verifying the chain of title for the company’s patents and patent applications, and for any patents that may be exclusively licensed by the company.  At the other end of the spectrum, a potential investor could be interested in the valuation of the company’s patent portfolio and patent counsel may therefore be proactively performing extensive analysis of patent claims, file wrappers and prior art.

4. Is there a good reason to conduct IP due diligence?

IP due diligence should only be performed when the reasons and expected value of the analysis are reasonably clear.  For example, I unfortunately see on a regular basis companies that invest significant financial and personnel resources to perform IP analysis that is not only unhelpful, but could actually be detrimental to the company.  A classic example is a noninfringement patent analysis that is unlikely to provide the company with any unequivocal product clearance conclusions, yet is virtually certain to bring to the knowledge of the company the existence of some patents that are potentially relevant to the business of the company.  Avoidance of willful patent infringement and treble damages should always be in the back of the mind of corporate management.

A good reason for performing proactive IP due diligence is to simply provide to the executive team a sense of comfort that the company’s corporate IP and product portfolios are in good shape. 

Another reason for engaging in IP due diligence in advance of a significant corporate event is to identify problems at a stage when they could still be rectified.  For example, a classic patent chain of title problem occurs when one or more employees of the company are named as inventors on a significant patent, yet the employees never executed a written invention assignment agreement.  This issue tends to arise with particular prevalence in companies that have activities outside the U.S., where employee hiring is sometimes handled in a less formal manner.  In a case like this, if the problem is identified early, the employees could be asked to retroactively execute assignment agreements, possibly in exchange for minimal additional consideration.  Unfortunately, many such issues are identified late, during an IPO or acquisition process, and some of these employees may have left the company and cannot be located or are unwilling to collaborate with the company.  The problem of an inventor who has not properly assigned ownership interest in a patent to the employer is a potentially-fatal issue for the respective patents, as a prospective infringer could obtain a license directly from the inventor and therefore make the patent virtually worthless to the company as an exclusionary instrument.

Another reason for performing proactive IP due diligence is to enhance the company’s image in front of potential investors or acquirers.  This is a subtle issue and requires careful execution, as the company will not want to provide the results of its internal IP diligence analysis and then extend any warranties based on the sufficiency or accuracy of the analysis.  Instead, the company could present selected aspects of the IP due diligence conclusions in a manner that suggests solid corporate IP management processes, careful and competent legal management, and generally inspires confidence in the company’s business and management team.  If the prospective investors or acquirers are represented by experienced IP counsel, selected results of prospective IP due diligence analysis coupled with prompt and organized IP due diligence document production may also reduce the duration and scope of the due diligence analysis by such investors or acquirers, and may buy good will for the company.  Additionally, a careful presentation of selected IP due diligence conclusions may even save the company some money: a small secret in the world of equity transactions is that the target company often pays for the cost of diligence, normally post-event – so any reduction in the time and cost of the due diligence process could provide multiple benefits to the company.

Scope of Intellectual Property Due Diligence

So if the company carefully weighs the pros and cons and then decides to perform proactive IP due diligence, what should the scope of the analysis be? 

The answer depends on many factors, including the nature of the expected corporate event (e.g., the due diligence analysis for an IPO will differ from the analysis for an acquisition), the technology and industry segment of the company (e.g., a fabless company will pose different issues than a manufacturing entity), and the nature of the Intellectual Property and product portfolios of the company (e.g., an organically grown business will raise different concerns as compared to Intellectual Property and technology licensed from third parties). 

At a minimum, the IP due diligence analysis will include a review of the Intellectual Property portfolio of the company, including the chain of title of copyrights and patents.  The analysis will further involve a review of confidentiality procedures across the company with the goal of evaluating the degree of protection afforded to trade secrets of the company and of third parties doing business with the company, including a review of standard and nonstandard nondisclosure agreements, employee and consultant agreements, corporate policies and confidentiality clauses in commercial and IP agreements.

The IP due diligence review may further consider inbound IP agreements (e.g., to assess adequacy and degree of certainty associated with any Intellectual Property licensed by the company and integrated into company products), outbound IP agreements (e.g., to ensure the valuation of the company’s IP portfolio was not inadvertently impaired by overbroad outbound IP licenses), vendor and supplier agreements (e.g., to assess supply chain security and Intellectual Property indemnification coverage), and customer commercial agreements (e.g., to evaluate any Most Favorite Nation (MFN) rights extended to third parties and potential IP indemnification and liability exposure of the company).

For companies with special profiles (e.g., companies with an existing or prospective Intellectual Property royalty-generating licensing business, companies in dynamic industry segments that rely heavily on innovation and Intellectual Property leadership to prevent cost-based competitive erosion), a deeper dive into the company’s patent portfolio may be desirable.  This could include claim analysis, file wrapper analysis, validity analysis and offensive product infringement analysis.  Potential side benefits of such patent analysis could be the decision to offer certain patents to third parties for purchase or exclusive licensing, to abandon certain patents in specific jurisdictions, to modify the company’s domestic and/or foreign patent filing strategy, or to make patent assertions against certain third parties.

In conclusion, while proactive Intellectual Property due diligence should not be perceived by any startup company as an unequivocal necessity in advance of a corporate exit or financial event, it will often not only be beneficial for such an event, but may actually provide benefits that extend to other aspects of the company’s business.  The only rule about proactive IP due diligence should be that each startup company should consider its desirability at some point, in advance of a potential IPO, acquisition or investment event.

For a wide range of cases, however, it is easy to demonstrate that progressively escalating litigation is an inevitable outcome given a plaintiff and defendant that make rational choices.  In this post, I will use game theory to analyze this issue.

Let’s consider a situation of asymmetrical revenue exposure with a smaller plaintiff company initiating patent litigation against a larger competitor.  To simplify the analysis, let’s assume that the plaintiff company (“MicroCo”) has approximately ten times less revenue than the defendant company (“BigCo”). 

MicroCo has a business unit manufacturing and marketing a particular class of medical equipment with revenues totaling $10M/year.  MicroCo considers this business unit fundamentally important to its future and believes that in the next three years revenue from this business unit will increase significantly, while revenue from the rest of its business units will stagnate or decline.  MicroCo’s total annual revenue is currently $100M/year. 

BigCo also has a business unit focused on the same general class of medical equipment with revenues of $100M/year.  The total annual revenue of BigCo is $1B/year.  BigCo also believes that this particular class of medical equipment will experience significant demand over the next few years and is therefore strategic to the company.

MicroCo and BigCo have developed solid patent portfolios over the past five years in anticipation of potential litigation, and each company can reasonably assert patents against all the revenue of the other company, including against each other’s business unit focusing on this particular class of medical equipment.

MicroCo has now decided to file a patent infringement lawsuit against BigCo.  MicroCo’s executive team, however, is split over whether to assert patents against all of BigCo’s revenues, or whether it should limit the assertion to BigCo’s medical equipment business unit. 

Using an extensive form game, we can use game theory to demonstrate that the optimal strategy for MicroCo is to assert patents against all of BigCo’s revenues.  Further, we can show that the Nash equilibrium for this situation is the case where BigCo retaliates by asserting patents against all of MicroCo’s revenues.

To develop a decision tree for our extensive form game, we need to define a set of payoffs for the various possible outcomes.  Let’s assume a 3% reasonable royalty rate for damages likely to be awarded by the court in this technology space (the actual royalty figure is irrelevant for our have theory analysis).   Let’s also assume that each company has sufficient patents with adequate claim scope to reasonably expect a decision of infringement against all the target revenue.  In reality, this assumption would need to be carefully evaluated, and may need to be replaced with a probabilistic model involving expected values.  The expected values would then be factored into the payoffs, with the game theory analysis still taking place as illustrated below, and the Nash solution being found the same way.  This same concept also applies to legal costs incurred by each company during litigation – each company’s expenses would be factored to decrease the expected payoffs, but the analytical approach would remain the same.

The following payoffs can be computed in this case:

1. If MicroCo (the plaintiff) asserts patents only against the particular medical equipment business unit of BigCo (the defendant), MicroCo will get a payoff of $3M (3% of $100M).  We will consider this a small scale assertion by plaintiff, denoted “P Small.”
2. If MicroCo asserts patents against all of BigCo’s revenues, MicroCo will get a payoff of $30M (3% of $1B).  We will consider this a large scale assertion by plaintiff, denoted “P Large.”
3. If BigCo asserts patents only against the particular medical equipment business unit of MicroCo, BigCo will get a payoff of $0.3M (3% of $10M).  We will consider this a small scale assertion by defendant, denoted “D Small.”
4. If BigCo asserts patents against all of MicroCo’s revenues, BigCo will get a payoff of $3M (3% of $100M).  We will consider this a large scale assertion by defendant, denoted “D Large.”

We will denote payoffs for this analysis as (plaintiff payoff, defendant payoff), computed in $Millions.

The following decision tree can now be constructed for this analysis:

If plaintiff makes a small assertion, we will be in the left half of the tree.  Defendant then has to choose whether to make a small assertion or a large assertion.  It is obvious that a rational defendant will choose to make a large assertion.  This is because in the left half of this decision tree, the payoff for the plaintiff (which is strongly correlated to defendant’s loss) is always $3M.  So defendant will rationally choose to maximize its payoff, and the way to do that is to make a large assertion.  The local equilibrium for this decision branch is therefore (3, 3).

If plaintiff makes a large assertion, then defendant will also rationally make a large assertion following the same analysis as in the left decision branch.  The local equilibrium for this decision branch is therefore (30, 3).

As a result, when deciding whether to initiate litigation, plaintiff has to choose between two specific local equilibrium solutions, where the payoff for the defendant is identical.  So the rational choice for the plaintiff will be to maximize its own payoff.  Consequently, the Nash equilibrium for this analysis is (30, 3), where MicroCo asserts against all of BigCo’s revenue, and BigCo asserts against all of MicroCo’s revenue. 

In conclusion, if the parties are acting rationally, the inevitable outcome in a broad range of patent litigation disputes, including cases with distinctly asymmetrical revenue exposure, is large scale patent assertions on each side.  This suggests that as a general rule, given the likely escalating legal expenses and long-term financial uncertainty, a product or services revenue company should seriously consider all reasonable alternatives before launching into patent litigation even if its initial decision is to make a small scale assertion with the hope that litigation will be contained.  The odds are high that litigation will escalate significantly.

Let’s pretend that we just went through a patent infringement lawsuit, where we were sued for infringing a portfolio of patents.  Our company could be selling medical equipment, wireless end-user devices, infrastructure systems, chip-level components, or any other product.  The dispute started with a typical letter of assertion, offering us a license, which we declined.  We argued with the patent owner for a while, and we eventually ended up in actual litigation (maybe we initiated a lawsuit in a jurisdiction of our choosing, seeking a declaratory judgment of noninfringement with the defendant countersuing, or maybe we just got sued for patent infringement because we refused to take a license).  We then refused to settle during the course of litigation, and we eventually went through a jury trial. 

Let’s pretend that we are now retroactively analyzing our choices during the course of this dispute, and we can construct the following chart of events:

A bit more about this chart of events:

• On January 1, 2005, we received a letter of assertion, and we estimated that the probability that the patent owner would prevail if we were to get sued was 70% and we had about $50 Million of our revenue exposed to these patents;  we chose to refuse to take a license, however, maybe because we were concerned about establishing a precedent that we are easy victims or because we were concerned that the same party would come after us again with another set of patents as soon as we settled this particular assertion.
• On June 1, 2005, we went through an internal analysis where we decided that based on our understanding of the patents at that time, our revenue exposed was actually larger, $90M.  We strongly considered taking a license to put this behind us, but decided against it again.
• We eventually started to feel a bit better about our chances, and on June 1, 2006, we initiated a lawsuit for declaratory judgment of noninfringement in a jurisdiction that we thought would be more convenient for us (maybe a District Court in Northern California);  the patent owner did not blink and countersued us for infringement.
• On June 1, 2006, our impression that things were not as bad as we had initially feared started to come true: we had a favorable Markman Hearing, and based on the way the court construed the claims, we assessed that the probability that we would eventually lose the trial was going down to about 50%.  Now we were starting to feel better about our decision not to settle.
• On June 1, 2007, during depositions and discovery, we had a good development – say we found out that one of the inventors was aware of some relevant prior art for all of the patents and had not disclosed it to the U.S. Patent Office, therefore raising a question of fraud committed towards the USPTO.  At this point, we reassessed our chances and decided that the probability that we would lose this trial was now only 30%.

So here is the question: did we do well to not settle through June 1, 2007?  If we had another chance to do this again, would we still make the same choices?

Here is a chart that shows the revenue we had exposed, as far as we knew, during this time:

What can we say about our litigation performance based on this chart?  The answer is: “Not much.” 

First, when dealing with patent litigation, we have to distance ourselves a bit from the absolute numbers.  The question that naturally comes up in connection with a patent lawsuit is, “How much revenue is exposed?”  We then tend to compare that with the expected cost of litigation and make an instant mental assessment of the best choice.  This type of a snap analysis, however, misses the fact that it is only a FRACTION of the revenue that is truly exposed in the lawsuit.  Indeed, if litigation goes through all the way to a verdict, the court will usually vet through a lost profits analysis and move on to a “reasonable royalty rate,” which tends to be in the range of 1-5% for the semiconductor industry.  So the actual loss that we could reasonably expect to incur with a verdict of infringement is probably less than 5% of the revenue exposed.  (Let’s ignore the risk of willful infringement and treble damages in this discussion, especially as recent case law seems to make this a harder target for a plaintiff.)  Consequently, the numbers shown in the chart above would have to be scaled down significantly.

Second, when making decisions regarding an optimal course of action, we can’t act based on absolute numbers.  Instead, we are interested in expected values – expected gains, or expected losses.  In this case, there is additional information in our table that we need to use to improve our risk analysis: we need to compute the expected value of the loss at each of those points in time, which we can do by multiplying the probability that we lose the lawsuit by the revenue exposed (and let’s not forget to scale the results down by about 5%).  Here is the new table we obtain:

With this information, we can generate a second graph of our risk exposure profile:

 From the enhanced chart and graph, we can now see that our best time to settle was January 1, 2007.  Taking into account the cost we incurred taking this case through trial, however, it would have been even better from a financial standpoint if we had settled on January 1, 2005.  So retroactively looking, our best decision would have been to take a license from the beginning, assuming we could have done it sufficiently cheaply.  Or at least we should have tried, knowing what our maximum cap was at the time.

But we initially contemplated the possibility that rejecting a license could cost us more, and we made a strategic decision not to take the license.  So then, we could walk along the graphic showing our Expected Value of Loss and try to find the most advantageous position once our initial reason for not settling was moot.  We could arguably think that once we initiated a lawsuit for declaratory judgment, we have shown all other prospective patent holders that may be contemplating assertions against us that we will not just roll over and pay money. 

In this case, the question becomes when after January 2006 we should have settled, if at all.  Well, the first time something changed in our analysis was July 2006, when our expected value of loss dropped.  So that would have been a reasonable time to take a license and pay some money.  But obviously, we would have done even better if we had pursued the litigation further and gotten to the favorable development in January 2007. 

And just to make this analysis harder while helping us feel good about our hypothetical decisions, let’s pretend that we knew what we were doing all along: in July 2007, the court issued a judgment that the patents asserted against us were invalid.  So we won in the end.

With that in mind, obviously, pursuing the litigation to the end was the right course of action.  Unless….

Well, unless the litigation cost ended up being higher than the cost of the license.  The cost of litigating a trial like this could conceivably exceed $2M.  If we could have taken a license in January 2005 for less than $2M, that would have been a better financial decision.  But then, how about not establishing a track record of being an easy victim?

There was some value associated with not giving in to the patent assertion immediately, and to make this analysis more rigorous, we should also claim some credit for that decision.  How would we value this credit?  We could expect, for example, that our decision to not give in, and actually initiate a lawsuit against the patent owner, will deter some lawsuits in the future.  That would mean that we could claim now some cost savings from having avoided such suits in the future. 

So can we just claim that each suit we avoided would have cost us about $1M and that we avoided two suits in the next five years, for a total of $2M in savings?

Not really.  We would need to think in terms of expected values again.  So we would need to assess the probability that we would have gotten sued again, and then multiply that by the cost we estimated for the first suit, $1M.  That would give us the expected value for the savings in the first suit we avoided, which we could now properly claim as a credit in our analysis.  That amount would certainly be lower than $1M.

By the way, please note that by conservatively estimating a cost of $1M for this suit, we actually avoided a more complex statistical evaluation.  To be more precise, we would need to actually guess on the revenue exposed in that particular suit, on the probability that we would lose that suit, figure in our 5% scaling factor, and then come up with an expected value for our loss, which would now be further multiplied by the probability of getting sued in the first place to obtain the credit that we can rightfully claim.  In this analysis, we again made an assumption – we assumed implicitly that the probability that we get sued in the first place and that we lose that lawsuit were independent random variables.  That assumption may not be actually correct (i.e., someone who is likely to sue us is also likely to have a stronger case against us, so the two probabilities may have some correlation).  But this assumption is good enough for this type of inherently difficult analysis.

So what do we conclude here?  We did well to litigate to the end, but if we had not been concerned about establishing a precedent that we are an easy target, we may have been able to do even better by settling early on, before litigation, for an amount below $2M.

This type of an analysis can help us improve our decisions before and during litigation – the question is always whether there is an optimal point to settle, and if so, what that point may be.  Unfortunately we always work with incomplete information sets when making this type of decisions in real time.  Further, we tend to second-guess ourselves, especially if the analysis is suggesting a course of action that just does not feel right (e.g., sometimes it is emotionally unsettling to accept a license offer early on given our tendency to wonder whether we can do better by waiting, and sometimes we start dreading the uncertainty of litigation and feel inclined to pay more than we think we should just to put the matter behind us). 

And finally, one thing to remember is that this type of an analysis could be very useful, and may very well average out uncertainties over a longer period of time and multiple individual disputes, but in the short term, when dealing with human reactions that are not always rational or predictable, there is no substitute for human instincts and subjective choices.  We just have to make sure we hedge our bets and limit our downside exposure as much as we can.  Oh, and build an offensive patent portfolio.

- Marius

The Customer Perspective

From a fab operator standpoint, it is imperative to understand the differences in the IP infringement risks posed by the purchase of materials and equipment given that the fab operator’s goal is to divest as much of its IP infringement risk as possible to its vendors.  Further, the supply chain manager of the fab operator needs a rigorous and quantifiable model for benchmarking its performance when negotiating IP indemnification coverage from such vendors.

A critical concern of a fab operator is the threat of an injunction that could shut down the fab.  As every fab operator would testify, that is pretty much the worst possible scenario for a fab considering the tenuous economics of fab operation and equipment utilization.  A fab should simply never stop for more than absolutely necessary, and even then only partially, to the absolute minimum necessary, be it for a defect in manufacturing process or due to partial retooling;  but in no event should a fab even pause because of an IP-related, court-ordered injunction. 

The Materials Vendor Perspective

From the point of view of a materials vendor, the fab operator is making a lot of money by buying materials that cost a lot to develop, posed various environmental challenges to produce and transport, and are constantly under threat of obsolescence from marginally-better, constantly-improving competing materials.  Add to that the constant pricing pressure from increasingly consolidated fab operators, and the last thing that a materials vendor will want to do is extend broad IP indemnification coverage to a fab operator.

The Equipment Vendor Perspective

Equipment vendors selling increasingly-complex tools to fab operators who demand 100% up-time and highly efficient equipment also find themselves under pricing pressures, constantly fearing the next economic downturn or fab overcapacity indicators.  Faced with increasingly expensive technology development, investor pressures to grow sales despite stagnating global semiconductor demand and an inherent need to consistently innovate to solve problems posed by shrinking geometries and power efficiency requirements, equipment manufacturers will not take lightly to demands from their customers for broad IP indemnification protection. 

Add to this difficult operating environment the inherently litigious nature of the semiconductor equipment industry, both as a result of patent threats posed by competing manufacturers among themselves and by trolls and other patent holders that do not directly participate in the commercial chain. 

The resulting economic picture is not rosy for the equipment manufacturers, so assumption of any further IP risk by this segment of the industry is not a trivial issue.  Needless to say, when I represent a semiconductor equipment manufacturer, IP indemnification is something I think about carefully. 

The Materials and Equipment Vendors Risk Matrix

So put it all together: what are the relative degrees of concern that materials and equipment vendors should have in connection with IP indemnification covenants?  Here is a matrix that breaks this discussion down by the elements that are normally found in an IP indemnification covenant extended in a contract for the sale of materials or equipment:

1. A semiconductor design is created in the U.S.
2. The design is exported to Taiwan for production.
3. The semiconductor product is sent to Malaysia for assembly.
4. The assembled product is moved to Thailand for packaging.
5. The finished semiconductor product is sent to China for integration into larger systems or products.
6. The larger systems or products are then exported from China around the world, including to the US. 

This chronological view of the journey of a semiconductor device, from the design and wafer stages to the integrated system level, mirrors the Semiconductor Industry IP Model that is the central focus of this blog (please click on the graphic to the right or follow the IP model link at the top).  In my IP model, each node of the semiconductor supply chain could very well involve a cross-border transaction, from the sale of semiconductor equipment by a company like TEL to a U.S. fab operator, to the packaging of a wafer in Asia, to the sale of a finished system in Europe. 

The ICAF report concluded that the U.S. semiconductor industry is a leader in globalized industrialization, with innovators, producers, suppliers, markets, and users of semiconductors spanning the globe.  Previously, the U.S. competitive advantage was in the design, construction, packaging, and selling of microchips.  That is no longer completely true because the U.S. advantage does not necessarily include production and packaging.  

The ICAF report’s focus on globalization provides an excellent backdrop for looking at international patent exhaustion for the semiconductor industry.

1. What is International Patent Exhaustion?

International patent exhaustion is a legal theory under which patents issued in a country and reading on a product being imported into that country are exhausted if the product was initially sold with the consent of the patent owner.  Alternatively stated, a first sale of a product outside a country that observes international patent exhaustion acts to exhaust patents in that country if the initial sale abroad was authorized by the patent owner. 

Other terms that are associated with international patent exhaustion include “parallel importation” (describing the concept that products are imported in a country in parallel with, but outside the reach of patents issued in that country) and “gray market goods” (referring to the flow of goods through distribution channels that are not authorized by the patent owner). 

International exhaustion issues traditionally occurred in the context of trademarks and copyrights but have recently become more common in connection with patents, including by playing a central role in the debates involving the parallel importation of medicinal drugs.

2. Theories behind International Patent Exhaustion

International patent exhaustion is a consequence of encouraging global trade at the expense of intellectual property protection.  Market economics suggest that from a pricing standpoint, permitting parallel importation of products would prevent pricing discrimination by product manufacturers across country borders, with a direct benefit for consumers. 

While some studies have suggested that parallel importation does indeed benefit consumers by reducing prices in certain jurisdictions, critics of international patent exhaustion have argued that most of the benefits of parallel importation may be realized by the distributors handling the importation of the goods.  A fundamental argument against international patent exhaustion is that parallel importation diminishes the value of patents, which can eventually lead to a decrease in innovation, with detrimental long-term consequences for both consumers and industry.   Another criticism of international patent exhaustion is that it runs counter to the principle of territoriality of patent rights, a basic tenet in the legal tradition of many countries.

From a legal standpoint, courts reviewing patent exhaustion issues tend to employ an analytical framework analogous to the U.S. theory of implied patent licenses.  For example, when addressing international patent exhaustion, courts in various jurisdictions focus on whether the patent owner imposed any restrictions when the first sale of the product took place, which can also be a line of inquiry in implied patent license investigations. 

At the international level, the ability of each country to determine its own position on the issue of international patent exhaustion is supported by Article 6 of TRIPS, which, as a result of intense debate on this issue during negotiations of the treaty, explicitly disclaims any intent to regulate international patent exhaustion.  This leaves each country, be default, free to set its own policy with respect to international patent exhaustion.

3. International Patent Exhaustion in Select Jurisdictions

a. The Unites States

The U.S. does not recognize international patent exhaustion.  The exclusive right to “import” patented products conferred by 35 U.S.C. 271 provides patent owners with the right to control importation of patented products into the U.S., regardless of whether these products had already been sold outside the U.S.  U.S. courts analyzing international patent exhaustion issues have generally rejected parallel importation and have upheld the right of a patent owner to control importation of patented inventions into the U.S.  It is noteworthy that U.S. courts are rejecting international patent exhaustion despite the fact that national patent exhaustion is a fundamental principle of U.S. patent law.

b. The European Union

The E.U. has adopted a territorial application of the patent exhaustion doctrine, known as “Community Exhaustion.”  Under the Community Exhaustion doctrine, patents issued in any E.U. member country and reading on a product are exhausted upon the sale of that product in any other E.U. member country, as long as the initial sale took place with the consent of the patent owner.  Beyond Community Exhaustion, however, the E.U. appears to reject international exhaustion with respect to products sold outside the E.U..  A number of commentators have suggested that E.U. laws may nevertheless apply to exhaust patents issued in E.U. member counties with respect to products initially sold outside the E.U., but nothing in the E.U. judicial or normative record appears to support that position. 

Consequently, a product sold in any E.U. member country with the consent of a patent owner can be freely imported into any other E.U. member country;  if the product is initially sold outsider the E.U., however, the patent owner is free to block parallel importation of that product into all E.U. member countries.

c. Japan: International Patent Exhaustion Observed

Japan has adopted a broad interpretation of international patent exhaustion.  In 1997, the Supreme Court of Japan ruled that when a patented product is sold outside Japan with the consent of the patent owner, “the patentee should not be allowed to enforce its patent in Japan against the buyer unless the buyer explicitly agrees to exclude Japan from the place of sale or use.”  The court further held that once the initial buyer resells the product purchased outside Japan to any subsequent buyer in Japan, patents can not be asserted against the subsequent buyer in Japan unless a notice of the agreement excluding Japan as a country where products may be resold “is clearly placed on the patented products.”  (see BBS Kraftfahrzeug Technik AG v. Kabushiki Kaisha Racimex Japan, Supreme Court of Japan, July 1, 1997).

d. Other Countries

A survey of parallel importation laws in a number of other countries reveals that international patent exhaustion has been adopted by some countries, generally subject to limitations similar to those present in Japanese law, and has been rejected by others.  There seems to be a slight trend towards the adoption of international patent exhaustion, especially among developing countries that tend to perceive international patent exhaustion as a way to minimize the impact of their patent imbalance with more developed countries. 

The following table provides a summary of how international patent exhaustion is treated in various countries:

4. Thoughts about the Semiconductor Industry

As discussed in the beginning of this posting, the inherently global nature of the semiconductor industry makes it very likely that a semiconductor product will journey on its way to a final end-user product through both jurisdictions that have adopted the doctrine of international patent exhaustion and jurisdictions that have rejected it. 

So how can a semiconductor company that is investing in a global patent portfolio minimize the dilution of its patents by international patent exhaustion?

Let’s break down the answer by focusing on a few discrete areas:

a. Market Segmentation and Parallel Importation

To the extent that a company contemplates relying on patents to prevent the importation of its products into particular jurisdictions, international patent exhaustion may interfere with such an attempt if the respective jurisdictions observe international patent exhaustion.  This would be the case, for example, if the company sold a product in a certain jurisdiction but sought to prevent the importation of that product in a particular country that has adopted international patent exhaustion.  

In this case, the company may be unable to assert a patent granted in that country against a distributor importing or selling the product in that country, or against a customer using the product in that country.  Please note that most companies would normally NOT assert patents against their own customers, but the ability to make such an assertion may be critical in a defensive situation, where the customer initiates a patent infringement lawsuit against the company and the company has to defend against such a suit.

The company’s ability to prevent unauthorized parallel importation into such a country may be enhanced by imposing explicit geographic restrictions regarding the sale and/or use of the product at the time of the initial sale.  This is because many countries adopting the international patent exhaustion doctrine tend to permit the manufacturer to impose geographic restrictions at the time of the initial sale and because these countries tend to accept such restrictions as barriers to parallel importation. 

Some of these countries, however, also require that the restrictions be reiterated at each level of the commercial distribution chain and that subsequent buyers and users receive explicit notice with respect to such geographic restrictions, with some jurisdictions requiring that the products themselves be appropriately marked (e.g., Japan).  Further, attempts to impose geographic limitations on the subsequent sale or use of products should also be evaluated in light of any other limitations that local law may impose against geographic division of markets.  For example, market segmentation coupled with multiple company arrangements and market power may raise the interest of antitrust authorities, especially given the recent antitrust assertive trend in the E.U. and some Asian countries.

b. License Agreements

To the extent that patent licenses granted by a company include geographic restrictions (e.g., in SIG intellectual property rights agreements with hybrid RF/RAND regimes or in patent cross license agreements), international patent exhaustion may interfere with the intended geographic restrictions in countries that observe international patent exhaustion.  For example, if a patent license granted by the company sought to curtail the licensee’s ability to sell or use licensed products in a particular country, if that country has adopted the doctrine of international patent exhaustion, the licensor may not be able to prevent the importation into that country of products that were first sold in a jurisdiction authorized under the agreement. 

As discussed above in connection with market segmentation, requiring the licensee to impose geographic restrictions and appropriately marking products at the time of the first sale may help prevent parallel importation of licensed products into unintended countries, but the logistical complications associated with marking products and with policing compliance at each level of the commercial chain may be significant.

c. Foreign Patent Prosecution Strategy

Companies may consider reevaluating their foreign patent prosecution strategies in light of the limitations that international patent exhaustion may place on enforcement of such patents in particular countries that observe international patent exhaustion.  For example, to the extent that the value of particular patents in a specific country that permits parallel importation would be derived from preventing importation and use of products initially sold by the company outside that country (or manufactured outside that country under a license from the company), the need to prosecute and maintain those particular patents may be correspondingly reduced. 

This situation would be most likely to arise in a developed country that permits parallel importation, which is a significant commercial market for the company, and which tends to import company’s products (e.g., Japan for most semiconductor products other than flash memory);  in this case, if the products being imported into that country are generally produced by parties covered by cross licenses with the company, the company’s ability to assert patents issued in that country against such imported products could be significantly reduced, and the reason for maintaining those patents should be reconsidered.

So the question arises then for the larger companies: if their relative position versus smaller companies has been altered in a negative direction, then have large companies recorded a net benefit from this change in cost structure?

The answer to this question is pretty complex.  Here are some relevant factors for this analysis:

• Cheaper patents have placed increased demand on patent attorneys and agents, especially in the U.S., where these professionals are expected to write strong patents, with broad claims and flawless specifications, on ever-decreasing budgets.  My subjective impression is that the quality of patents on the average has been going up despite the shrinking budgets, and I credit this to better trained professionals with enhanced access to technical materials, legislative materials and legal decisions. 
• On the other hand, common sense suggests that a decrease in funding would result in lower quality in the end product.  And indeed, unless the system is nonlinear (maybe because patent professionals are doing a better job, as I suggested above), this would be the natural outcome.
• But even if the quality of patents were lower on the average, a large company may be able to make up for that by prosecuting a larger volume of patents, stretching further in scope and capturing a broader range of technologies.

There are other factors that should be considered if we were to perform a comprehensive economic analysis, including for example the relative commercial position between small and large companies given that the dynamics that affected the patent prosecution cost structure have also affected the product and services competition in the industry.  But I will remain focused here on the narrower question of whether large companies are better or worse off as the cost of patent prosecution has gone down in the U.S. and other jurisdictions.

The best way to answer this question would be to take the patents issued to a few large companies between 1993 and 2000, map their claims to then-corresponding products in the industry, then perform the same analysis for patents and products between 2000 and 2007, and compare the two.  Such an analysis would be prohibitively expensive and require information that may not be available anymore (e.g., who remembers much about Ethernet routers or car phones available in the early 90s?). 

An approximation of this answer could be achieved by limiting the analysis to statistically-sufficient samples of patents for each company, but even that would require significant analytical resources and product information.

A second order approximation could be obtained by ignoring the mapping of claims to products and simply comparing the older patents with the newer patents.  Such a comparison, removed from an actual infringement analysis, is inherently more subjective and imprecise, but could still be useful in assessing the relative quality of the patents.  A further refinement of this approximation would ignore issues relating to validity (e.g., prior art analysis) and P 112 issues (e.g., enablement) and simply compare the scope of claims between the two classes of patents.

Comparing the scope of claims of older and newer patents would still require significant human analytical resources, so ideally we would like to find an automatic, computer-based approach to this problem.  There is currently no Artificial Intelligence engine capable of evaluating the scope of patent claims for purposes of such an analytical exercise, so we would need to come up with some more discrete metrics. 

One approach would be to run batches of patents through a semantic engine that parses out the independent claims of the patents and simply counts the words in those claims.  The problem with this approach is that it is difficult to correlate the number of words in claims with the scope of the claims.  For example, less words would tend to indicate broader claims, given that more elements in a claim tend to imply more limitations, and therefore a narrower claim.  On the other hand, the semiconductor space has arguably become more crowded over the past 7 years, and longer claims may very well be inherently necessary to write valid claims. 

So according to this analysis, if we were to find that on the average, recent semiconductor independent patent claims issued to large companies have fewer words than older claims, we could conclude that the relative quality of patents has increased.  Coupling higher, or at least equal, quality of patents with a larger absolute number of patents issued to large companies, we could then conclude that large companies have benefited from the decrease in patent prosecution costs.

On the other hand, if we were to find that on the average, the number of words in patent claims has increased, the result would be inconclusive.  We would need additional analysis to understand whether longer claims mean lower quality or are simply the result of a crowded space.

Can you think of any other metrics to assess whether recent patents are better or worse than older patents, taking into account the larger density of patents in the semiconductor space?  I will follow up on this topic in the future.

Default Applicable Law and Information Asymmetry

In most jurisdictions, commercial law generally provides default mechanisms for allocation of risk, generally favoring the buyers.  In the U.S., for example, the Universal Commercial Code (UCC) provides the buyer of goods with a default warranty of noninfringement of IP rights.  That is a rational approach, considering that from an information theory standpoint, the seller will generally know asymmetrically more than the buyer about the subject of the transaction.  Consequently, it is fair to allocate to the seller the risk of IP infringement.

Commercial law generally allows the parties to override the default applicable law.  In a further attempt to protect the buyer, however, overriding those default provisions may need to observe additional requirements.  Under the UCC, for example, disclaimers of the warranty of noninfringement of IP must be written in a conspicuous manner, which is why most agreements governed by U.S. law include such a disclaimer in capital letters.  Parties to commercial transactions will usually go further, and will include in agreements affirmative obligations of indemnification extended to the buyer, with various limitations and specific remedies.

Hybrid Good and Services Agreements

It is important to realize that many master product purchase agreements entered into by companies in the semiconductor space also include service provisions.  This is normally done for the convenience of the customer.  The IP issues raised by service transactions differ from those raised by the purchase of goods, yet such hybrid master purchase agreements do not normally differentiate between the two.  In many such transactions, however, the buyers will be able to impose their own template agreements, and these agreements will normally have very broad, overreaching indemnification provisions benefiting the buyers. 

It is critical for a seller of equipment or other goods, therefore, to understand its indemnification risks and the likely financial exposure in such a transaction.  The goal of the seller should be to minimize its risk profile by focusing on the areas of these agreements that will provide the maximum return on negotiation investment.  The IP indemnification section should generally be at the top of that list.  In many such transactions where the buyer is insisting upon broad indemnification covenants, the seller may be able to exploit indefinite language or the ill applicability of template provisions to the specific products or issues covered by that transaction. 

Negotiating IP Indemnification Frameworks

The customer should also understand the actual IP infringement risks that it is likely to incur once it places in service the respective equipment or goods.  Rather than simply asking for the broadest possible indemnification coverage it can extract from the seller, the customer should attempt to focus the indemnification clause on these particular risks.  Regardless of the leverage that the customer may have in such transactions, asking for broad and unilateral indemnification covenants from the seller has the potential to antagonize the negotiations, shift the discussions to the sections that address limitations of liability, drive the seller to take stronger positions on liability caps, and possibly even push the seller into a corner where the seller will feel justified in standing its ground on the IP indemnification negotiations, with a potential outcome that is suboptimal for both parties.

The Hidden Risk:  Liability for Use of Products

Early in my legal career, I was representing a fab operator in a transaction with a company selling semiconductor manufacturing equipment.  The transaction included a sale of goods component, and at some point the discussion turned to IP indemnification.  Given my limited experience at the time, I did not think much about IP indemnification, and instead I simply asked for a broad indemnification covenant – everything but the kitchen sink.  The equipment seller made all the objections that I was expecting, but then offered me a deal:  they were going to forfeit some of their objections if I agreed that they would have no indemnification obligations for our USE of their equipment.  That surprised me, and unfortunately they were unable to explain to me why they had asked for that.  They had apparently heard somewhere that indemnifying for use of equipment was “bad.”

Well, I ended up doing some research on that issue, picked up some cases on patent infringement damages, read some legal decisions that included IP indemnification disputes, and later in my career found myself involved in patent assertions that directly involved the issue of infringing use of equipment.

I now think about IP indemnification for infringing use in most transactions.  It is not a trivial issue, and in the foundry space it is something that both the sellers of equipment and materials, on the one hand, and the fab operators on the other hand, should consider carefully in most transactions. 

Applications beyond the Semiconductor Industry

Further, this is an issue that permeates most commercial transactions, even beyond fab operators and the semiconductor industry:  the IP infringement risks posed by use of goods sold in a commercial transaction differ fundamentally and generally from the issues raised by the USE of those goods.  Examples of these differences include the nature of the patent claims that are involved (e.g., method claims v. device claims), the geographical characteristics of the IP infringement, patent portfolios involved and relevant litigation forums (e.g., the location of the use v. the location where the equipment is manufactured and/or sold v. the percentage of revenue attributed to the infringing use in that particular geography, etc.), the complexity of the goods compared to the simplicity of the usage as it may appear to a judge or jury, and so on.

Privity of Contract and Patent Portfolio Strategies

Finally, the privity of contract and commercial relationships established via IP indemnification provisions are issues that could prove to be determinative of outcomes in infringement assertions and should be carefully considered by sophisticated patent portfolio developers.  For example, when developing a patent portfolio with a licensing component, it is critical to consider the mapping of the patent coverage to the various links in the commercial chain, superimposed on the probable indemnification relationships existing between the corresponding parties.  This is because IP indemnification obligations could fundamentally alter the economic calculus and the litigation strategy of all parties in connection with patent assertions or other licensing discussions, including those of the patent owner and of the target licensee.

Conclusion

IP indemnification is an issue that deserves careful consideration in most commercial transactions in the semiconductor industry.  Applicable law rightfully protects the buyer in most jurisdictions, but the parties usually go beyond that and enter into specific agreements that explicitly define the IP indemnification protections extended to the buyer.  Unfortunately, however, many of these transactions are governed by standard template agreements, with only limited effort by the parties to define and implement a realistic allocation of risk based on the information available to the parties at the time of the transaction.  The downside of this approach is visible in many transactions, where the parties either under-invest in negotiating these clauses, or one of the parties dictates terms that are overprotective of that party in some ways, but often nonspecific and vague in other ways.  The optimal solution for two sophisticated parties will always be to fully negotiate IP indemnification coverage by learning about each other’s products and use scenarios, and then realistically assessing each party’s risks and economic costs.

This got me thinking about a conceptually-similar phenomenon that has affected the IP world: the relative commoditization of patent prosecution, with the cost of patents being driven significantly lower over the past 10 years.  Has this been good or bad for our industry?  Has it altered the balance between larger companies and smaller entities?

First of all, we need to qualify this discussion with the observation that while patent prosecution costs may have decreased at the micro-level, this was offset to a large extent by a general tendency to prosecute more patent families on a global scale, taking full advantage of the Pacific Cooperation Treaty mechanism for deferring prosecution costs and expanding the global reach, and by increasing filing and maintenance fees around the world.  Consequently, as any company building an international patent portfolio will likely agree, the costs of international prosecution are dramatic and will likely continue to increase.

And yet, lower patent prosecution costs in most individual jurisdictions, especially the U.S., have certainly helped all entities stretch their patent prosecution budgets further.  This has meant different things for different entities:

• For companies in the semiconductor equipment and services space, which tend to be smaller (with some notable exceptions, however), this has allowed them to get more patents without dramatically increasing the percentage of their R&D budget allocated to patent prosecution.
• For larger companies, including the fab operators, this has allowed them to continue to get large numbers of patents on an annual basis without diverting financial resources away from the crushing costs of capital equipment and fab operation.
• For companies in the ODM, back end processing, and contract manufacturing spaces, which were not traditional patent powerhouses, this has allowed them to step into the field and start accumulating significant patent portfolios.
• Finally, universities and research consortia have been able to ramp up their patent prosecution efforts, developing significant patent portfolios that have become a new source of concern for product companies that have found themselves at the receiving end of patent assertions based on such patents (although not always made directly by the initial patentees).

Many of these companies, especially non-US entities, would have increased their patent portfolios defensively anyway given the ever-increasing level of patent infringement assertions directed at them, but lower prosecution costs must have certainly helped across the board.

So there is no question that cheaper patents have meant more patents for everyone.  But back to my initial question: has this altered the relative balance between companies in the semiconductor industry?  I think so.

I would argue that this has helped more smaller companies and newcomers to the patent arena.  This is because of the inherently-asymmetrical impact of patents in disputes between companies with discrepant product revenues.  Simply put, a small company will only need a few patents to hit a significant amount of revenue of a large company, whereas the larger company will need a comprehensive patent portfolio carefully mapped to all aspects of the smaller company’s product space, and even then, that may still result in relative advantage for the smaller company.  So for any industry segment, if all companies are getting more patents, I think that as a general rule, the relative value of each patent will be inversely proportional with the product revenue of the patentee.  This means that if you are a smaller company, you don’t need to get too many patents, you just have to map them well to the right competitor revenue.  If you are a big company, you have to be ready to spend some money.  Now, of course, both smaller and larger companies may get creative and find better ways to fine tune their patent portfolios depending on the particular industry dynamics, but this is a discussion for another day.

The Economist article suggested that, “A widescreen plasma television is lovely, but you do not need one to laugh at Shrek.”  In our industry, this may translate into, “A large patent portfolio is nice, but you only need a couple of targeted patents to keep at bay a bigger competitor.”