Sisvel’s new Head of AI Pools explains why the data centre is the next big frontier in SEPs

Last month, Sisvel appointed Willy Chang as Head of AI Pools. In this newly created role, he is charged with driving the development of programmes relating to AI infrastructure and high-performance computing (HPC) technologies. These include the Ultra Ethernet pool envisioned in a call for patents published in April, and related efforts that have yet to be made public.

Prior to joining Sisvel, Chang was Vice-President, Global Strategy and Growth at Via Licensing. Earlier in his career, he spent five years at Qualcomm as a Senior Director, legal counsel, and, before that, was a senior legal counsel at TSMC.

In this conversation with Sisvel Insights, Chang discusses his career change from litigation to licensing, the keys to pool formation in fast growing fields and why data centres are the new smartphones.

Willy Chang, Head of AI Pools, Sisvel

Tell us about your path into the IP world.

I doubled-majored in computer science and history, so while I really enjoyed tech and wanted to go into that field, I was equally drawn to the humanities. That led me towards law school and from there into intellectual property. I followed that path and ended up practising patent litigation – first at Fish & Richardson and later at Baker & Hostetler.

Your first in-house role was with TSMC in 2014 – how did that come about?

I was just trying to do something a little bit unconventional. Today, everyone knows TSMC and it’s one of the 10 largest companies in the world. Back in 2014, it was far less prominent even though it was already at the leading edge of chipmaking. It was an exciting move, quite different from what my peers were doing.

Working at TSMC was a great experience because it exposed me to the entire industry and placed me well ahead of the curve on technology trends. I was working there during the first big NVIDIA boom, when GPUs were being used to mine Bitcoin. By the time I left in 2019, we were beginning to see the first signs of a new use case for GPUs: AI model training.

You then joined Qualcomm – what was behind that move?

I had wanted to move into the licensing side for a long time, and I am grateful to Qualcomm for giving me that opportunity. Most of my experience to that point was in litigation, but I had a strong track record of working closely with companies across Asia, particularly in China. So, I got the chance to do a lot of work supporting Qualcomm’s cellular licensing negotiations in greater China.

Why did you want to make the jump from litigation to licensing?

It was clear to me that litigation is first and foremost a tool for obtaining favourable terms in a licensing agreement. The commercial negotiation is where the highest stakes lie, so that’s where I wanted to be.

What is the biggest difference between being a litigator and a licensor?

The licensing world is much smaller; everyone knows each other and works together closely. Licensing executives on opposite sides of the table are always looking for common ground, and often they can maintain positive personal relationships over the long term. You do not always see that dynamic in the litigation world, which is much more adversarial.

What drew you to the pool space, first with Via LA and now with Sisvel?

The pool model has stood the test of time because it is an efficient, pragmatic approach – and it truly does work for both licensors and implementers. I know this is a well-worn argument, but I genuinely believe it is important to reduce the friction of implementing standardised technologies while simultaneously ensuring just compensation for patent holders. That does not always happen in bilateral licensing negotiations because in almost every case, one party has leverage over the other. I think pools and aggregation are the better way to do things.

You have been looking at patent pool opportunities in AI for several years now. Why has infrastructure emerged as the key focus?

When LLMs first captured public attention, many licensors were initially focused on the AI models themselves and the software and application layer. But much of the financial value has been created by chipmakers and other companies on the hardware side. There are many different components that make a data centre work, and standards play a huge role in bringing all these parts together and ensuring that they are interoperable. That is what makes AI infrastructure such a strong opportunity for patent pools.

The licensing market was somewhat slow to appreciate this at first, but the memory shortage changed that almost overnight. The industry now recognises how much of the value of AI is tied to unsung components, and the licensing ecosystem is far more attuned to the importance of AI infrastructure. The positive response to Sisvel’s Ultra Ethernet pool effort is evidence of that shift. High performance networking is increasingly understood to be a critical bottleneck or enabler for AI – when the network underperforms, the entire data centre underperforms.

You joined Sisvel in June as head of AI licensing. What made you feel Sisvel was the right place to pursue AI pool-building efforts?

Many factors attracted me to the role. I wanted to continue working on AI networking in particular and Sisvel has a strong track record in networking. Sisvel’s success in Wi-Fi made it clear that the firm had already built technical expertise in similar technologies; and the leadership understands the value of the networking associated with AI infrastructure.

Why do you think patent pools are the right licensing model for this relatively new market?

It is clear that pools are the best model. You can envision the data centre as a product category, almost like a smartphone. In fact, data centre spending far exceeds that of the worldwide smartphone market. One of the biggest licensing challenges around smartphones is fragmentation – the market developed through extensive bilateral licensing, and a pool solution never fully took hold. We can avoid those issues by establishing pools at an early stage in four or five technology areas relevant to the data centre. Implementers in this space will have an opportunity that smartphone makers never had: to address SEPs upfront through an aggregated solution.

Who are the patent holders of the AI-related technologies you’re looking at? Have they traditionally engaged in licensing or participated in pools?

There is considerable overlap with companies that are already active in patent pools, but there will be new players as well. For example, some of the semiconductor firms that do not work on codecs or cellular lack experience as pool members or licensees. At the same time, some of the top innovators in cellular and Wi-Fi also hold patents related to wired networking, optics and related fields. Many founding licensors in these new programmes will be patent owners with a long track record of pool participation.

What’s the process like as you go about creating new programmes in this space?

Much of the engagement with licensors involves identifying new areas and prioritising which ones to enter. There are many technologies and a lot of standards, and you do not want a single pool to include so much that it becomes unwieldy. The aim is to secure implementers the rights they need and we will have to take a lot of hard decisions to achieve that. We are also engaging with implementers who understand the risks in this area and want to see pool solutions emerge.

What’s the key to pool formation in a new and fast-growing field such as AI?

The biggest difference between these pools and others is the importance of speed. Pools in the cellular or codec sector often spend several years in formation, but that simply will not work here. These technologies are developing much faster than wireless or video standards. We want to take these pools from conception to launch in less than a year – ideally within nine months – to maximise their impact.

That means we do not necessarily need a large group of founding licensors for these programmes. We need four or five founders that are aligned and can launch the pool in a timely manner. We are also ensuring that licensees are involved from an early stage. If a pool can launch with founding or early licensees, it will attract additional patent owners over time.

What’s the big-picture opportunity here for both licensors and licensees?

Imagine you could go back 20 to 30 years and create SEP licensing solutions for smartphones before anyone else had tried to license that sector. You might establish separate pools for cellular, Wi-Fi, codecs, charging, and so on. Imagine the friction and transactional costs that could have been saved over the ensuing decades. That is the unique generational opportunity we have with the data centre. I am not sure that has happened before in the history of licensing, and I think Sisvel is in the best position to build out this ecosystem successfully.

Rapid-fire questions…

What’s a typical weekend activity for you?

Usually shuttling kids between activities, but I try to play a round of golf at my home course at Green Hills in Millbrae.

What should someone with a free day in San Francisco make sure they see, eat and do?

Start and grab some breakfast or lunch in Chinatown. It’s my favourite Chinatown in the US, and then just head north and just spend the day on a wine train. Can’t beat it.

Where is your favourite holiday destination?

Going to Taiwan, it’s my home away from home. I lived there for years when I worked at TSMC, and visiting it, even briefly, always brings back a ton of memories.

What’s the most memorable live performance you’ve witnessed?

The 2012 Sugar Bowl, Michigan v. Virginia Tech. I think there’s just something about being in New Orleans and watching football when you’re young. Definitely my favourite event—and after party—of all time.

What’s your favourite TV series?

I haven’t watched TV in years! Probably one of the Law and Order series, I guess.

What’s the best piece of professional advice you’ve received?

“You are good at what you do.” Probably the first 10 years of my career I struggled with self-doubt that I felt that I wasn’t good enough, either as an associate or as inhouse counsel. Dan Goettle and Mike Shen really were the two that instilled the confidence in me to allow me to step into my own.