CASPA: Innovation with Chinese Characteristics

A couple of times per year, the Chinese American Semiconductor Professionals Association (CASPA) holds a meeting. Despite the name, you don't have to be Chinese to go, and all the presentations are in English (although sometimes there are slides with some Chinese that slip in). The most recent one was on July 15th (yeah, a Sunday, but at least it didn't start until the World Cup Final was over).

It is surprising that CASPA can put on such an interesting afternoon, pretty much limiting itself to Silicon Valley people, and normally with a bias towards Chinese "semiconductor professionals". This time it was in one of the Intel (old Altera) buildings just off Zanker. The presenters are on the graphic below. As you can see, not that much bias towards Chinese this time, with an Indian, a Pakistani, a Brit, and a couple of American white guys, as well as 5 Chinese.

At the end of the day, Craig Johnson presented Cadence Cloud. I won't cover that since it went over the same material that I already covered in my post Cadence Cloud on the day of the announcement. In fact, I won't cover all the presentations, just pick and choose from some of the most interesting stuff.

Ajit Manocha, the CEO of SEMI (and the Indian) kicked off the afternoon. If he was a politician, he could say that for years semiconductor was a sleepy mature industry, but since he's been CEO of SEMI, it has gone on a growth spurt that shows no sign of stopping. Machine learning has made semiconductor a fast growing, exciting industry again. It hit $400B for the first time last year, and is on track to $500B, perhaps as soon as next year. Ajit is convinced it is on track for a trillion dollar industry by 2025 or so.

The keynote speaker for the day was Lung Shu, whose boss is Ajjit, I suppose. He runs SEMI China. He talked about what is going on in China in semiconductor and put it into a global context. I'll cover what he talked about in its own post next week. There was also a panel session at the end of the day with most of the presenters, but I had another commitment so I can't report on what was said.

Naveed Sherwani

Naveed is the CEO of SiFive (and the Pakistani). He talked on The Dramatic Impact of RISC-V on the Speed of Innovation. I assume you already know RISC-V, if not then the best place to start is my blog post RISC-V Workshop, Milpitas from last December. Naveed talked a little about Chisel too, which I covered at CDNLive EMEA this spring.in my post Agile Development of Custom Hardware.

Naveed pointed out that software companes can get to a minimum viable product (MVP) in about 3 months, building on top of a stack of open-source libraries. For chip design, that is hard for a number of reasons, a big one being that you need too many experts in different domains.

One of the goals that SiFive has is to reduce the cost for a RISC-V in a non-leading edge process like 180nm from several million to a few hundred thousand dollars. Chisel is very productive for the things that it does, and it has a broad library of components (and growing) that can be reused easily. Of course, one group of components are SiFive's RISC-V processor cores, which are all designed in Chisel, and so can both be dropped into other designs or modified to produce derivative designs.

Naveed also said that they are working on a JV in China to address some pre-selected verticals. The idea is to address those verticals even more efficiently so that they can tape out a chip in one of the verticals where all the libraries exist within a month.

Stuart Ching

Stuart is a VP at KULR Technology (and the Brit, not sure how he has a Chinese surname). KULR is pronounced "cooler" because what they do is thermal management. He titled his presentation High Energy Lithium Batteries: Friend or Foe?

KULR's background is in the aerospace field, but now there is much more widespread interest since lithium batteries are everywhere (especially in cars). As he put it:

It’s a great time to be a thermal engineer…everything we do generates a lot of heat.

Lithium-ion batteries are much more dangerous that people realize. The problem is that lithium loves the heat, and the hotter they get the more likely they are to experience thermal runaway, and either explode or catch fire. This is especially a problem with big battery packs like those in cars, since if thermal runaway is mishandled, then one battery can explode and the heat eventually causes the next one to go into thermal runaway, and the whole battery pack will eventually burn up but will take several days to do so.

Oh, and he had a PSA about battery fires, should you be unlucky enough to encounter one. Most of us saw during Chemistry lessons when lithium (and sodium, potassium etc) were dropped in water and reacted strongly with it. So there is a reluctance to use a water-based fire extinguisher on a lithium battery fire. But there is actually very little free lithium in a battery, and water is the best way to extinguish a battery fire. Very few people know this.

Talking of battery fires, Stuart showed a video of NASA's robot humanoid having a battery failure. This is thermal runaway during charging. The battery is not punctured or mechanically damaged. In fact, the NASA people just left it to charge and went to lunch. Oh, and you will see that one of the groups of people who doesn't know that water is the best way to extinguish a battery fire is...the fire department at NASA.

So the good things about lithium batteries are that they have high energy density, high watt-hours/kilogram, high power (low internal resistance), and cheap. That's the Dr Jekyll side, the nice side.

On the other hand, there is Mr Hyde. Thermal runaway can be caused by:

• internal short
• manufacturing defect
• overcharging
• heating
• physical abuse, such as puncturing

The question, Stuart emphasized, is not if a battery is going to runaway someday, but when, and what are you going to do about it to contain it.

What KULR do about it, is build on work they originally did with NASA, to create what is called a "passive vaporizing heat sink". This goes between the individual cells of a big battery pack. The idea is not to stop a single cell going into thermal runaway, that is the if/when thing, eventually one will. The idea is to contain it so that there is no chain reaction of one cell kicking its neighbor into thermal runaway too. NASA spacecraft mandate use of this type of heatsink, since losing a single cell is a minor problem, losing the battery pack is the end of the mission.

How does this relate to China? Stuart asked himself. China is clearly leading the race. Tesla's GigaFactory in Nevada is small compared to some Chinese battery factories. If you want to be in batteries, you have to be in China.

Big battery packs in cars have active cooling, but to make the whole battery system fail-safe there really needs to be passive too. 

 Mark Ding

Mark is CEO of SITRI and talked about Innovate and Win at China IC Industry. In the past he had Lung Chu's job as President of SEMI China. SITRI is the Shanghai Industrial μTechnology Research Institute. They are a sort of incubator for new businesses, as will become clear below.

He started by pointing out that businesses require two types of funding, what he called funding to get across the entry barrier, and then funding to achieve critical mass. Too often, there is the initial funding to get to market, but not enough funding to "get traction" as venture-capitalists like to say.

I don't know if he had anyone in mind, or was just talking generally, but he pointed out that it is $2-3B to build a non-leading-edge fab. But then they forget that is just to get over the entry barrier. It requires continuing investment to achieve critical mass. If you've ever worked in a semiconductor company that has a fab, then you know that the #1 business imperative is to fill the fab. Most of the cost of a fab is depreciation, and that is the same whether the fab runs wafers or not. In fact, in modern big fabs, running wafers is such a small incremental cost that they run wafers all the time, just gathering data if nobody wants the end-product.

Mark went on to say:

There are many cases where small investment crossed the entry barrier, but then there was no continuing investment. Many companies disappeared after a few years. Many government people don’t understand this, and the projects they fund don't achieve critical mass.

Mark went on togive his breakdown of innovation into four categories, putting them in the context of China:

1. Geographical innovation: for example, Chinese internet companies are basically just copying the US,or a successful JianBing (a type of pancake) restaurant in New York is copying JianBing stores in China.
2. Cost innovation: most of China's IC companies are in this category. It is low risk, but also low margin by definition, and often companies end up with just one product and no money pump to fund a second.
3. Revolutionary innovation: quantum computing, AI, superconductivity. High risk and a requirement to build an industry cluster to support it. Plus the problem that "we don't know when it is coming".
4. Evolutionary innovation: most successful companies fall here. Low risk, advanced (but well understood) technology with higher margin.

The three ingredients for successful innovation are Platform, Talent, and Capital. There are three types of platforms (Mark is really into taxonomy):

• Platform companies like Apple, Google or Intel
• Industry R&D clusters like Albany NY, CEA LETI Grenoble, imec, ITRI Taiwan
• Industry clusters

Mark gave two examples of what he meant by industry clusters (as opposed to R&D clusters): silicon valley and Shenzhen (a city in China literally just over the bridge from Hong Kong). See the two slides above for details.

Mark wrapped up with his commercial fo SITRI. They are focused on More than Moore technology, 3D packaging, wafer bonding, and the like. Their business model is to provide almost everything in a sort of one-stop-shop: manufacturing technology, process technology, marketing, CAPEX, investment, biz dev, exit strategy. You just have to have the idea and do the product development.

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.