- 3 days ago
AI chips are getting more powerful — and hotter. Malaysian start-up nanoSkunkWorkX is developing technology for advanced chip packaging that could help tackle the heat challenge. But this is also a story about what happens when science leaves the lab. Can more Malaysian deep tech make it into industry and into the AI chip value chain?
Category
🗞
NewsTranscript
00:12As AI chips become more powerful, managing all that heat is becoming a bigger challenge.
00:18A Malaysian deep tech company, Nanoskunk Works, believes that it may have part of the answer.
00:24They have come up with a technology designed to help AI chips manage heat and power more efficiently.
00:31And the company has just recently raised $2 million to take that technology closer to commercialization.
00:37Joining me today are the founders of Nanoskunk Works, Dr. Amani Salim and Iqbal Shamsul.
00:43Thank you so much for coming to the show. Really excited to have you here.
00:46Likewise, thank you.
00:47So I want to start with the problem because we hear constantly about the raise for more powerful AI chips,
00:54but perhaps less about the physical limits that come with all that computing power.
00:59So that is a problem that both of you are trying to solve.
01:01Let me start with that. Dr. Amani, tell us more.
01:04Sure. As you know, as you can see now, for example, AI is actually pushing the boundaries
01:13of all of these materials and hardware that is involved in compute.
01:17Okay. And what we're doing is that because, you know, you're doing the compute is getting
01:22much more faster and demanding. Therefore, you know, electricity has to, you know, move faster
01:31and heat is generated, you know, thermal management becomes an issue.
01:34So what we're solving is actually at the fundamental level, which is of the materials
01:43and interface of materials that actually conducts electricity and therefore how, where heat
01:49is being generated. So this is the problem that Nanoskunk Works is actually solving with
01:55our technology.
01:56Okay. Let's go further into that because the material that you're referring to is at
02:03the baseline is the copper, right? That which we use at conducting heat. What does your technology
02:10allow it to do to make it more efficient?
02:14Sure. So if you look into copper, for example, is used a lot in interconnects, you know, for microchips,
02:23advanced packaging and so forth. And devices are become smaller and smaller, wires are becoming,
02:29you know, smaller and smaller, packed into small spaces. And because of that, you know, as dimensions
02:36getting smaller, you hit the limitation of these materials because things get amplified when things
02:41become smaller. For example, if there's surface roughness, if there's defects, for example, you know,
02:46the problem of copper conducting electricity and generation of heat become amplified.
02:51What we do is that we're using our technology, we have a process, room temperature process
02:58that's called SUSAN, where we engineered the copper as its formation. So we control how these
03:05materials are being formed at the initiation of the materials so that you can actually engineer
03:11and improve how these copper materials actually conducts heat, signal, you know, electricity, how
03:20it moves throughout this material. This is what we are doing with our technology.
03:23Let me bring Iqbal in. How does that actually translate into a real AI chip? Help us visualize
03:28this. Okay. Well, we used to think about Moore's law, right? You're cramming more and more
03:34transistors. The basic building blocks of all chips, transistors, but you're, you know, it's tiny little
03:40circuits, wires essentially, right? And when you double the amount of compute, for example, that your data and the processing
03:47that you're moving through these microchips, the brains, if you will, the physics of it, when you double the amount
03:54of current
03:54and electrons, you're quadrupling. So if the number of processing goes two, the amount of heat goes up four times.
04:02It's a nonlinear effect, which is why we have all these complaints all over the world, right? The amount of
04:07water data
04:08center uses, the amount of heat that it generates, and of course, the amount of power they simply consume
04:14on top of limited and scarce land, for example. The copper wires where these data, these signals and how power
04:21is delivered to the chip is carried across using copper wires. This has been true for the last 30, 40
04:26years.
04:27And that's the point where we intervene. Okay. So based on my understanding, this is your technology,
04:33at least claim to say that it can move heat across the surface at more than twice the rate of
04:39a standard
04:40copper, the copper baseline, right? Help us understand how does that work? And has that been commercialized?
04:52Where are you at the moment? And what are the gaps that still exist?
04:55Sure. So currently, we have prototypes that we've done in labs, okay, that we basically test, you know, in formats
05:08where you can compare at least with industry. So we've seen all these amazing effects, you know, for example,
05:16moving heat twice as fast compared to copper that we engineered in-house. So of course, we have a baseline.
05:23Of course, we're looking into literature and industry standards and so forth. Our next stage is actually
05:28bringing this into industry validation. See, this is the most exciting and also very important
05:36and challenging step. Because you can prove something in our little startup in our place.
05:43But having industry to adopt it requires it to be scaled, you know, to be able to just test it
05:48in the right environment
05:50where this phenomenon actually has to be tested and validated by another independent parties, which is industry.
05:58So this is where we are at right now.
05:59So this is where it's very exciting because your company recently raised $2 million to move this closer to commercialization
06:05from a regional funder, Tinman Capital. Talk to us about what specifically does that capital allow you to prove that
06:14you have not proven yet?
06:16Sure. So picking up from where Dr. Amani left off, right? The lab, there's a scale issue, but there's also,
06:22I would say, comparability issue.
06:26Once you, you know, what Dr. Amani and the team has been able to do was to sort of mimic
06:30some structures which we believe are representative
06:33of what is deployed in actual data centers, integrated into the chips that you might have heard about.
06:39Huawei, Ascend, NVIDIA, Vera Rubin, etc. Of course, these are proprietary designs. These are proprietary architectures and implementations.
06:48And so the next step would be for us to form another interface, not just about materials now, the interface
06:54with the engineering part of industry.
06:56How do we get closer to their implementations? Can we then apply our innovation onto their implementations and test it
07:05in the field, so to say, in a realistic environment like a data center?
07:09Dropping in that technology to existing manufacturing process.
07:12Correct.
07:13Okay.
07:14Help me understand it. How much does the manufacturing of current, let's say, AI chip producer, how much of that
07:21has to change to use your technology?
07:24Well, like we said, we don't replace the copper. We actually enhance the copper.
07:29So it's actually what we're developing is a drop in module where you don't need to do, you know, excessive
07:38retooling.
07:39It's just plug in our modules, you know, on top of what your copper process actually already done.
07:44The reason is because I've already looked into, you know, we've already certainly looked into is that taking technology, a
07:51new technology into the fab process is not an easy feat.
07:54Yes.
07:54Therefore, the engineering and the work is done is to minimize that friction as much as possible, which is, you
08:01know, to develop something that doesn't require retooling, but this drop in module that I just said just now.
08:06Okay. And what do you think would be the, in your opinion, the challenge to get to that?
08:14One of the big challenges I think is structural, if you will, Malaysia's position in the semiconductor value chain has
08:20so far been, up until quite recently, what we call the back end, right?
08:24Yeah.
08:24Assembly testing, outsourced assembly and testing specifically, right?
08:28And so, and this is of course being addressed by the industry themselves with government policy, which is how do
08:34we leapfrog and bring ourselves farther up the value chain?
08:37One through IC design, the other is through advanced packaging, what we've been talking about, and that's also an entry
08:41point for us.
08:43The challenge for immediate challenge, I should say, is that where our technology can make a difference, the clients, the
08:50people who need it, who use it, who will pay for it, are just not in Malaysia.
08:53Okay.
08:53And that's why Tin Man, you know, visionary investors, you know, long friends of ours, you know, that capital is
09:02helpful because it allows us to, I would say, along with Silicon Star, this program with Silicon Catalyst and MTDC
09:10Kazana, allows us to, number one, access international global semiconductor industry, but with some firepower that allows us to sort
09:17of try to meet industry halfway, which is to say, look, you, you know,
09:23for startups or anyone bringing something new, you're always having to meet the market halfway, they're never going to say,
09:28oh yeah, yeah, great, come in.
09:29We hope we can get that, right? But you probably have to sort of work with them and continue to
09:35fund our own work in, to be able to get to a place where they can readily try and test
09:40out.
09:41People are going to always ask for more proof, more proof, more proof.
09:44And, you know, taking the risk of trying something new in the production line.
09:48Correct. And reducing that risk, a lot of it falls on us.
09:51Right.
09:51How do we anticipate their objections? How do we address them before the questions are asked?
09:58So that's kind of where the money is going to. And of course, IP protection, right?
10:01Of course.
10:03What I found really interesting is, of course, Dr. Amani, you're a former NASA scientist, Iqbal, computer scientist, and also
10:10a former Wall Street trader.
10:11So a very unique combination. And you both started the company and you chose to do it in Malaysia, alluding
10:17to your point earlier about how funding could be a bit more challenging in this part of the world.
10:23So why? Why Malaysia? Why not take it over to a place? I mean, you worked in the US, you
10:31as well, where the thriving, innovative community of tech companies are there mostly.
10:38Yeah. So one of the main questions I asked myself, so I, you know, the satellite sports set, I was
10:45a principal investigator, was launched on April 18th, 2014.
10:49It's a while, a while ago. The question I asked myself is that, you know, can we build a deep
10:57tech innovation in Malaysia?
11:01A lot of it come from my curiosity that can we do this, you know? So this is the reason
11:07why I came back to Malaysia.
11:09I start first in academia to basically explore, you know, what is the potential of our talents? What are our
11:17advantages and limitations?
11:19Because you can't start things without knowing your baseline. Right.
11:22So I came to find out, you know, we are not lacking of talents. Malaysia have, you know, all of
11:28this ecosystem, you know, that we need to connect the dots.
11:31So I said, wow, you think, you know, that this thing doesn't, you know, can't be made in Malaysia.
11:36But once you know what Malaysia has, and you connect the dots, and this is where the building starts to
11:41happen.
11:41Okay. So, as you pointed out, we have the talent, we have the R&D grants, and even development of
11:48scientific talent.
11:49But I think maybe research problem is not so much the big issue here, it's the commercialization problem.
11:55Well, I think commercialization is an outcome of the whole ecosystem, right?
12:00It is not simply, you know, everyone likes to make out commercialization as an end goal, right?
12:05Of course, not all research has commercial as an end goal to begin with, right?
12:11And I think even Dr. Amani's case, right, the fundamental sort of the scientific insights that lead to where we
12:17are did not have a commercial goal in mind.
12:20And so I think it's first important to separate the research question versus the commercial question.
12:26It's not, one does not necessarily lead to the other, right?
12:31But what's perhaps more instructive is the research itself, because it was trying to solve certain things,
12:40it had to attack certain fundamental issues about how energy and material sort of interact with each other, right?
12:47That led to these insights, but of course, we have to do this matching, if you will.
12:54Where does the science get married to an industry problem, into an industry pathway, right?
13:01Not all scientists want to do that. Not all science can do that.
13:06So the gaps here are kind of the structural gaps, clearly, you know, in terms of funding, but also in
13:13Malaysia, no one, as I was mentioning earlier, right?
13:16Industry, but for the most part, domestic industry, none of them truly got there, if you will, by being smarter,
13:24sort of scientifically smarter, technology smarter or more innovative, right?
13:27And that's a big mental block, if you will, from people understanding this conversion of knowledge generation into commercial outcomes
13:36or industrial outcomes.
13:37And someone has to take that early risk.
13:40Perhaps my question could be a bit pointed, but who should be doing more if we talk about the environment?
13:46We have the government, venture capital firms, established Malaysian companies, even researchers themselves.
13:52So in your words, like what needs to work better together?
13:57So in my view, Iqbal might have another view, but in my view is that to basically bring something from
14:03zero to one,
14:04the thing from the lab to market is actually an ecosystem building exercise, you know.
14:11For example, you can't expect, say, a young researcher to basically head on doing this without her or him taking
14:21financial risk, career risk, and so forth, you know.
14:25So therefore, for deep tech especially, you need, this is just a long gestation period, so you need patient capital.
14:32And if you look into even things like, you know, a lot of the innovations that happen in the world,
14:38for example, you have government involvement,
14:40which is important at a certain stage of the company, especially in the early stage of the company,
14:45before a private investment or venture investment can come in.
14:49I truly believe that this is a collaboration of the ecosystem between government, between the universities, between the entrepreneurs, between
14:58the VCs.
14:59You know, it has to come together to understand the requirements for the deep tech to get into market.
15:03So this is my view.
15:04Iqbal.
15:05So it's a very multifaceted question, of course, right?
15:10Even when we talk about commercial, right, what drives commercial decisions and what, you know, solutions get traction depends on
15:18the time as well, right?
15:19If we were to talk about AI, when I first studied AI in 1998 at MIT, right, I mean, AI
15:25was a dead field.
15:26People were like, why are you taking AI, right?
15:28And of course, today, very, very different story, right?
15:30In 1998, the internet was the big thing, right?
15:34So one can never tell where the commercial interests are going to be, right?
15:39Even for our company, right, when we got our first sort of external funding, we were solving the COVID problem,
15:45right?
15:46The technology that we had originally developed for glucose.
15:49So before, we thought we were going to go to market as a glucose sensor company.
15:53We developed, we prototyped, we published a technology that allowed us to measure glucose in your saliva.
15:59But before we were able to get funding for that, COVID happened.
16:03So we pivoted.
16:04We thought to ourselves, can we adapt the technology to address the challenge of the time?
16:08Yes.
16:09COVID, diagnostics, so we did, we converted, we pivoted.
16:12And we were successful, perhaps a bit late to address, the vaccines came faster than our technology went to market.
16:18Why we are able to build longevity and sort of resilience in our business model and our companies?
16:23Because the core technology that came from her insights at NASA, we figured out all the different things it could
16:31address.
16:31And depending on what the world needed at the time, so the mission stayed the same.
16:37We need to figure out where the impact can be brought to bear, how we can make a difference.
16:41But depending on what crisis are you facing at the time, what the market needed, what we could get funding
16:47for, we adjusted.
16:49So a country can do that sort of thing that we're trying to do in miniscule here.
16:53We're trying to do in a microscopic level, but as a country, as a society, that's the approach you have
16:58to take.
16:59You have to seed a lot of things, create a portfolio of knowledge and expertise, and then create the demand
17:06for it.
17:06We had to find, oh, today our solution is good for some other thing.
17:11Tomorrow is another thing.
17:12But the country can do this at much bigger scale.
17:15So you have to create these institutions that allow you to sort of, this long relay race.
17:20People talk about value of death and it's funny.
17:23Well, this is where institutions need to come in and plug those gaps.
17:26Other countries have to get it out better.
17:28Clearly there's a role for government because as the underwriter of first and maybe last resort,
17:33but also as a demand generator, right?
17:36Where can technology help with our healthcare system?
17:39Where can technology help with our sovereignty of energy, of food supply, of defence, you know?
17:46Where do we start? Well, you know, leadership perhaps.
17:49Well, I want to just bring it back to semiconductor first because I know your application is across the board.
17:54There are a few areas. I want to get to that a bit later.
17:57But Malaysia already has a major semiconductor manufacturing base.
18:02But if you want to capture more value, advanced packaging, designed, Malaysian-owned IP.
18:10I want to get your insights into how can Malaysia realistically move up the value chain?
18:20Should I start?
18:21Yeah.
18:22I think Malaysia is already responding to that challenge, right?
18:25But it's still early days. I think, you know, our friends in Penang, right?
18:29Coming together to form the, you know, MPAC, right?
18:32The Malaysian Advanced Packaging Consortium with the support of the government.
18:36But even then, right, the industry knows its needs better.
18:41The industry knows the competition better, right?
18:43And I think what's nice here is that not only is the industry taking that first step, government is also
18:49supporting, right?
18:50With funds, et cetera, right? And we're now with policy support.
18:53So that's part of the solution.
18:56And then, of course, the government has the National Semiconductor Strategy focusing on IC design.
19:01Those are also parts of the solution.
19:03But I think we have to also recognize that this is in response to the market needs, et cetera, right?
19:09In a way, it's a reaction.
19:14So, semiconductors, if I were to be very blunt about it, did not get the love that it does today,
19:21even five, ten years ago, right?
19:23And if you look at the history of the Penai Semiconductor Industry, it never got that kind of support the
19:28way the steel industry or the car industry, et cetera.
19:31So, how do we catch up from that?
19:34Is it too late? Time will tell, right?
19:37I think there's a lot of motivated people with the right support and right intentions.
19:42So, that strategy, you know, we might say is already in place, right?
19:47But how do we go farther? There are other parts of the ecosystem as well that we have to sort
19:53of address.
19:54Some of our sort of part suppliers, manufacturers, right, are serving different parts of the value chain from what we
20:01see in Penang.
20:02Our part suppliers are supplying, clearly, lamb research, you know, but they're also getting qualified to ASML, who's not in
20:09Malaysia.
20:10And that spillover of expertise, right? You know, we have staff at TSMC.
20:14You know, some of them have come back, they're building up this in Malaysia, people like us.
20:18So, there's the core strategy and there's the core stakeholders, but there's also the spillover here that we have not
20:25fully capitalized on.
20:28That deserves further consideration.
20:30Dr. Money, catching up, I think that's a key word here.
20:32Yeah, so it's not just enough to, you know, build more factories, manufacturing factories and so forth.
20:40It's also important to actually bringing in something that actually solves the hard problems, the problems of heat.
20:48You know, and this comes from really fundamental work and physics that leads into IP.
20:57Okay, so this is when people talk about patents and so forth, you know.
21:00We generate a lot of patents at university.
21:03Now is the time that, you know, can we look into what Iqbal said about portfolio.
21:07We have portfolio of R&D, patents and so forth.
21:11Choosing the problems to solve is one of the hardest things to actually identify for a scientist like myself, you
21:18know.
21:18So, I think to move up the value chain, as Iqbal said, yes, talent and so forth.
21:23But at the same time, it's actually we must do things that actually solve the hard problems.
21:28And usually that thing is actually not something that's already existing sometimes, most times.
21:34It's actually bringing the zero to one thing, you know, from the R&D.
21:38The deep science and physics inside is what create that IP that creates the moat.
21:44And that moat is actually what we need to own as Malaysians.
21:47Not just taking blueprints from other people and building it, you know, for someone else.
21:54So, just to add is that since we're doing this nanoscunk works, when we're building things from zero to one,
22:00bringing the IP to life, it basically motivates the ecosystem.
22:03The contract manufacturers said to me, you know, wow, this is the first thing that Malaysia IP come to ask
22:09them to build something for them.
22:10This is something that we need to create more.
22:12So, as we do this more, our contract manufacturers, our small SMEs can actually build high value parts and items
22:21in Malaysia.
22:23Do you feel that you see yourself as part of the shift or do you still see yourself as an
22:26outlier of the ecosystem?
22:30Well, it's a little bit of both, right?
22:33A successful or a startup that tries to be successful, that wants to get and does get venture money.
22:41Venture works with the power law model. I'm sure you're familiar, right?
22:44They are always looking for the outlier.
22:46And I think that's one disconnect in Malaysia sometimes, right?
22:49Culturally, we don't like outliers. We don't like people who don't fit into a box.
22:53Here's the problem though, right? If you already have a box for people,
22:57if you have the same starting points and categories, you're going to end up with the same outcomes.
23:00So, if you say, I want a different outcome today, I have to start with a different starting point.
23:06And that means you have to work with, find, cultivate the outliers, which is a thing that our current system
23:12and perhaps society is not geared up for.
23:15So, this is why, one reason why America sometimes is better for this is because they love the outliers.
23:20They like the crazy ones, right? You know, Steve Jobs goes out, right?
23:23Who are the crazy ones?
23:26Who are the crazy ones who think they can change the world?
23:30Yeah, we don't have enough of that. So, that's one part of it, right?
23:33But of course, you start out as outliers, and then perhaps you have a community of outliers, right?
23:41But once that community grows and thrives, suddenly, that's also the ecosystem.
23:46You become a shift.
23:47Exactly, that's the ecosystem.
23:49After a while, right?
23:50Do you share the same view as Ingbal?
23:52Yeah, see, I came from academia, okay?
23:55Of course, you know, academics, we have our own way of doing things and so forth.
23:59But here's the thing, this is also for my friends who are academics, right?
24:05The academia geared you towards a certain path.
24:08You know, you get your promotion, become professor, you know, you get promoted, you build spin-outs, maybe, if you
24:17want to do those type of things.
24:18But here's the thing, the thing for every scientist that they need to ask themselves is that you want to
24:24make a difference with your science.
24:26You want to make the impact with your science.
24:28This is the core thing that you hold.
24:30This is I hold, at least me.
24:32There's multiple paths to make that happen.
24:36And academia is maybe one.
24:38But there are other paths that you can venture into to make impact with your science.
24:43So, for me, I make a deliberate choice.
24:46Academia has offered me so much.
24:48My foundation is in academia.
24:50I'm not discounting that's important.
24:52But then maybe my second path is when I said pivoting every 10 years or something.
24:56My second path is actually, I want to make the greatest impact with my science to humanity and the world,
25:02especially for Malaysia and Malaysians.
25:03And maybe my path is maybe not leaving academia to do that.
25:07You know, it can go hand in hand.
25:09Again, the model doesn't need to fit one.
25:11It's when Ibal said, you know, you don't need to fit into a box.
25:14You can create your own path.
25:15I mean, if you look at your technology, I mean, you're working across semiconductors, hydrogen, diagnostics, biosensors.
25:23These are all completely different technologies based on one science.
25:28One core science.
25:29Core science.
25:29All right. Amazing.
25:30OK, so just to wrap up, if we were to have this conversation again in a year on up, right?
25:36What does success look like to both of you?
25:39Our science is being adopted by the industry, you know, solve the issue of compute, heat, thermal, you know, I
25:49want to see that so that if you look at the issue of data center, you have to use water
25:53to cool the data centers and so forth.
25:55And now our technology can actually, you know, be more forgiving to the environment.
26:02As much as I want to bring my technology to the world, I also want humanity, you know, to strive
26:06on things, you know, that matters, you know, taking care of an environment, taking care of our people so that
26:12people have high quality of life.
26:14You know, whatever it is, technology can be used, whether it is a double edged sword, you can be one
26:19thing and another.
26:19But I do hope that the technology, the science I'm bringing to the world is actually making better for humanity
26:25and the world.
26:25And Malaysia, we have yet to have a track record of deep tech companies, especially scaling it big and, you
26:30know, making a big exit, for instance.
26:32Iqbal, your opinion, what the success would like to you in the year?
26:35Well, that's in the long run, in the medium run, perhaps even, right? So, yes, an exit would be nice.
26:41But, you know, we have investor milestones to meet and I think, which I think is already sort of clear
26:47for us as well, right?
26:47How do we get this technology to a place of industry validation and adoption in a year, right?
26:53The trajectory, I think, is, I would say, fairly well defined at this point.
26:57We want to be able to have a key partner, a tier one global partner, hopefully on the downstream side,
27:03but certainly in the upstream semiconductor manufacturing side, right?
27:07To be able to validate, test and adopt with them and a pathway where this can go to scale with
27:13them and the likes of them.
27:14So, that's our immediate, I would say, objective for the next year, setting us up for the next funding round,
27:19perhaps,
27:20or even towards like a trade sale or a partial exit, right?
27:23Where, you know, our dream is, of course, for the industry to adopt our IP through some kind of a
27:28licensing regime.
27:31Look, we, you know, we know we're small, you know, we have our value, right?
27:34We're small, we're nimble, we can take the risks.
27:36But now, the transfer of that knowledge into industry, that can happen in many ways.
27:41Do we need to be the ones building a million of these Suzannes, for example, across the global fabs?
27:47It may not have to be, right?
27:48So, we're always on the lookout for partners, strategic partners who can, you know, who have their own leverage,
27:53who have their own footprint that marries well with ours, right?
27:56So, you know, that's kind of what we're looking at.
27:59Truly very exciting. And thank you so much to both of you for coming on the show.
28:03And we look forward to speaking to you again.
28:07Thank you, Dr. Manin. Thank you, Yuba.
Comments