“From a fab perspective, you’re not only looking at submicron structures, but then some of these fiber interfaces may be centimeters. You have to be able to inspect and fabricate structures which vary in orders of magnitude.”

Photonic integrated circuits (PICs), which use light instead of electricity to transmit and process data, will enable faster, more efficient communication critical for applications like data centers, telecom, and advanced sensing. GlobalFoundries, a major semiconductor foundry based in Malta, N.Y., is at the forefront of the industry’s PIC work. In this episode of Manufacturing Matters, TECH B2B’s Dan McCarthy and Aaron Hand talk with Vikas Gupta, senior director of product management for silicon photonics, about the GF technology that combines photonics with a 45 nm RF CMOS technology — all on a single piece of silicon. Gupta details the challenges faced in manufacturing electronics and photonics on the same chip, the trickiness of optical interfaces, packaging complexity, thermal mismatch, and more.

Dan McCarthy: [00:00:06] Well, everyone, welcome to another episode of the Manufacturing Matters podcast. We talk about the trends and technology shaping industry, logistics, and other sectors of the global economy. I’m Dan McCarthy of Tech B2B Marketing. And joining me today are my colleague Aaron Hand — say hi, Aaron — and Vikas Gupta, senior director of product management for silicon photonics at GlobalFoundries in Albany, New York. Welcome to the podcast, Vikas. Thanks for joining us today.

 

Vikas Gupta: [00:00:33] Hi there. And by the way, I’m based in California.

 

Dan McCarthy: [00:00:36] Oh you are? Even better. Oh my gosh. If I’d known that, we would have done this a little later in the day.

 

Vikas Gupta: [00:00:40] But no worries, no worries.

 

Dan McCarthy: [00:00:42] I’d like to kick off a little bit by finding out what our guest’s company does, but I think that is somewhat obvious to many people. If it’s not, it’s the largest semiconductor manufacturer in the U.S. But maybe instead of that, tell us a little bit about where GlobalFoundries fits within the semiconductor manufacturing value chain.

 

Vikas Gupta: [00:01:03] Right. So GlobalFoundries, as you mentioned, is a foundry. And so we manufacture chips for in some cases fabless semiconductor companies or in some cases companies that are looking for additional resources with respect to manufacturing. And in about 2018 we sort of pivoted away from leading-edge nodes, single-nanometer nodes, to manufacturing essential chips for the industry. And that constitutes about 80% of the semiconductor market. It’s a fairly large market for GlobalFoundries to play in.

 

Dan McCarthy: [00:01:43] Excellent. And some of that too was to focus a little more on the silicon photonics side.

 

Vikas Gupta: [00:01:49] That is true.

 

Dan McCarthy: [00:01:50] I know that wasn’t a major strategy, but you are a little bit deeper into that I think, or a little ahead, shall we say, of the game than some of your competitors. The GF Fotonix platform, which you launched in 2022, and that’s where we want to focus mostly on today, because I think semiconductor manufacturing is a highly refined manufacturing process. But silicon photonics ups the game, makes it even more complicated. For the sake of our viewers, could you kind of walk us through the idea of what the GF Fotonix platform does specifically and maybe a little bit on photonic integrated circuits in general, also called PICs, for future reference.

 

Vikas Gupta: [00:02:27] So it’s kind of interesting. I’m not sure if your audience knows, but this is sort of the third generation of silicon photonics at GlobalFoundries. So we started working on silicon photonics almost 10 years back, and it was a combination of work that we were doing in Singapore, in our Singapore facility. We still actually have a silicon photonics line in our Singapore facility. And then, with the acquisition of the IBM assets, we got access to an advanced silicon photonics technology, which we called 90WG. Now since then 90WG, end of life to that technology, and that has morphed into what we call our GF Fotonix platform. GF Fotonix runs in a fabbing facility in Malta, New York. And it’s a unique technology in that it combines the features of photonics along with a 45 nanometer RF CMOS glass technology. So it combines both the electronics and photonics on a single piece of silicon. And we’ve intentionally made it very feature-rich and very flexible. The technology has a lot of components that typical photonic designers would need, design components, ranging from — we start off with the data centers, but the application can range from lidar, sensing, communication. And it’s on a 300 millimeter wafer, which is, as I said, in a multi-facility. Also we’ve tried to make, from an enablement perspective, make it look very electronic-like from a design ecosystem perspective. So there is a lot of work that we’ve put in in terms of having a very advanced PDK to allow designers to design to it using tools that are available in the market with respect to electronic design. So there’s a lot of focus on time to market. So there’s a lot of components that we have pre-characterized, which are part of the PDK, that allow customers to design these PICs and get to market very quickly.

 

Dan McCarthy: [00:04:55] So PDK is, for our listeners, a fairly common tool foundries offer to designers so they can design a chip to the manufacturing process. I’m curious: What’s the threshold between designing and manufacturing the electronic side of chips versus photonics? I mean, how much overlap could you draw from? And how much did you have to reinvent?

 

Vikas Gupta: [00:05:17] So I would say 90% of the tooling is very similar between electronics and photonics. And that’s the reason why we are running it in our most advanced factory. And so there’s a significant overlap with respect to manufacturing capability for electronics and photonics. Some of the more unique features with photonics is specifically how do you get light in and off the chip? And I think that’s where all of the differentiation comes in. And that’s to some extent where we see the innovation from the industry. But that is also the reason why these chips are so difficult to manufacture.

 

Aaron Hand: [00:06:00] So I’m going to jump in as somebody with more of a background in standard IC production, which is probably true of a lot of our listeners too. Can you talk about how that translates into that front-end manufacturing process, of the differences that you have to deal with there?

 

Vikas Gupta: [00:06:20] That is correct. So I think for the most part, from a starting material perspective, the first thing is, of course, we use SOI wafers. Silicon on insulator. So that’s sort of the first difference right from the get-go. The processes in the fab with respect to lithography, etch, implantation are very similar with the standard electronics. Now where things start to differentiate is, as I said, about lighting in and light out. Again, you have to create these sort of optical interfaces. The fiber could be through the edge. Our plan of record mechanism is using V grooves, so we actually do a crystallographic etch right along the edge of the chip to provide that interface for photonic ICs to the fiber. And as we get into this whole area of packaged optics, there’s this demand for detachable fiber solutions. And that, again, mandates some unique ways of processing the wafers to get detachable solutions for fiber interfaces. So in general, I would say a significant portion of the electronics processing is used for photonics also.

 

Aaron Hand: [00:07:46] Okay. Well that’s good, but I understand that scaling up to production levels is particularly difficult. Can you talk about some of the challenges that you face that make that scaling difficult?

 

Vikas Gupta: [00:08:01] I would say from a fab perspective, scaling is challenging from the point of view almost like MEMS-like features that you have to create. And from a fab perspective, you’re not only looking at sub-micron structures, but then some of these fiber interfaces may be in centimeters. And so you have to be able to inspect and fabricate structures which vary in orders of magnitude and be able to inspect these. So I think from a manufacturing perspective, I think that’s the biggest challenge. Given the fact that, as I said, it’s third generation, we’ve now resolved those sort of manufacturing challenges. The remaining manufacturing challenges lie sort of outside the walls of the fab with respect to getting packaging done for these photonic ICs. And so how do you create an automated flow where you’re attaching these fibers. And then you add in the complexity of 2.5D or 3D stacking. So I think we’ve resolved a lot of the fab-related issues. We are now working or sort of growing our OSAT ecosystem to address the manufacturability outside the walls of the fab.

 

Dan McCarthy: [00:09:29] And just to step back a little bit, because we’re talking about packaging, which is the back end. Front end is the wafer processing. Obviously when you talk about silicon photonics, you’re still working with a silicon wafer.

 

Vikas Gupta: [00:09:40] That’s correct.

 

Dan McCarthy: [00:09:41] But you’re integrating other materials, which is not necessarily typical of a CMOS fab. You have the semiconductors for generating light, all those components. Silicon doesn’t lay. So I guess my question is: Is that part of your makeup as well? Obviously you’re developing these wafers. You’re layering all this material on it, but they don’t match. It’s not like silicon, where it’s the atomic-level matching of the matrix. Tell me a little bit more about your processes there, because I know that’s very complicated to do.

 

Vikas Gupta: [00:10:15] Right. So there are different ways of generating light that is required for the photonic IC, so we do have a process where we actually can bond a fully formed DFB laser to the photonic IC.

 

Dan McCarthy: [00:10:30] So you bond it. Okay.

 

Vikas Gupta: [00:10:30] There are other companies that have actually done, they’ll place a coupon of an indium phosphide die on the PIC and process it further. We’ve chosen the ability to sort of put a DFB laser on the photonic IC. Now, especially when you start talking about more advanced optics, like co-packaged optics, the fact that these optics are sitting so close to an ASIC, which is a very large source of heat, a lot of companies are looking to move the laser interface off the chip. And so there are different solutions to getting the light generated, to get it onto the photonic IC. And again, Dan, you talked about materials. So a lot of the materials at this point on the photonic IC are silicon based. Of course, the laser or an optical amplifier would be a common material. Now, when you get to the higher and higher speeds, beyond 200 gig per wavelength, when you start seeing that silicon sort of runs out of steam as a modulator. And that’s where things start getting even more interesting, because now you have to start looking into materials such as thin film, lithium niobate or barium titanate or organic polymers. So that’s sort of the next frontier of materials beyond the laser.

 

Dan McCarthy: [00:12:02] Right. And just a quick follow-up on that because your building material starts to grow. And the supply chain becomes much more complicated. These materials are not always necessarily as readily available as silicon. How much has GlobalFoundries adapted to the growing complexity in the supply chain, particularly these days?

 

Vikas Gupta: [00:12:24] Again, there’s been a big thrust from the GlobalFoundry side in terms of establishing the ecosystem. And while we continue to work with external OSATs to make sure that these chips are manufacturable, that there is a flow that is established, there’s also a supply chain where we’re looking into laser, fibers, connectors, etc. So we’re working to establish that ecosystem to make sure that we can manufacture PICs, not only us – our customers can manufacture PICs. And so we’re establishing that external sort of ecosystem. But we’ve also made an announcement with respect to this advanced photonic and packaging center in Fab 8 that allows sort of a U.S.-based footprint of manufacturability of packaging of these photonic ICs.

 

Dan McCarthy: [00:13:22] Interesting.

 

Aaron Hand: [00:13:23] I’d love to get more into the 2.5, 3D packaging aspects, but first I’m just wondering actually about kind of the thermal aspects. You talked about being too close to the ASIC and the heat you get from that. I would think with different materials, factoring in that you might have some thermal mismatch, tell us a bit about what you’re dealing with in terms of thermal constraints.

 

Vikas Gupta: [00:13:51] Yeah. And that’s what makes silicon photonics interesting in terms of innovation, because it’s a multi-physics problem and a multi-material problem. And some of these are not unique to silicon photonics. So, for example, if you have an electrical IC which is sitting on an organic laminate or a PCB, there are different thermal coefficients that are in play. So that’s not unique to silicon photonics. What does become unique for silicon photonics is, for example, the adhesives or the index matching fluids that are attached or placed on the photonic IC to interface with the fiber. Those have to survive not only the reflow temperatures of getting bumping done but then the fact that they’re constantly sitting next to an interface that is generating a lot of heat. So these are the problems that we’ve sort of prefetched as part of our technology qualification. So when we do technology qualification, we are making sure that devices are characterized to the higher temperatures that they would see under mission mode. We’re making sure that as part of our qualification with respect to packaging that we are comprehending the thermal profiles that these photonic ICs will see. And so a lot of that is comprehended as part of the technology offering that GlobalFoundries provides. And towards that, it does require a level of sophistication that we have developed over the years of understanding the system application. So it’s not just the photonic IC. You have to understand where the photonic IC is placed and what are the mechanical limitations, thermal limitations that this photonic IC will see under mission mode. And so that’s sort of the sophistication that we have gained by being the third generation now of silicon photonics.

 

Aaron Hand: [00:15:56] Okay. So I don’t know if it’s too soon to jump into 3D packaging, because actually I think that what you talked about there, Vikas, kind of begs the question of which is a better construction once you go to that package. So 3D packages would stack chips one on top of the other and connect through silicon vias. Whereas two and a half IC packaging would be more side by side? Is that right?

 

Vikas Gupta: [00:16:33] That is correct. That is correct. So, again, when we talk about photonic ICs, when I said feature-rich technology, there’s a lot of attention that we have paid on prefetching the system-level applications. For example, we make sure that our photonic ICs are compatible with copper pillars or copper receive pads. We now offer through silicon vias in our technology, which serve not only as signal propagation for supply currents and as sort of a heat source to be able to dissipate the heat from the electrical IC that is sitting on top. So, yes, certainly there’s sort of the electrical components of the electrical IC or sitting on top of the photonic IC, but there are also these other thermo-mechanical aspects that we need to consider when we start talking about 3D. Now the interesting part – again, none of this is unique to silicon photonics. You know, the electrical ecosystem also has some of the challenges. The thing where it becomes complicated is the fact that now you’re also dealing at the same time with an optical interface.

 

Vikas Gupta: [00:17:57] So how do you deal with pigtails that are hanging off your photonic IC as you’re going through the process, and that creates this unique need for a manufacturing process that comprehends the photonic interface. Now, there’s a lot of work that we’ve done in terms of making the OSAT processes or the packaging processes, wafer, wafer level, and they look more and more like fab processes. So a lot of the copper pads, development that is done is at wafer level. Through silicon vias are generated at wafer level. The reveal is done at wafer level, and so a lot of these OSAT processes are also now starting to look like fab processes. There’s less and less work which is being done at a die level, including, for example, the fiber attach. So we’re coming up with detachable processes, where the first interface is actually created at the wafer level. And so we’re trying to make these processes, as you sort of scale them, it’s more and more important that you do more of this manufacturing or processing at wafer level. And that helps with making the PIC wafers also look more like electrical IC wafers.

 

Dan McCarthy: [00:19:18] That’s interesting. So when we talked about packaging, we talked about thermal management. Touched upon V grooves. My next question’s leading up to alignment. Obviously the packaging typically takes most of the cost from what I understand, in the manufacturing of these chips, and a lot of the bottlenecks have to do with alignments of the optical signal, because obviously it’s not a contact. You have make sure that’s perfectly aligned down at the nanometer level in order to keep that signal strong. So I’m curious, how was that figuring into not only your processes but the industry in general? What are the trends there as far as, and you talked a little bit about embedding alignment on the wafer level itself, but typically today it’s all done with attachment of fiber.

 

Vikas Gupta: [00:20:05] So you will hear the term “active alignment” versus “passive alignment.” So active alignment is where you actually power up the fiber. There’s a light coming, and you’re looking for sort of maximum power transmission. And that would constitute the active alignment. Now the beauty of the V grooves, which is our plan of record, is that it’s all passive alignment. And you create this sort of regroup-based structure that provides this interface that self-aligns the fiber core to the spot size converter. And again, when you’re talking about small-count fibers, it may not be that compelling for customers. But then when you start talking about having to align 30 fibers on the edge of the PIC, the value of the V grooves becomes more and more apparent. Where now you put the full fiber array in the V groove array, and they’re sort of self-aligned across the board. So I think that’s one mechanism. The other mechanism, as I was talking about in terms of detachable solutions, one of the turning mirrors is placed on the photonic IC itself. And we will use fiducial marks to make sure that – so we use high-precision pick-and-place tools to place these turning mirrors on the photonic IC. And again, that is all passive alignment. And the fact that it’s done at wafer level allows for that scaling to occur. That typically has sort of plagued this industry right from the beginning. So there are mechanisms that are being put in place. Again, this is in collaboration with a lot of innovators in the industry, and you can actually sense now that even the OSATs have realized that there is value in this industry. And so they have started participating.

 

Dan McCarthy: [00:21:57] I’m curious what the threshold is there for the level of alignment, even with a V groove. You’re doing pick-and-place. I mean, the pick-and-place machines have a certain amount of accuracy in how they can place. What’s the threshold for alignment that you need to achieve that?

 

Vikas Gupta: [00:22:16] I would say sub 1 micron.

 

Dan McCarthy: [00:22:18] Wow.

 

Vikas Gupta: [00:22:19] And there are tools that are able to do that at pretty high speeds. Again, you will hear the term “UPH,” units per hour. And again, that is a metric that we are all keeping our eye on to make sure that these solutions are manufacturable at scale and at speed.

 

Dan McCarthy: [00:22:42] Excellent.

 

Aaron Hand: [00:22:43] And you bring up the point of the ecosystem, of more people realizing this is worthwhile to get involved with. I’d like to take a look at the design side of that. So I see chip design is pretty well established. How does the ecosystem for PIC design compare?

 

Vikas Gupta: [00:23:04] So I would say it’s on one hand, as I sort of mentioned earlier in this conversation, from GlobalFoundries we’ve tried to make the PIC design as much as an electronic design in the sense that we in our PDK we support layout. We support LVS — that is the layout to schematic verification. So a lot of the tools that have been established as part of the electrical IC ecosystem is what we are leveraging to do the design for the photonic ICs. Now the area where the EDA is immature is sort of the multi-physics problems associated with device simulation. And the fact that you have to comprehend not just the electronics side of things but also the photonic behavior of these devices. So there’s a lot of innovation on that front in terms of these multi-physics solutions, where you have to comprehend thermal aspects, electronic and photonic aspects of things. And that is where I think a lot of the innovation is coming in. And you can see  companies now starting to also promote more electronic-based design for photonic ICs. And the other thing what we’ve done, as I said, is we use industry standard tools like SPECTER-based models for photonic ICs. Sorry, Verilog-based models for our photonic ICs, SPECTER-based models for our electronic components. So we’re trying to use as much of the electronics ecosystem to ease this transition in terms of photonic design.

 

Aaron Hand: [00:24:54] Okay. So how does that translate into advice that you might have for PIC designers in order for them to shorten that cycle time between design and shipping?

 

Vikas Gupta: [00:25:03] Yeah. So I think the PIC design has similar challenges to analog circuit design in the sense of you can provide tools to designers, even analog designers, to simulate circuits, but it does require a level of sophistication from the designer to understand process variations or for that matter wafer-to-wafer variations to be able to create manufacturer designs. So I think it does require somebody who is experienced in photonic design to be able to use these tools effectively and create manufacturer designs. We’ve done everything we can in terms of PCells for these devices, that are models associated with these photonic devices. We also, as part of our documentation, provide details about the model, to hardware correlation. So there is a lot of information that we have provided to the designers to enable them. But the designers do need the experience and sophistication to be able to take advantage of these tools.

 

Dan McCarthy: [00:26:17] Since you brought up multi-physics multiple times, I want to talk about a multi-physics challenge. To your point, there’s your testing for both electrical and optical signals on the chip as you’re going, which is not typically part of IC production. Does that require two processes or are there ways to integrate the testing for both signals on a chip? I’m just curious. Is testing a bottleneck out there?

 

Vikas Gupta: [00:26:47] So testing certainly is — I would say the next frontier of innovation is packaging and testing. We’ve been doing electro-optical testing for a while. And so I’ll just break the testing down into two pieces. The parametric test that occurs at wafer level. So similar to these curved structures that foundries will place in die, the chip-to-chip area, where they will create these structures to be able to characterize device performance at wafer level. GlobalFoundries has a similar methodology, which we call ILOTs, so in-line optical test systems, where we will place devices, photonic devices in a pad case that has grating couplers and an electronic interface. So we’ve been doing this sort of electro-optical testing for a while at GlobalFoundries. I think the next frontier is to be able to do electro-optical functional testing. And whether it’s low speed or at high speed, that is the capability that is sort of being developed. You know, there are companies that have, as recent as the OFC this year, that have made announcements of being able to do both the electro-optical testing at the same time. So I would say the functional testing is still immature, but I think this sort of gap is being observed by several name brand test companies that have now provided solutions that are available.

 

Dan McCarthy: [00:28:25] Interesting.

 

Aaron Hand: [00:28:27] I want to go back to a point you made about what’s required of that PIC designer and really understanding the physics, understanding the process variations. To me, that gets into this whole electrical engineering versus optical engineering. So they really need to know both aspects, right?

 

Vikas Gupta: [00:28:52] That is correct.

 

Aaron Hand: [00:28:53] So how do you deal with that at GlobalFoundries in terms of building a workforce? You know, we hear so many issues with even just the basic technical level of the workforce. So how do you develop that workforce that can handle all of the physics involved there?

 

Vikas Gupta: [00:29:11] Right. So I think some of it is organic. You know, within GF there is certainly a lot of experience that we’ve had on the electronic side. And from the processing side, for example, a lot of our engineers who are working in the fab come from the electronic side, and they have learned as we have worked with customers what things are important to customers. And so we’ve grown our knowledge base based on the fact that we’ve been working on this for so long, but we also have this vibrant ecosystem with working with universities and funding universities as part of our university program where we will choose projects from professors, where we provide them access, and they become that source of the future talent that we want to attract to come and join GlobalFoundries. So some of it is organic growth within GlobalFoundries. But then also this enablement through research institutes, through university programs where we are sort of trying to create this next generation of talent that will go, and of course we would love for them to all join GlobalFoundries. But also there’s value in them seeding the ecosystem to allow this industry to grow.

 

Dan McCarthy: [00:30:40] You’re talking about university programs. Are there degree programs that bridge electric and optical physics?

 

Vikas Gupta: [00:30:48] So there are certainly research programs and degree programs. The other source, in Europe there’s this huge push in terms of university programs and funding that is available. And again, the whole industry benefits from these sort of investments that are made by the governments and by the individual universities. And it just grows the pie. So yeah.

 

Dan McCarthy: [00:31:19] And to your point, partnership is a big part of this, right? Not just within the universities, not just within government, within the industry itself. I know GlobalFoundries is collaborating with some pretty heavy hitters out there, including Nvidia and Cisco, as well as some startups like IRLabs. How critical is that? You can’t go it alone.

 

Vikas Gupta: [00:31:45] So I mentioned earlier this whole, the level of sophistication of understanding the system-level implications of where these photonic ICs actually fit. And I think towards that, it is very important for us to work with some of these leaders in the industry. So even for GF Fotonix,  while we were developing GF Fotonix, we had established collaboration with some key startups that we thought had the sophistication and the understanding of the photonic design, which would allow us to work with them to refine the technology based on what their requirements were. And  so there was this symbiosis in terms of learning on both sides, where we would provide learning from the foundry side, they would provide learning from the design side. And so, yes, this is not possible for a single company to just grow this knowledge base by itself. It does require collaboration. So that’s the other source. And to your point, it’s not just the universities or funding. It’s also working very closely with startups that we’ve done over time that has increased our understanding of what the photonic industry needs.

 

Dan McCarthy: [00:33:09] Is it fairly unique for the photonics side? Does the semiconductor industry — and I guess what I’m thinking is you’re talking about partnering with startups in the semiconductor, conventional semiconductor, side. Do startups contribute that much or is this sort of unique to the photonics element?

 

Vikas Gupta: [00:33:26] I think we’ve leveraged a lot of learning from our colleagues on the RF side, for example. I think they’ve done an excellent job of growing the ecosystem. So again, while a lot of this conversation may seem unique to silicon photonics, what GlobalFoundries as a foundry is trying to leverage is learning that we’ve done on other product lines to sort of see what is applicable to silicon photonics and add in whatever layer that we need that may be unique to photonics. But we are leveraging a lot of work that our colleagues have done in other product lines. So there are other examples of this where we worked with some of the industry leaders and a vibrant startup ecosystem to get that sort of learning. So it’s not unique to silicon photonics. I think our RF friends, product line, has done an excellent job similar to what we are doing.

 

Dan McCarthy: [00:34:23] It takes a global village.

 

Vikas Gupta: [00:34:24] Yeah, exactly, exactly.

 

Aaron Hand: [00:34:28] So I do want to raise one point – Dan mentioned the partnerships like with Nvidia or Cisco – of what this is all about. I think we haven’t really spelled out what this means, what these developments mean for data centers, for example, that are using so much energy today. Can you put this all into context for our viewers?

 

Vikas Gupta: [00:34:56] So this conversaion about the use of photonics. Initially it was all telecom. With the growth of data centers, it sort of started off with how do you get data from the racks to the switch on the top of the rack? And that’s where the initiation of that conversation, specifically I would say about co-packaged optics. That’s where we started talking about, hey, as the data rates increase, you start running into signal integrity problems when you’re transmitting data over copper. And so the solution was, let’s move to photonics because photonics I would say has sort of two unique properties, amongst others. One is the loss of the photonics signal through a fiber is minuscule compared to what you would see in a copper-based transmission system. That’s one thing. And the other beauty of photonics is that you can transmit data on multiple wavelengths and they don’t actually interact with each other. And that’s where you start getting into this wave division multiplexing. So you can actually transmit a lot of data through the same fiber by multiplexing signals at different wavelengths.

 

Vikas Gupta: [00:36:15] So it’s the combination of getting to a lower power mechanism of transmission. So you don’t need re-timers, regenerating circuits, or repeaters to be able to transmit data long distances. And the fact that you can actually slow down in some cases the electrical interface and then multiplex a lot of optical data on the same fiber. So this trend is sort of what is driving this, this adoption of photonics in the data center. So it is a combination of the need for lower power, the need for a lot more data to be transmitted across distances, and the fact, for generative AI, as you start building these larger and larger farms of GPUs, they need to talk to each other with very low latency. And the fact that you don’t need these re-timers in the path of an optical interface. That sort of helps with being able to build larger and larger farms of GPUs together.

 

Dan McCarthy: [00:37:30] Excellent. This is a constantly emerging and evolving sector.

 

Vikas Gupta: [00:37:35] Exactly.

 

Dan McCarthy: [00:37:35] So there’s still a lot of manufacturing challenges that are going to enable that. But progress seems to be accelerating from what I’ve seen. Vikas, I could talk to you all day about this sort of stuff. It’s pretty fascinating. But we know your time is precious, so we don’t want to keep you too much longer. Is there anything we didn’t cover that you feel we should touch on?

 

Vikas Gupta: [00:37:55] No, as you said, I think this is an exciting industry to be in, and certainly GlobalFoundries has been in this industry for a while. As I said, third generation. And the fact that it’s starting to pick up now, I think it is actually gratifying to see all of this effort actually come to fruition. And the industry recognizing this effort.

 

Dan McCarthy: [00:38:22] Absolutely. We appreciate your time. Big thank you to Vikas Gupta and GlobalFoundries for joining us today. If you have any questions, please feel free to post them in the LinkedIn thread and we’ll share them with Vikas. Or you can respond directly. If you’re interested in past episodes of Manufacturing Matters podcast, you can find them on our website, manufacturing-matters.com. And for anyone wishing to be in the loop on upcoming conversations about automation, please follow us on LinkedIn. And thanks for joining us today and thanks again, Vikas. I appreciate your time.

 

Vikas Gupta: [00:38:50] Thank you both. Thank you. Take care.