# 2026 Optica Executive Forum @ OFC Panel:  Breakthrough Technology

https://www.youtube.com/watch?v=9Kn2q_-v-sU

[00:00] Highlights from the Optica Executive Forum at OFC.
[00:02] Thanks to these sponsors, but the data center's ging deep inside.
[00:08] Copper running hot.
[00:11] Hear the signals getting f and cold package light is how we win.
[00:20] We're releasing seven sessions from Optica's executive forum because the world is watching.
[00:25] Feedback encouraged.
[00:29] Now enjoy.
[00:33] Hello everyone and welcome back from lunch.
[00:35] How was lunch?
[00:35] Was it okay?
[00:39] I must admit I didn't eat anything because I was just talking to all of you trying to connect all of you.
[00:43] Many of you told me that this lunch was actually the most business activity they had in a long long time.
[00:51] And for that this is what we're trying to do here.
[00:53] We're trying to achieve new things.
[00:54] One of the biggest things, the biggest feedback that we received last year was you have to bring
[01:00] new technologies, new companies, new kids of the block because the industry is changing.
[01:04] So you ask and we give it to you.
[01:07] The next panel is on new technologies and is moderated by nobody else but Tim Dy.
[01:11] Tim, thank you very much for being with us.
[01:14] The room is waiting for your energy.
[01:15] The floor is yours.
[01:24] Welcome back.
[01:24] Welcome back everybody from lunch.
[01:26] Uh thanks for joining.
[01:30] When you take a look at the journey we've been on since this morning, right,
[01:32] we heard from OpenAI
[01:35] and we heard what's happening um inside the data center and some of the connectivity challenges.
[01:40] We then took a look at scale across
[01:41] and um how does scale across and DCI uh differ and then we also looked at scale out
[01:55] and uh after our lunch break here we're going to keep shifting our perspective
[02:01] and at this case we're going to look at uh three companies who are committed and have committed their time, their energy, their resources to helping solve the scale up and some of the scale out challenges that are there.
[02:18] We'll have um three uh folks come up and present one after the other and then we'll come back for a moderated uh Q&A session.
[02:26] Certainly welcome your questions as part of that.
[02:29] I have some that I've developed as I uh saw their presentations and the rest.
[02:36] So we'll start with uh Nick Nick Harris from Light Matter.
[02:39] We'll go to Marco Casadi from uh Aviscina and then Hamid Arabzad from uh Ranavas.
[02:47] Welcome aboard, Nick.
[02:52] Okay.
[02:53] Well, I'm Nick Harris, co-founder CEO of Light Matter.
[02:55] Uh we're a spin out of MIT almost nine years ago now.
[02:58] Um and I'm excited to be here.
[03:00] I think this is the most exciting OFC event of all time.
[03:03] Uh a little bit about the company.
[03:06] Uh so founded in 2017, we've raised about 850 million.
[03:11] Uh we've got great investors headquartered in Mountain View, 200 people there, 100 in Boston.
[03:15] Uh we opened an office in Sinshu in Taiwan to be near TSMC and some of our ODM partners uh recently.
[03:23] And uh we have a team in Toronto as well.
[03:25] We're a little over 300 employees and we've invented a lot of great technologies that I'm excited to tell you about today.
[03:32] The central challenge that we're solving at light matter is this gap between the rate of progress in computing and the rate of progress in memory and interconnect.
[03:43] Uh so if you look over the past 30 years compute has a millionxed 10 millionx on some scales.
[03:49] Now that accounts for precision changes and other things like this but call it a millionx.
[03:53] Memory and interconnect have progressed at a much much slower rate.
[03:58] Uh this gap results in low utilization and at this point it's limiting our ability to continue scaling performance on AI and
[04:04] that's why this is such an exciting OFC because that moment is here.
[04:09] Um the last thousandx in computing performance definitely has come from optimizing the compute and the next thousandx is 100% going to come from fetonics scale up networking and scale out networking.
[04:21] The little red dot there is our M1000 chip that was released last year.
[04:25] Uh and it was at supercomputing 25 conference and our customers have been looking at that platform for quite a while now.
[04:33] So the big challenge uh in the space of computing I always look at this whole area not just as an interconnect technology but as a computing technology is that computer chips are getting bigger and the compute elements live in the area.
[04:44] The area is growing as length squared, but the perimeter where the IO happens grows with the length.
[04:51] There's the fundamental mismatch between the fact that computers scale now through area and through big packages because of the end of Moors law and the requirements for IO.
[05:00] We're building giant machines that can't really talk very quickly but are very smart.
[05:06] So, Fetonix is the solution to this challenge and I'm going to walk you through the road map as we see it and the timeout timing for the go to market.
[05:13] Uh right now the world is dominated by pluggables something like 17 to 20 picoles per bit.
[05:17] These are optical engines that are 18 inches away from the processor from the switch.
[05:23] Um we just announced uh a couple days ago our L20 near package optics engine.
[05:28] Maybe a surprise for people that light matter would release a near package optics solution.
[05:32] Uh but L20 is a birectional very high energy efficiency solution.
[05:37] It's actually 4.7 picjles per bit, 224 Gbit per second by per fiber, 6.4t per optical engine.
[05:46] 27 is going to be the big year for near package optics and there was this XPO uh consortium that we joined.
[05:51] Uh there's a new MSA there.
[05:53] We think that's going to be a big deal.
[05:55] It's a new form of pluggable and it's something that L20 fits into nicely.
[06:00] In 2028, we see TSMC coupe technology being really dominant uh for fetonics.
[06:03] uh we're designing a bunch of chips in that
[06:08] space right now and it's going to be a big improvement both in terms of energy efficiency but also total bandwidth per package and then unique to light matter our interposer technology that we invented in 2018 hundreds of patents uh 3D CPO interposer it's here at the conference at OFC it was at supercomputing last year and our customers have been seeing it for quite a while two pajles per bit 114T in a single optical engine so this is the road map for optics And light matter does everything from NPO all the way to interposer and you can come to our labs our validation data center and see the racks anytime.
[06:45] This is what they actually look like.
[06:47] Uh so on the left we have the near package optics engine.
[06:48] That's what L20 looks like.
[06:50] In the middle we have CPO.
[06:52] This would be TSMC coupe largely also Global Foundaries.
[06:57] And then on the right are M series fatonic interposers.
[06:59] So NPO and CPO from light matter are called L series and interposers are called M series.
[07:03] Earlier this year in January, we launched our new product line, which are
[07:09] Our laser technology called Guide.
[07:11] The challenge that we ran into is our optical engine is 70 times faster than a 1.6T pluggable.
[07:13] That means it needs 70 times more light.
[07:18] And so we've invented a new kind of laser and it's sampling with customers today.
[07:21] And we couldn't be more excited about this technology platform.
[07:24] We're calling it very large scale photonics.
[07:25] Hundreds and hundreds of components per chip.
[07:29] We're taking laser diodes from a single discrete element to integrated circuits with hundreds of lasers per chip.
[07:41] So we have three uh generations of hardware in our lab sampling today and you can check out our website lightmatter.co.
[07:46] We just had an update and you can check out the platforms, check out their specifications and you know if you're a customer you can come check them out in the lab or a partner.
[07:55] Um we have passage m1000 our passage 50 EVK.
[07:58] You probably heard about this OCI MSA passage 50 is 800 gigabit per second, 400 gig each direction, 16 lambda by at 56 gig.
[08:04] So we saw what was
[08:10] coming with OCI several years ago and we built that platform and it's been with customers uh you know for a while now.
[08:16] and we're designing multiple generations of chip on that kind of architecture and you know not to outdo ourselves but Passage 100 EVK we just did a press release with Qualcomm.
[08:25] This is 1,600 gigabit per second per fiber.
[08:28] 16 lambda at 112 gigabit per second PAM 4 per optical fiber.
[08:33] So huge advancements here.
[08:36] And we're not thinking about things just at the chip level.
[08:38] We're looking at rack scale.
[08:39] Validation is the frontier.
[08:41] Building these systems at large scale, proving firmware reliability, link flaps, all of these details associated with deployment is the frontier and it's all enabled by MRM technology.
[08:51] Um these are some of the partners that we work with uh on photonic integrated circuits EIC's and our packaging partners.
[08:58] We just had some great announcement with ASSE where we showed a new advanced packaging compatible detachable coupler.
[09:04] You can see a render of what a detachable coupler looks like on the slide there.
[09:07] Uh we're very excited about it.
[09:09] What it
[09:11] means is that coals sike or silicon substrate type technology solutions can now have fetonics on board.
[09:17] It can go through mold and grind and it comes out the other side with a very very low loss detachable coupler.
[09:23] Go look at our booth uh at OFC and you can see the hardware.
[09:28] Uh you can see both M1000 with detachable solutions as well as our uh our chip platform for this V-click technology.
[09:35] Very very low loss and essential for the future of Phatonix.
[09:38] And that's it.
[09:41] I'm excited to be here.
[09:43] Thank you.
[09:49] Good morning everybody.
[09:51] So my name is Marco Kizery.
[09:53] My story with a visa start in 2002 2022.
[09:56] At the end of 2022 I invested um in my previous uh role in my previous company um leading series A into the company.
[10:05] Um I've been on the board for two years in Dulf and about six months ago I joined as a CEO.
[10:07] But the story of Avisa really starts
[10:11] About five years ago.
[10:15] The group of founder um invented really um or or came with a very bold and innovative idea.
[10:23] The idea was to use micro to replace laser for high speed interconnect and of course um leverage this technology for a very special application for scaling and and scale and scale up.
[10:38] Today I'm pleased and really proud to announce that Abbyenna is announcing here at OFC the world first microd interconnect evaluation kit.
[10:46] Uh in this kit we will have 300 channel at two 2.5 gigabit per channel um at over half alpha terabit and very soon this this very same kit will host micro at 3 3.5 Gbit per channel for a total capacity throughput of a terra a terabit per second.
[11:06] Seeing this technology becoming real, of
[11:13] course, has been a huge testament of what the team has done.
[11:15] And the team has been able to do this in a very short period of time with $70 million.
[11:20] And the team today is 100 people strong, 30 PhD strong, but it's really still a small team which really says a lot about the potential of this technology.
[11:29] And in potential technology, the goal is technology are really very simple.
[11:37] The goal is to simplify the system.
[11:39] The goal is to achieve uh lower power, lower cost and uh and overall simplification of the system replacing laser.
[11:54] I want to spend just a few minutes to give a sense of how the technology works um and how really what really the fundamental component of this technology and how much simpler compared to laser system this technology really is there really five fundamental components um there is the micro um the micro really
[12:15] it's it's the innovation that we bring
[12:17] to that we bring to market these LED are
[12:20] as you know capable of switching 4 8 GB
[12:23] per second
[12:25] um the the photo diode um and our in
[12:30] single photo avalanche photo diode
[12:32] microl lenses that we 3D stamp directly
[12:35] on the um on on the arrays
[12:39] the fiber bundle and here every channel
[12:43] of course has its own bundle so you know
[12:45] as I as I was saying as as you will see
[12:48] in a minute it's still a simplification
[12:50] of um modulation that we in the channel
[12:53] as well. And then of course the ASIC for
[12:57] the micro driver control and the uh
[13:00] receiving part.
[13:05] From a technology perspective, the key
[13:07] message here is that uh we the
[13:10] technology allow us to achieve a very
[13:12] significant throughput and density in a
[13:15] very small
[13:17] um micro emitting area.
[13:20] In the chip that we're presenting here at OFC um the one at Alfa and soon hopefully at full terra.
[13:28] the micro is just about just a little bit smaller than 1 millm square and the message here is that the density will be roughly one 1.6 terab per millimeter square of emitting micro area.
[13:44] Here you have a a picture of um our micro under the microscope and this micro are entirely manufactured by Avisa.
[13:53] Avisa is unique in the sense that as a startup we have a nano fab a 35 gallium nitrite fab on which we grow our gallion nitrite on safire and uh the entire system in it's entirely packaged entirely assembled entirely bonded in our facility in a
[14:18] Um, so what are the applications we're chasing?
[14:22] Well, I I know we talk a lot about this morning about scale up, scale up, and scale out, but the first of course and most promising application of this technology is actually scale in CPU for die to die and chip to chip interconnect.
[14:36] So computer to compute and compute to memory, anything from few millimeter to a meter interconnect within the board.
[14:49] Um, and then of course the second application is the scalab, the connectivity, um, um, of of computer to computer switch up to 20, 25, 30 m for this scaling.
[15:03] Uh, I really, we really believe the technology is going to be significantly disruptive starting of course from the reliability, no laser, um, to begin with, but then intrinsically because we can afford so many LEDs.
[15:20] Uh, the promising technology is actually to use a lot of spare LED channel and use the spare channel to replace LEDs that are, um, that are not working.
[15:33] So the optical engine and the system in, in itself is significantly simpler, um, than any laser solution that, uh, um, you, you have seen for chip to chip interconnect.
[15:45] Of course, in general, the promise is a significant reduction in power and the reduction in power comes also at system level.
[15:54] We can transmit wide and slow, and, uh, we can transmit pretty much the natural speed of the IO's of the, of the chip, which basically means we really don't need for chip to, for die to die interconnect series.
[16:10] And we really don't need any complex, uh, high order modulation P four or even more complex, um, so the advantage really is to transmit.
[16:22] into the fiber one channel per fiber.
[16:25] um uh with no service um uh service requirement.
[16:28] and because we can transmit so many channel of course we can also transmit the clock.
[16:32] and by transmitting the clock we remove the need for uh complex DSP retiming.
[16:38] as I said this the footprint we believe in the density is extremely high and will be much higher in the future.
[16:47] today we use micro 50 micrometer pitch with 50 micrometer fiber but of Of course in our road map we envision to use a 25 micrometer pitch uh which means a quadruple of density of throughput per unit of area.
[17:03] Um and what this con this this technology we believe will allow in a chip to chip interconnect is a completely new rethinking of the design of the board.
[17:12] because with the elimination of service we will be able to uh interconnect chip at centimeter of distance many centimeter and half a meter even a meter.
[17:21] Um and that will
[17:24] change of of course completely the possible configuration of uh the architecture starting from the remoting and potentially this the disagregation of HBM.
[17:35] Um um and of course offering HBM a completely new band bandwidth which means significantly increasing the throughput of uh of the compute system.
[17:48] From a mechanical alignment perspective also we believe this one will be a game changer compared to laser.
[17:52] We're demonstrating in our lab um full full signal at 10 to the - 9 10 to the - 10 bit rate as we move the fiber by 20 micrometer both in the x-axis and the y-axis and the z-axis uh which of course would be unthinkable with uh with laser.
[18:14] And last but not least, of course, this system offer very significant reliability with respect to temperature as gallon nitriton sapphire can sustain a very high temperature at 250 Celsius
[18:25] and above.
[18:28] But we believe also there is a value proposition also with respect to scale up both in comparison to active electrical cable and we are developing now of our first product at 1.6 six terab.
[18:39] And with respect to a copper cable, of course, the advantage would be significant longer reach up to 20 25 meter reduction in power.
[18:47] We expect at least 35% power reduction um at 12 watt per cable end, significant reduction of the thickness of the cable and more importantly significant reduction of the weight.
[19:02] Um the fiber cable match exactly the footprint of the micro and if the micro is 1 millm square the bundle cable is 1 mm square and that means of course that the cable weight will be instead of 250 m 250 kilogram somewhere in few tens of kilogram for the overall interconnect bind uh bind Iraq um we do expect though
[19:29] also significant advantages also with respect to 1.6 six terra laser.
[19:34] Starting from reliability, as I said, intrinsically because of the higher reliability of a an LED compared to um a laser diode.
[19:48] Um but also more importantly because of the redundancy we can finally introduce logically uh structurally in the system with spare channel and again we expect reduction of at least half of the power with respect to laser and lower cost lower cost I I I saw a presentation today where um near was mentioning that the possibility or the cost the system level of the fiber would be very high.
[20:13] We tend to disagree with that because the fiber we're going to use is a fiber that has a very different index profile for 15 20 25 m interconnect than the fiber that you use for kilometer interconnect.
[20:28] And so we believe that the cost road map of those
[20:30] fiber um uh will allow for a very very
[20:33] competitive cost structure also of the
[20:35] active optical cable. What really excite
[20:37] us though it's really the scalability of
[20:39] the technology and this one is an
[20:42] illustrative way and I want to emphasize
[20:45] an illustrative way to think about the
[20:48] scalability of the system in the future.
[20:51] Our first product we start at 1.6 six
[20:53] terra and you can think about and so
[20:55] what you see in this diagram in the y-
[20:57] axis of course is tra bit per millimeter
[21:00] um of shoreline and on the x-axis the
[21:03] speed of each micro 4 gigabit 8 gigabit
[21:06] 12 and 16
[21:09] we start the first product at 1.6 terra
[21:11] and product at 1.6 six terra with 4 GB
[21:15] per micro D means 400 microL
[21:19] um but you could think about 400 micro
[21:23] at 8 GB and that would be 3.2 to terra
[21:27] and then as a next generation
[21:29] u the idea will be to scale the the the
[21:33] the
[21:34] pitch of the LED. In the blue line you
[21:37] have a pitch of 50 micrometer
[21:39] um um with fiber of 50 micrometer
[21:42] diameter. On the red line we we will be
[21:45] going at 25 micrometer pitch and 25
[21:48] micrometer fiber and that will means
[21:50] quadrupling the density per unit of area
[21:53] and that will be 12.8 a bit per
[21:55] millimeter and then again with 400 micro
[21:59] at 25 micrometer pitch and 16 GB per
[22:02] micro that would be the road map for
[22:04] 25.6 six terab that's still with 400
[22:07] micro per array we could double of
[22:10] course the number of micro per array and
[22:12] this one again in our view could become
[22:14] a very very scalable and quickly
[22:16] scalable technology for be the uh the
[22:20] coming generation to serve scale up but
[22:23] also um uh CPO scaling
[22:27] this of course is very exciting for us
[22:29] but it's not the most exciting part of
[22:31] the road map what really excite us about
[22:33] this technology is the possibility
[22:35] ability actually to use the entire area
[22:38] of the the the chips to trans uh to
[22:42] transmit. And the idea is actually to
[22:44] use our light bundle um um solution at
[22:49] the bottom of of of
[22:53] the compute and and for and and memory
[22:55] to transmit under. Yeah, this is again a
[22:58] purely illustrative picture. But we
[23:01] believe that um with this technology we
[23:04] could achieve easily um um um 10 terab
[23:09] per millimeter square and and you can
[23:11] and you know as you as you consider the
[23:13] area of HPM and computed you can scale
[23:16] up you can imagine what would be the
[23:18] scaling of um um and the density and the
[23:22] capacity of uh of this solution for uh
[23:25] for area to area interconnect
[23:31] So the value proposition here that we
[23:32] believe um micro can offer is of course
[23:36] first and foremost higher reliability by
[23:38] eliminating laser by introducing spare
[23:40] channel that intrinsically will provide
[23:43] higher redundancy temperature tolerance
[23:47] um from a power prosperity again the
[23:49] elimination of laser but at system level
[23:51] particularly for dieto- interconnect
[23:53] it's elimination of the service
[23:56] elimination complex I order modulation
[23:58] the modulation and the elimination of
[24:01] the SP timing
[24:03] and overall in a system package that has
[24:06] a significantly uh simpler level of
[24:09] integration and very small uh footprint
[24:12] and high level of scalability
[24:14] um at OFC um as I said um at boot 3 to4
[24:19] um uh we'll be demonstrating the first
[24:22] uh um I think a worldwide um evaluation
[24:26] kit with over 300 channel um again will
[24:30] be between two and 2.5 gigabit for now
[24:34] um um at a row rate of 10 to the minus 9
[24:39] 10 to the minus 10 um and when I say bit
[24:42] rate bit rate of course I mean no
[24:45] forward error correction required to
[24:47] achieve this level of bit rate um the
[24:50] the demo currently run roughly at three
[24:53] at three per bit but we hope of Of
[24:56] course, in the next few weeks and months
[24:58] with more engineering um to refine uh to
[25:01] refine the performance of the system at
[25:03] 3 3.5 GB per channel um to achieve 896
[25:08] GB overall system or close to one terra
[25:11] at or below 2 pju per bit. Thank you so
[25:14] much for for the opportunity and
[25:18] I pass you to you.
[25:21] Super. Well, thanks a lot uh for the
[25:23] invite for this panel. Uh, it's great to
[25:25] be here. I've never seen the OFC
[25:27] executive forum be so so busy ever. Like
[25:31] I've been in this industry for 35 years
[25:33] now from optical systems to optical
[25:35] components. Uh, and uh, really this is
[25:38] really a massive great great work. It's
[25:40] fantastic work guys. Uh so before I get
[25:43] into uh Ranovis and what we're doing uh
[25:46] in the industry, I wanted to talk a
[25:47] little bit about standardization.
[25:49] And um I think um the most um dramatic
[25:54] change I've seen over the past uh
[25:56] several years really happened this year
[25:59] with uh standardization and you saw the
[26:02] effects of it from OCI MSA and we know
[26:05] we all know that standardization leads
[26:07] to uh cost reduction and uh I uh was
[26:10] part of the uh telco guys uh uh one of
[26:13] the old guard uh from Nortell and uh
[26:16] most of our team is from Lucent and
[26:18] other other companies and system side
[26:20] and we saw how in the telco days uh
[26:22] everything was bookended.
[26:24] All the system vendors were developing
[26:26] systems for themselves and nothing was
[26:28] interoperable.
[26:31] Then after the hyperscalers joined the
[26:34] party um they all disagregated that uh
[26:37] hold that these companies have and most
[26:38] of them went bankrupt and rolled up into
[26:41] some other company. Nokia is the
[26:42] beneficiary of multiple of those uh uh
[26:45] companies and they started to uh
[26:48] introduce interoperability and
[26:50] interworking between parts and then
[26:53] driving the volume and being able to
[26:55] build something that would be
[26:56] generational change in the optical
[26:58] systems and optical modules and that's a
[27:01] phenomenal work they did and in the AI
[27:04] world uh things went backwards the
[27:06] person who had the lead built a closed
[27:08] ecosystem
[27:10] to make sure that they can develop the
[27:12] technologies they need for their
[27:14] performance and I can completely relate
[27:16] to that because you know you you
[27:18] generally start with that but then uh
[27:20] what happens right after that if if you
[27:23] want to take it to uh you want to let's
[27:25] say unbundle that
[27:28] you you forced to start delivering new
[27:30] standards that the whole ecosystem could
[27:33] design into
[27:35] so we even though our background all of
[27:37] the people you see on this chart uh uh
[27:40] as I said they come from um um let's say
[27:42] the telco days and uh bookended systems
[27:45] and building DWDM system. They've been
[27:47] involved in many world's first
[27:49] technologies from ideiation to mass
[27:52] deployment.
[27:54] We in 2018 decided to switch gears and
[27:58] go for standardized development Ethernet
[28:00] compliant interoperable technologies.
[28:04] Don't go for the bookend at once because
[28:06] it's going to be a long long time before
[28:08] you can see mass volume and you have to
[28:10] get the entire ecosystem behind you.
[28:13] This is not a simple task. No matter how
[28:15] much money you raise,
[28:18] 500, 600, 800, 1 billion, doesn't
[28:21] matter. You got to get a customer with
[28:23] you. And the customers with you, they
[28:26] all want to have dual sourcing.
[28:28] Nobody will design you in as a single
[28:31] source to change their entire data
[28:33] center to be based on certain
[28:34] technology. So these were the uh the
[28:37] effects of it is white hair and no hair
[28:39] that you see in most of our uh
[28:41] management team and we start the the
[28:44] whole design cycle uh since our
[28:47] background is from systems is really
[28:48] from a system design which is the
[28:50] foundation to all of the technologies
[28:53] that you're going to be uh developing
[28:55] further and the foundaries you're going
[28:57] to be working with the customers you're
[28:58] going to be working with all of that
[29:00] starts from the system design and then
[29:03] based on that you work with your
[29:04] customers you see who is who's
[29:06] interested in what technologies, what
[29:07] sweet spots are they trying to hit. And
[29:10] that intimacy with those customers gets
[29:13] you access to specifications that are
[29:15] not readily available in the market.
[29:18] Then you take that and you translate it
[29:21] into different technologies. You say,
[29:22] "Oh, well, for this I need this type of
[29:24] laser, this type of um photonic engine.
[29:27] I need this um driver TIA, you know,
[29:30] forward error correction, DSP, all of
[29:33] the parameters that goes into creating
[29:35] that, including the serties of the
[29:37] customer
[29:39] and this becomes a code-design activity
[29:43] and then you can move this forward and
[29:45] start predicting what are the next
[29:46] things in line. So all of our
[29:49] relationships with all of our customers
[29:51] has involved skin in the game. All our
[29:55] relationships with our partners,
[29:57] foundaries, OSATs has been alongside of
[30:00] skin in the game. Because with the skin
[30:03] in the game, what you get,
[30:05] you don't get just some money they give
[30:07] you from NRE or licensing or things of
[30:09] that nature. You get their commitment
[30:12] and their validation of where you're
[30:14] going. And that part is really really
[30:17] important. So we took that to heart and
[30:20] uh went through our journey of uh uh
[30:23] developing our platform. Uh just to give
[30:25] you a little bit of update on where we
[30:27] are. We're 160 people. Uh we broke even
[30:31] last year. This year we'll be doubling
[30:34] our revenue
[30:36] and next year we have $500 million of
[30:39] funnel.
[30:41] majority of that funnel is NPO and some
[30:45] CPO and in 2028 there will be half half
[30:49] between NPO and CPO
[30:52] so getting through that journey and
[30:54] trying to build different pieces of
[30:56] course needs uh strong partnerships
[30:59] partnerships that we generally don't
[31:00] announce in the market because you know
[31:02] announcing partnerships is just telling
[31:04] everybody else what you're trying to do
[31:06] u but it really requires a lot of
[31:09] engineering hard work many many days of
[31:12] code design code spec specifying systems
[31:16] and taking it through the chiplets to
[31:18] your IP cores all of the elements that
[31:21] you have that you bring together and
[31:23] even our laser designs. We designed our
[31:26] own lasers and we have uh you know we we
[31:29] have an um we have a clean room at our
[31:31] facility where we do all of the
[31:33] machinery for assembling the lasers into
[31:35] into our optical engines and then we
[31:37] hand it over to an OSAT to do it in mass
[31:39] volume. There's a lot of things you have
[31:42] to do yourself
[31:44] when you're trying to do a world's first
[31:47] but the most important part of it is not
[31:49] leaving the ecosystem behind.
[31:53] By leaving the ecosystem behind, you may
[31:56] be very happy with what you've achieved,
[31:59] but you'll not get where you want to go.
[32:03] So, so very quickly, um this is the size
[32:05] of the market. Uh you you must have seen
[32:07] this chart from some other folks. Uh uh
[32:09] pretty massive market for AI. U 84
[32:12] billion. The lower portion is the
[32:14] modules and then the top portion is the
[32:16] NPOC CPO. Half of that top portion by
[32:19] the way is the is the lasers as you can
[32:21] imagine. Uh and that's a pretty uh
[32:23] pretty strong vote of confidence for
[32:25] people who build lasers in this audience
[32:26] and sign strategic relationships with
[32:29] other players. uh um and uh the way we
[32:32] build our technology is uh and we've
[32:34] done that at 100 gig uh Ethernet is uh
[32:36] we build it such that it fits into
[32:38] modules
[32:40] NPO and CPO
[32:44] and we use it the design criteria of it
[32:46] working for CPO
[32:49] because that's where you get the most
[32:50] benefit in terms of power consumption
[32:53] size footprint all of those parameters
[32:56] and then you bring it down to NPO it
[32:58] looks world class you take it down into
[33:01] module it beats everybody hands down. So
[33:04] um so maybe just to go through one I
[33:07] have to talk about scale up scale up. So
[33:10] let me let me just go through this very
[33:11] simple diagram. The uh so the first
[33:14] thing we did in co-ackage optics this
[33:16] was uh in 2023 where we took one of the
[33:19] accelerators and we packaged the 5 pjle
[33:21] per bit optical engine interoperable
[33:24] with uh Ethernet. This was a 800 gig um
[33:27] uh Odin platform uh and we brought that
[33:30] to market. We demonstrated it can
[33:31] interoperate with different uh lasers
[33:33] with different modules with different
[33:35] entities on the other side and uh this
[33:38] one had a uh laser this one was a laser
[33:40] outside. Uh then we partnered with
[33:42] MediaTek in 2024 and we built a 6.4
[33:46] terabit version of uh of the same uh
[33:48] construct at 4 pjle per bit. These were
[33:51] all industryleading pieces at the time
[33:53] we're building them. uh and uh now of
[33:55] course the the standards have gone even
[33:57] further and our our next generation uh
[33:59] platform which we're showcasing at the
[34:01] uh at OFC uh you'll see it it does much
[34:04] much better in terms of power
[34:05] consumption here and we also uh
[34:08] partnered with players in the industry
[34:09] to do near package optics the NPO that
[34:12] we introduced in 2025 uh this is a four
[34:14] pico jewel NPO uh 3.2 two terabit. So
[34:18] all of these elements were we're working
[34:20] to try to find out what is the intercept
[34:22] of uh GPU you know CPU and also the
[34:26] memory interconnect to find the right
[34:28] specification that we can help the
[34:31] ecosystem get to where we were where our
[34:33] customers wanted to go. In parallel with
[34:36] that uh Broadcom of course uh 2024
[34:39] fantastic team in Broadcom uh uh
[34:41] building the uh first CPO and being the
[34:43] tip of the spear of the largest
[34:46] companies that were going to move the
[34:47] ecosystem
[34:50] in the shadow of those there's a lot of
[34:52] companies going to grow right and that's
[34:55] Broadcom was one of the lead leaders in
[34:57] that space and these were all it e uh
[34:59] standards they're all 200 100 gig
[35:02] interoperable and you've seen of course
[35:04] uh Nvidia
[35:05] announcing last year and that was a
[35:08] major major announcement to bring
[35:09] another uh major foundry into the
[35:12] ecosystem and then that foundry
[35:16] has significant leverage on the
[35:18] ecosystem to bring OSATs to bring
[35:21] various uh various partners in there. So
[35:23] that's really uh what they brought in
[35:25] and they're they're pushing it forward.
[35:27] I was just checking to see if Jensen
[35:28] announced anything on the GPU side but
[35:30] nothing nothing of yet. Yeah. So, so
[35:33] we're good on that. And then the OCI MSA
[35:36] which I mentioned and this is a major
[35:38] major breakthrough for uh for this space
[35:40] for it to go to mass volume. Really
[35:43] congratulations to the whole team there
[35:45] to uh to bring it together so quickly.
[35:47] We we didn't expect this. We didn't
[35:49] expect this. We looked at the uh 100 gig
[35:52] market, the 200 gig market and you can
[35:54] see is a massive market. uh this is the
[35:56] existing um interfaces and um per
[35:59] wavelength type of uh 200 gig um and you
[36:01] can see it's just starting to uh to ramp
[36:04] and and then 400 gig coming right after
[36:06] that. So what we are uh what we're
[36:09] looking at is two sets of
[36:10] configurations. Uh this is 2027. Uh what
[36:13] is behind our our business plan uh is uh
[36:16] the where you have a merchant ASIC uh
[36:19] merchant or ASIC NPO on one side and
[36:22] then you have a switch on the other side
[36:24] and these NPO high bandwidth modules uh
[36:28] could be co-ackaged. Well, this is the
[36:29] NPO which is on the on the um on the PCB
[36:33] and uh what it allows you to do is
[36:35] allows you
[36:37] for the GPU person to disagregate the
[36:41] switch IC's dominance
[36:44] in the supply chain. It's a dis
[36:46] disagregation play which is phenomenal
[36:49] and you can see a whole slew of
[36:50] ecosystem players there and I think this
[36:53] is going to be a major major we we're
[36:55] involved in three three big big programs
[36:57] in this area and they're all 200 gig um
[37:00] electrical 200 gig per lane uh
[37:02] birectional and also DR u interfaces. So
[37:05] that's for 2027. for 2028 29 the ramp is
[37:10] based on these two configurations and
[37:12] you can see now other players who have
[37:14] the GPU and the CPU and they have the
[37:17] switch and the CPU so those are the
[37:19] players there and you can see the
[37:21] standards there there's e and OCI MSA is
[37:24] gaining a lot of momentum so our program
[37:27] for that uh 2028 uh mass volume
[37:30] deployment is based on the OCI MSA
[37:34] yep so that's pretty much it scaleup is
[37:36] are uh is where we where we are and
[37:38] we're going to be moving into scale out
[37:41] and um and also in the access market in
[37:43] the scale across uh but maybe I'll I'll
[37:46] stop it here. Looks like we're ready for
[37:48] the Q&A. Thank you very much.
[37:53] Yeah. So I put some questions together
[37:55] but I again this is our panel all of us
[37:58] together and uh first of all uh thanks a
[38:00] lot uh speakers great job. Great to see
[38:03] the technology and this is a journey
[38:06] that they've all been on. This isn't uh
[38:08] instant success. I think you can all
[38:11] acknowledge and you have the investment
[38:13] in time, energy, resources and your your
[38:16] passion uh to bring it all forward. Um
[38:18] I'll start with a few questions. Um
[38:21] we'll talk around here a little bit and
[38:22] then I'll open up to the audience and
[38:24] see what you all have in mind as well.
[38:27] So I wanted to kick off with u with
[38:29] Marco. Um you know there's a number of
[38:33] different hurdles in the existence of a
[38:36] of a company like yours and a and a
[38:38] startup uh along the way. Um maybe
[38:41] what's the biggest technical hurdle that
[38:44] you guys had to overcome to get where
[38:46] you are now? Of course, there are a lot
[38:50] of technical hurdles, but uh but I think
[38:52] the most important one is really to to
[38:54] prove that the system can be done. When
[38:57] when you have a new technology never
[38:58] done before, I think the key challenge
[39:00] is really to prove that the entire
[39:01] system can come together and can work.
[39:04] And that's I think really where we are
[39:05] today. Um you know, think about when
[39:08] electric vehicle came out. You know, I
[39:11] was the first one to to say, you know,
[39:12] I'll never drive one. And today I don't
[39:14] think I'll ever go back to
[39:16] >> electrical cars. you're talking about
[39:17] >> yeah to as a as an analogy right so the
[39:22] the the point is to prove that the
[39:23] technology that we have put together
[39:25] really can can work into a system and uh
[39:27] and that's I think it's the biggest
[39:29] technical challenge that of course we
[39:30] will face and of it's a customer um
[39:33] challenge as well but I think it's the
[39:36] biggest one altogether that we'll have
[39:37] to prove yeah
[39:38] >> guys your journeys any big technical
[39:42] >> your biggest technical hurdle maybe to
[39:43] to get there
[39:44] >> I would say uh early on in in the
[39:46] company's is this Yeah.
[39:47] >> Yeah. All right. Good.
[39:48] >> I would say early on in the company's
[39:50] history, I remember trying to source
[39:51] lasers and that was incredibly tricky,
[39:53] especially if you're trying to do DWDM.
[39:55] >> Yep.
[39:56] >> Uh so it was very hard to find anything
[39:58] like that. Uh it felt like there was a
[40:00] lot of advancement that was required in
[40:01] the field. That's why we ended up
[40:03] building our own lasers in the end.
[40:05] >> Um and you know, packaging technology,
[40:07] photonix is a packaging technology. So
[40:10] we've been working on that for you know,
[40:11] eight and a half, nine years.
[40:12] >> The whole time really in some ways,
[40:14] right?
[40:14] >> Pretty much the whole time. uh and so
[40:16] I'd say lasers, packaging and there are
[40:19] so many other intricate little things.
[40:21] Uh today it's validation at the data
[40:23] center scale and firmware optimization
[40:25] and these sorts of things. I mean
[40:28] >> maybe I'll talk about the non-technical
[40:30] items. How about that? Should I switch
[40:31] to the next one?
[40:32] >> Yeah. Yeah. Yeah. I was going to I was
[40:34] going to head there which is um that's
[40:37] the technical piece, right? But but
[40:39] these guys have to manage the business
[40:40] side of their business too and that's
[40:42] investors and and getting to cash flow
[40:44] positivity and and all the rest. So um
[40:47] what about the non-technical piece of
[40:49] the business? Hame
[40:50] >> uh I I think the uh the biggest part was
[40:53] uh having so many solutions out there
[40:58] you know uh there was a time in um in
[41:01] the optical communication where I was
[41:02] saying oh my god we have too many CEOs
[41:04] there are too many smart CEOs and
[41:06] everybody's creating their own little
[41:08] thing and they're doing fantastically
[41:09] well so the customer starts to get
[41:12] confused in terms of which one of these
[41:14] technologies not only customer but the
[41:16] market starts getting confused and which
[41:18] one of these technologies has the
[41:19] highest chance of being adopted and I
[41:21] think that was the biggest hurdle we had
[41:23] at the beginning. Um we had um we had
[41:27] our own we the company started with
[41:28] developing our quantum dot laser taking
[41:30] in mass deployment you know all the
[41:32] silicon photonics all these other things
[41:34] but when we started working with the
[41:36] customer they said can you make this
[41:38] interoperable with existing technologies
[41:40] and that disconnect say you know cost us
[41:44] a couple of years couple years of
[41:46] working on something that was really low
[41:48] power doing all the specs everything
[41:50] right checking with the customers so I
[41:52] think um there was just too many too
[41:54] many technologies in the market and as a
[41:57] result of that we didn't have a unified
[41:59] view from the customers like these are
[42:01] the two tracks we want you guys to
[42:02] innovate on.
[42:05] >> Awesome. Um let's talk ecosystem for
[42:08] just a minute. I think almost all of you
[42:10] kind of touched on that but I'll I'll go
[42:12] back to you Nick. um when you guys began
[42:16] the journey, uh I feel like you probably
[42:19] had to had to build an ecosystem, but
[42:22] yet you're a you're a smaller player.
[42:25] You're you're starting up. You're trying
[42:26] to get mind share uh from folks. Um
[42:30] let's talk a little bit about kind of
[42:31] the journey of of partnership and
[42:34] ecosystem
[42:35] um that you've been on and then maybe
[42:37] you guys can chime in as well, Nick.
[42:39] >> Yeah, I mean I think that's a that's a
[42:41] really important one. Uh, how do you get
[42:43] a little startup, you know, when we
[42:44] started, we were worth something like
[42:45] $30 million. We raised $11 million. How
[42:48] do you get a giant foundry, how do you
[42:50] get a giant packaging house to care
[42:52] about working with you? For us, it was
[42:54] about delivering value to them through
[42:56] our technical skills and driving what
[42:58] they thought was valuable. And then, you
[43:00] know, obviously for us, it was getting
[43:01] our stuff to work at all. Um, so I think
[43:04] that those partnerships have been, you
[43:06] know, uh, hard fought, hard one and
[43:09] we've done good work and over time we've
[43:11] built a lot of friends and that's what
[43:13] an ecosystem ends up being. Uh, pushing
[43:15] the frontier of the technology. I think
[43:17] that's how we do it. Uh, so I'm curious
[43:20] what other people think. Marco,
[43:22] >> yeah, look, it takes a village. In fact,
[43:23] it takes the entire village. If you have
[43:25] a new technology, you cannot go out and
[43:27] do it by yourself. You need to carry
[43:28] with you the supplier from an ecosystem
[43:31] perspective. You need to build the
[43:33] customer and and and and the customer
[43:35] need to be supportive and help you
[43:36] really to drive pricing. Um so um it was
[43:40] exactly the same for us and we have a
[43:43] little bit of um our fortunes that the
[43:45] system is a little bit simpler. So um I
[43:49] we don't have the need to have such a
[43:51] huge list of supplier and uh but
[43:54] definitely carrying with us the
[43:56] ecosystem is one of the most important
[43:57] thing that uh we do.
[43:59] >> Yeah, for for us was a little bit
[44:00] different. Um I guess uh we had built
[44:03] many different multiple uh um world's
[44:08] first type of technologies in system
[44:10] companies like Lucent and Nortell in
[44:12] core optics DSP in Cortina high gain fck
[44:16] in bine optics pawn uh optical chips for
[44:20] pawn network so there's a lot of this
[44:23] network and also I forgot to mention uh
[44:25] JDSU rodms and optical cross connect so
[44:29] our executive team had these networks of
[44:32] people that worked uh for let's say 30
[44:35] years that they had trust in when they
[44:38] were behind the product when they're
[44:40] behind something there must be something
[44:42] behind it. So we didn't have to go
[44:44] spread money around the ecosystem we
[44:47] just went in with our credibility and
[44:49] then start staying the course. I
[44:51] remember u our first discussions IBM was
[44:53] for example one of our big one of our
[44:56] big suppliers at Nortell we did SIG
[44:57] chips with them and then they were a
[44:59] strategic investor in core optics. So
[45:00] they had seen how we evolved through
[45:02] there and when I talked to global
[45:04] foundaries team uh in 2018 I think it
[45:07] was we knew everybody we knew everybody
[45:09] and they knew when we would be behind
[45:11] something we're going to see it through
[45:13] we wouldn't switch foundaries and that
[45:16] was a major state statement of
[45:18] commitment and of course they replied
[45:20] back in return uh being highly committed
[45:22] to what we what we do so there are
[45:24] different techniques I mean you can get
[45:25] raise money and contribute to your
[45:28] ecosystem so they work with you you You
[45:30] could use your reputation, you could use
[45:31] your customer. There are many different
[45:33] ways and I think all of them are valid.
[45:36] >> Uh I'm going to we have a question
[45:38] there. Uh Jose.
[45:40] >> Awesome.
[45:41] >> Thank you very much. U you know when uh
[45:44] >> name and company please
[45:45] >> Mark Mark with Fiber Reality. Thank you.
[45:47] >> Yep.
[45:48] >> I mean when we're talking about
[45:49] breakthrough technology uh you guys have
[45:52] been around from seven to 14 years. I
[45:54] mean uh I and I bring that up and
[45:57] believe me I understand how much guts it
[45:59] takes to start a company. I I really do.
[46:01] But uh is is it maybe not the lesson
[46:04] that uh uh that would start us maybe
[46:07] they should start out with solving a
[46:08] problem initially and then come up with
[46:10] the solution instead of uh if I could
[46:13] just finish I if instead of just
[46:15] starting with the technology looking for
[46:17] that killer app I mean Nick you've uh
[46:19] changed your business plan I believe at
[46:21] least three times and so and I'm you
[46:24] know at the end of the day I'm very
[46:25] concerned about the health of this
[46:26] industry and the fact that uh there are
[46:29] all these startups a lot of them CPO
[46:31] you're related and uh they're getting
[46:34] they're getting all this funding from
[46:35] the clueless venture capitalists. Uh and
[46:38] uh I don't know if that's good for us. I
[46:40] don't know if that's good for us. Thank
[46:42] you very much.
[46:42] >> Yeah, I I completely agree with you,
[46:44] Mark. I mean, that's spot on. And I
[46:46] think uh uh the uh I think the the the
[46:50] important thing is when do you see
[46:53] you're getting escape velocity?
[46:56] You're right. there's maybe a couple of
[46:57] turns and twists and turns like I know
[46:59] we did quantum dot laser and silicon
[47:01] photonics for DCI I guess scale across
[47:04] networks uh and uh and that didn't pan
[47:07] out because Marll team you know had a
[47:09] massive strategic uh investment in that
[47:12] area and we're competing with each other
[47:13] so anyways it's a long story but um but
[47:16] I would say u the important thing is to
[47:18] see if there's anyone getting critical
[47:21] um critical revenue in key accounts to
[47:25] be at scape velocity. If you see that,
[47:29] then then there's a there's a chance
[47:31] that this thing is going to unfold,
[47:32] right? So, uh no, I I I I hear you. We
[47:35] got to start with the with the problem
[47:36] at hand. That's why I was mentioning you
[47:38] got to be with the customer. You got to
[47:39] be around the uh customer's table and
[47:42] try to get them to commit and put a skin
[47:44] in the game or else what you're doing is
[47:46] just going to be a set of charts.
[47:48] >> Yeah, I respectfully I respectfully
[47:50] disagree. I mean the the importance of
[47:53] having startup that develop technology
[47:54] is exactly that you don't know where the
[47:56] industry is going to go next and you
[47:57] need to have many different type of
[47:59] technology and one of them ultimately
[48:01] will succeed
[48:03] >> I want to there would be no invidcom if
[48:05] there not been first a startup right so
[48:07] and and or or or Nokia or lum every
[48:12] company needs to stop from being a
[48:13] startup and then of course the market
[48:16] and the opportunity that come with the
[48:18] market will define which one will be
[48:19] successful I I think the fortune that we
[48:21] have today is that AI has finally
[48:23] accelerated the need for high speeded
[48:25] interconnect and I think will be a huge
[48:28] benefit for all of us and and will
[48:30] create a huge amount of opportunities
[48:32] for our technology to be successful.
[48:34] >> Yeah, just one final note. I think
[48:36] hardware companies famously take about a
[48:38] decade to build. Uh I think we're all
[48:40] about on time. It's about a decade.
[48:42] That's what it takes. Um I want I I do
[48:44] want to pick up on on some of this
[48:46] discussion and at the risk I don't want
[48:49] to turn this into Jerry Springer here.
[48:51] So uh I there is some caution in my next
[48:53] question. Um but I'm going to start with
[48:55] Nick and then everybody respond which is
[48:59] um as we look at the market as we look
[49:02] at what's happening with data centers
[49:03] and the scale up and the scale out and
[49:05] and where you guys are um is the market
[49:10] opportunity
[49:11] big enough and your technologies diverse
[49:15] enough that there's there's multiple
[49:18] folks that succeed here or how are you
[49:21] thinking about the market and its
[49:22] segmentation?
[49:24] Yeah. I mean, just to bound the market
[49:26] size quickly, so say something like 20
[49:28] to 30 million GPUs in 2030, a few
[49:31] thousand per GPU, and that's just on the
[49:34] CPU side. The laser side is about
[49:36] equally as big. It's a huge market. I
[49:39] think that these hyperscalers want to
[49:40] have multiple choices uh for where to
[49:43] go. Our spot in the ecosystem is at the
[49:45] absolute frontier. Yeah, we go back to
[49:47] the older technologies where it makes
[49:49] sense for us and we bring back some of
[49:51] the futuristic things that we do build
[49:53] to go address that space. We did that
[49:55] with L20 uh last week.
[49:57] >> But ultimately these companies want
[49:59] multiple partners and I think there are
[50:01] going to be a lot of big companies in
[50:03] the space.
[50:04] >> Yeah, Marco,
[50:05] >> I agree there is going to be a
[50:06] tremendous opportunity in scale up and
[50:08] scaling. um the size of the market as
[50:11] Nick said is going to be mindbending
[50:13] >> just microphone
[50:14] >> and the size of the market is going to
[50:16] be mindbending and the biggest problem
[50:18] that AI system have today it's really as
[50:20] Nick also showing his slides it's really
[50:23] the the massive disconnect between the
[50:25] compute speed and the interconnect speed
[50:27] right um if if if we can help the
[50:31] industry to fix the problem I think the
[50:33] the opportunity is so large that I
[50:35] believe all these different technology
[50:37] will be able to coexist
[50:39] Amid.
[50:40] >> Yep.
[50:40] >> I agree.
[50:41] >> You're in on coexistence.
[50:43] >> Absolutely. I I think as long as things
[50:45] are interoperable, I think there's
[50:48] there's a massive opportunity there. If
[50:50] you do bookended things, then you have
[50:52] to really convince a big player to buy
[50:56] into that. And you can you can I mean
[50:58] you can I mean I think Marvel convinced
[51:01] somebody that their bookended solution
[51:03] is going to work and they bought a
[51:04] company for it, right? So, I think it is
[51:07] possible to be done. Um, yeah.
[51:10] >> Um, you mentioned the um kind of M&A, so
[51:14] I'm going to I'm going to go ahead and
[51:15] pick up on that if I don't see other
[51:16] questions from anywhere else. Does
[51:18] somebody have a question over here?
[51:20] Okay. Um, you have one. Okay. Go ahead,
[51:23] John.
[51:24] >> Uh, hello. Uh, my name is Rafi Islam.
[51:26] I'm from Cactus Materials. Uh, we we
[51:30] actually manufacture the lasers in
[51:32] Arizona Tempe in 6in fab. I'm just
[51:34] curious from the Nick question that you
[51:36] had some challenges on the laser
[51:39] sourcing. I'm just curious if you can
[51:42] highlight what are those challenges. Is
[51:44] it the materials like indium phosphate
[51:46] materials or gallium arsenite in this
[51:49] kind of case or the manufacturing in the
[51:51] United States? Thank you.
[51:53] >> Go ahead.
[51:54] >> Yeah, happy to talk about lasers. So, as
[51:56] you look at DWDM, there start to be very
[51:58] strict requirements on the drift in the
[52:00] laser color. There's two things you want
[52:02] to know. what color is the laser line?
[52:05] Like what wavelength specifically? And
[52:07] then you'd like it to also not move very
[52:08] much. As you pack many colors side by
[52:11] side, they can't be sort of encroaching
[52:13] on each other or the links go down. And
[52:15] so meeting both of those requirements as
[52:17] well as power is one of the things
[52:19] that's been really tricky for the
[52:20] industry. People have traditionally
[52:22] built lasers in indium phospide
[52:23] foundaries. Laser bars are something
[52:25] that we've been sampling for probably
[52:27] seven years, something like that. We've
[52:29] always had challenges getting these
[52:31] solutions to uh not require heavy
[52:33] binning and therefore have low yield uh
[52:36] to meet those precision and accuracy
[52:38] requirements for the wavelengths. And so
[52:41] that's what we had trouble sourcing. Uh
[52:43] we invented guide and this has tunable
[52:46] lasers and we're able to boot up the
[52:48] platform, assign the wavelengths and we
[52:50] have you can see it at OFC two
[52:52] generations of that technology 5
[52:54] gigahertz error which is obscenely good.
[52:57] I'm very proud of what the team has done
[52:58] on that front. Uh, and this is required
[53:00] as we continue to scale. People are
[53:02] announcing 200 gig today in this OCI 400
[53:05] gig some coming soon. We're at 1,600 and
[53:08] we go back down to those 200 gig numbers
[53:10] when we need to. But as you continue on
[53:12] this road map, the laser requirements
[53:14] get stronger and stronger. And so we're
[53:16] making sure we have that technology
[53:18] today. And we're doing it through a
[53:19] technology platform that scales from a
[53:21] cost perspective from little tiny laser
[53:24] diodes that are assembled by hand on a
[53:26] substrate to an integrated circuit. Huge
[53:29] difference in assembly time. Uh big cost
[53:32] improvement. So we're really excited
[53:33] about it and that's what we couldn't
[53:34] source.
[53:35] >> Supply chain for you guys concerned in
[53:38] general with the ramping demand or
[53:40] you're managing that and like everybody
[53:42] else you just kind of work your
[53:44] relationships and work your ecosystem.
[53:46] Um maybe uh Hamid. Sure. So, so I think
[53:50] um if you look at the uh silicon
[53:52] photonics for example um it is it is an
[53:55] emerging let's say supply chain
[53:57] challenge right and um with TSMC having
[54:00] the lion share of the semiconductor
[54:03] of course they have a lot of leverage on
[54:05] the silicon photonic part allocation
[54:08] right so there's there's a lot of uh I
[54:10] think right now the world is all about
[54:13] supply chain management the innovations
[54:16] we talk about disruptive innovations and
[54:18] all that. Some of them are going to be
[54:19] shunted.
[54:20] >> Y
[54:21] >> because the supply chain would not be
[54:23] there because of some movements that
[54:25] some players might make.
[54:27] >> So look out for that.
[54:28] >> Yep. Marco.
[54:30] >> Yeah, we don't lo we don't use lasers.
[54:31] So you can do the math and calculate how
[54:34] many micro you can put on a four or six
[54:36] inch wafer.
[54:37] >> You heard it here.
[54:38] >> A 50 or 25 micrometer length for the
[54:41] micro number is mindbending. So uh we
[54:44] don't have supply issues. I'm going to
[54:45] stay here with you uh on there's been
[54:48] some recent M&A activity obviously I'm
[54:50] thinking about a couple of different
[54:52] companies Celestial AI maybe Nubis um
[54:55] has M&A activity affected you and your
[54:58] go to market business and and maybe I'll
[55:00] ask that to all of you but I'll I'll
[55:02] start here and maybe go down the line
[55:03] >> it's not affected go to well it's not
[55:05] affected go to market but clearly this
[55:06] M&A has created huge momentum with
[55:08] investors so
[55:10] >> it's a positive with
[55:10] >> it's super positive raising raising
[55:13] money has become signific ificantly more
[55:15] palatable for them and that's of course
[55:17] very important because allow us to raise
[55:20] money and which means accelerate
[55:21] dramatically productization and
[55:24] hopefully bring the pro to market
[55:26] sooner. So definitely huge positive.
[55:28] >> Same.
[55:28] >> Yeah, I I think from an M&A perspective
[55:30] this is very good to see you know
[55:32] acquisitions of these companies at
[55:34] reasonable prices and it validates the
[55:36] market opportunity. I think it's an
[55:38] absolutely enormous market. I bounded it
[55:40] earlier so this is all goodness.
[55:41] >> Yep. Uh so I I say on our trajectory you
[55:45] know um we look at um players like um
[55:50] Credo AEC was really what made them go
[55:53] public right um Bill must be here I
[55:55] don't know but uh Aster Labs I mean it
[55:59] was rettimer PCIe rettimer so a single
[56:01] product can take you public and for us
[56:06] NPO is that and then followed by CPO so
[56:09] that's sort of how we look at it the M&A
[56:11] market is uh is actually great support
[56:13] for that too.
[56:14] >> Um, we're almost to time. I don't see
[56:17] other questions. I had two two closing
[56:19] questions I wanted to kind of give this
[56:22] this group. And I know in this in this
[56:24] audience, we have folks who are dreaming
[56:26] about their own startup journey and
[56:29] their own entrepreneurial journey. Well,
[56:31] you have three of uh three of them here
[56:33] uh living it every day. So I'm going to
[56:35] go back to you Hamid and maybe uh you
[56:38] guys chime in as well. But what would
[56:40] you tell folks in this audience that are
[56:42] coming out of the universities or that
[56:45] are setting in a company dreaming of
[56:47] starting their own business? Um what
[56:49] would you tell them uh as they they
[56:52] think about doing that themselves and
[56:53] what maybe you wish you knew um before
[56:56] you got in the chair you're in? Go.
[56:57] >> Sure. So I'll tell them to follow Mark.
[57:00] You know there good good comment there.
[57:02] you know, don't don't come out with a
[57:04] technology from university. I mean, try
[57:06] to surround yourself with people who've
[57:09] done this thing before and have them as
[57:12] advisors. Look at how many cylinders
[57:14] you're missing in your car. Get those
[57:17] cylinders all filled in and they'll
[57:19] maybe maybe not just technical things
[57:21] because I think as you come from out of
[57:22] university, you feel technology is the
[57:26] core of everything.
[57:27] >> Yep. when you come out a little bit
[57:29] further in your life, you realize that's
[57:31] just one part of it. So, I would just
[57:33] say look out for the other parts. Nick,
[57:36] >> yeah, I think it's a pretty pretty
[57:38] complicated deep question. Uh, building
[57:40] companies is hard. I think if you have
[57:43] to ask, should I do this? Then you
[57:45] probably shouldn't do it. Uh, I think
[57:46] that's probably the the best test. Um, I
[57:50] think that to make a small outcome, to
[57:52] do something small in the world, you
[57:54] need to address a challenge that's very
[57:55] obvious and right in front of you. And
[57:58] that's a good way to never get any
[57:59] investment money. If you want to do
[58:01] something really big, something that
[58:03] actually will make a dent, you have to
[58:05] go for a big goal, something that's
[58:06] backed by the foundations of physics and
[58:09] where the world's going. And that's what
[58:11] we've done. I think that's that's what
[58:13] you need to go after if you want to go
[58:15] for a big outcome and do something
[58:17] really incredible and make a difference.
[58:19] >> Yeah, I think all are super important
[58:21] points. I would also add important to
[58:23] find the right investors, right? I mean
[58:24] you need to find investor really are
[58:26] supportive. They can really sustain
[58:28] long-term
[58:29] >> investor partners beyond just cash.
[58:31] >> Exactly. Investor partner way way beyond
[58:34] cash.
[58:34] >> Yeah.
[58:34] >> And they really can help you really to
[58:37] um navigate through many many years of u
[58:40] >> absolute last question I promise. Uh
[58:42] because we are here and uh I told these
[58:45] guys they're going to get one word maybe
[58:47] two maybe a sentence but one word or
[58:50] two. um what best describes how you feel
[58:53] where we are uh as an optical industry
[58:57] today and the and looking forward into
[58:59] the future and I'm going to go Nick
[59:01] Marco Hamid in that order
[59:03] >> game time
[59:04] >> game time
[59:06] >> lightning
[59:07] >> lightning
[59:08] >> optimism
[59:10] >> optimism all right give it up for the
[59:11] panel thank you guys appreciate your
[59:26] Everybody talking about scaling AI,
[59:32] but the data center's choking deep
[59:34] within.
[59:38] Copper running hot. Yeah, the signal's
[59:42] getting thin.
[59:44] So we flip the switch now.
[59:48] Optics is in.
[59:50] Bandwidth climbing fast. Racks are
[59:54] running red.
[59:56] Cloud demand exploding overhead.
[01:00:02] Pluggables fading as the limits closing.
[01:00:09] Cold package light is how we win.
[01:00:14] It's photo
[01:00:16] baby 2026
[01:00:20] riding that light way
[01:00:23] doing new tricks
[01:00:26] from the fiber in the ground to the chip
[01:00:28] in my hand. We make that sunshine jump
[01:00:31] on command. Yeah. Fenix, baby.
[01:00:37] 2026.