High-Speed Interconnect Opportunity (NASDAQ:COHR) | Seeking Alpha

Intels Silicon Laser Chip Promises Faster Data Transfer

Justin Sullivan/Getty Images News

Data traffic continues to grow at an unprecedented rate. According to Cisco (CSCO), data-byte traffic had a 25% compound annual growth rate over the last five years. Driven by artificial intelligence, video streaming, higher-capacity 5G networks, or the internet of things (IoT) – the amounts of bytes of data to be processed principally at large data centers and telecom networks are going through the roof.

chart of data-byte traffic

Cisco

Data centers are especially challenged to store and process these massive amounts of data in a fast and efficient manner. The limiting factor in data centers today is not chip processing speeds. It’s the limited data rate and attendant power consumption when driving data between chips. Part of this challenge can be addressed between memory and compute by various emerging in-memory processing technologies such as neuromorphic computing, where memory and processing reside together, similar to how the brain functions. The data center challenge also is addressed by high-speed interconnect solutions that are being driven by silicon photonics – silicon and optical integration. Both in-memory processing and silicon photonics technologies have been around for over a decade. Silicon photonics, however, should hit the mainstream over the next few years, creating a compelling investment opportunity for first movers.

All-optical networks for long-haul (>1000 km) communications have been deployed since the 1990s. The speed and data capacity of multiplexed light waves caused fiber to displace slower copper-wire networks. Over time, optical networks moved into metropolitan rings (~80 km) and within buildings (ˆ2 km). Shorter distances, however, require higher levels of integration between silicon and optics to realize the full potential of optical communications.

Data Interconnect Evolution

Semiconductor Acquisition Partners

The integration of specially-designed lasers and CMOS chips into compact, viable solutions will take several years to deploy. In the meantime, Mellanox (NVDA) – acquired by Nvidia for $6.9B in 2019 – has been the leading pure-play high-speed chip interconnect provider over the past decade, mainly using integrated solutions without silicon photonics – actually, Mellanox acquired silicon photonics startup Kotura for $82M in 2013, but subsequently wrote-off the investment. Other semiconductor companies like Intel (INTC) have been investing in silicon photonics for several years. Larger communication equipment companies like Cisco also are positioning themselves for the high-speed interconnect opportunity. Faster transmission of data at lower power is the ultimate goal of silicon photonics – that is, deliver the maximum bits per second for the minimum Joules per bit. To the victor goes the spoils – leadership in a $22B optical interconnect market by 2026, according to Mordor Intelligence.

Leading Innovators

I attended the European Conference for Optical Communications {ECOC} in Basel, Switzerland last month. At the trade show, Ayar Labs demonstrated a silicon photonics solution that can reach data rates of up to 2 terabits per second (Tbps). Their current product consists of the TeraPHY I/O (input/output) chiplet and the SuperNova light source, providing customers the first truly implementable silicon photonics solution. Furthermore, Intel announced that it’s supporting the CW WDM MSA rev 1.0 standard, of which Ayar Labs was an early adopter. Ayar Labs is the first mover, and with major industry players standardizing on the laser platform, customers will likely begin committing to the technology.

We’re still a few years from mass adoption of silicon photonics. For example, future versions of Ayar Labs’ product can integrate its laser and chiplet, creating even higher throughput and energy savings. Similarly, future versions of the CW WDM MSA standard will evolve to continually improve performance. But commercial adoption of silicon photonics is now visible.

In the meantime, several competitors with novel, but unproven, technologies are vying with Ayar Labs to become market leaders. Avicena is focusing on a low-power solution that can scale to very high throughput. Avicena uses arrays of gallium-nitride micro-LEDs that are connected via light-fiber bundles to silicon detectors on standard CMOS chips. This approach enables ultra-low power links of < 1pJ (picojoule) per bit with up to a 10-meter reach (at least 5x less power than what Ayar Labs claims, albeit at a shorter distance). When moving down the interconnect distance spectrum to within a system or a card, efficient power scaling becomes critical. Avicena is an earlier-stage company than Ayar Labs but has the potential to address not only data centers but everyday computers, LiDAR for autonomous driving, and other large markets. We observed heavy traffic at Avicena’s booth at ECOC.

data interconnect distance-efficiency tradeoff

DARPA

Enlightra (pka MicroR Systems), an early-stage startup that also attended ECOC, relies upon micro-resonator technology similar to that of Ayar Labs, but claims better power scaling. Enlightra, under-capitalized in a capital-intensive IT hardware business, is a long way off from commercialization but can be a good strategic fit for a larger player.

The big picture is that silicon photonics is inevitable because it’s much more efficient than copper beyond a certain distance, and that critical distance is shortening. That said, Kandou has a proven high-speed SerDes technology for electrical interconnects. SerDes stands for serializer/deserializer, and is chip-to-chip circuitry that converts a serial data stream to a parallel data stream, or vice versa. Kandou licensed its chip-to-chip interconnect technology to Marvell (MRVL) in 2016 and launched a commercial USB product in late 2021.

Kandou’s upcoming product for PCIe (Peripheral Component Interconnect express) for printed circuit boards (PCBs) will take fuller advantage of Kandou’s technology. Based on a novel modulation scheme, Kandou’s technology allows higher speeds mainly due to better design flexibility, allowing closer proximity of chips on PCBs (or even an integrated system on a chip), which in turn, leads to lower power. This PCB (or on-board) space also is being targeted by the aforementioned silicon photonic companies. These are not mutually exclusive solutions. As distances lengthen, Kandou’s advantage diminishes over silicon photonics solutions, and vice versa. So, both solutions can coexist.

Data Interconnect Players

Semiconductor Acquisition Partners

Coherent / II-VI

Coherent (NASDAQ:COHR), the merged II-VI – Coherent entity, had a large presence at ECOC. The company’s $3.2B acquisition of Finisar in 2019 combined with existing VCSEL capabilities makes Coherent a leading VCSEL optical components supplier. Other major optical component suppliers include Lumentum (LITE) (enabled by its NeoPhotonics and Oclaro acquisitions), Infinera (INFN), and IPG Photonics (IPGP). This newly-established Coherent entity has pro-forma CY 2023 consensus revenues of $5.7B (vs. $2.2B for Lumentum, $1.7B for Infinera, and $1.6B for IPG), making Coherent a force to be reckoned with. Potentially, Coherent can use its resources to move upstream from a component supplier, to a more integrated solutions provider. To do so, would require a major capital investment and likely acquisitions of key enabling technologies to leapfrog current innovators. Historically, II-VI/Coherent has thrived on growth by acquisitions with substantial experience integrating acquired companies into its product roadmap.

Lightwave Logic

At ECOC, I sat next to Michael Lebby, CEO of Lightwave Logic (NASDAQ:LWLG), at the IEEE/MSA presentation of the CW WDM MSA standard showcasing Ayar Labs. I met Michael over 20 years ago when he was the Co-founder/CEO of Ignis Optics, which was subsequently sold to Bookham Technologies at a fire-sale price of $15.5M in 2003 – Bookham subsequently merged with Avanex to from Oclaro in 2009, which was then acquired by Lumentum in 2018. He was quite positive on Lightwave Logic’s polymer-based electro-optic technology. Polymers (i.e., plastics) were previously thought to be too slow and unstable to compete in the optical arena. At ECOC, Lightwave Logic announced that it partnered with photonic integrated circuit startup Polariton Technologies to create a 250GHz high bandwidth electro-optical-electrical link. The demonstration at ETH Zurich used Polariton’s high-speed plasmonic modulators, Lightwave’s proprietary perkinamine chromophores and ETH Zurich’s high-speed graphene photodetectors.

While this news is encouraging, it’s unclear what contributed to the super high-speed link. Was it the graphene material (which is many years away from industrial production), the Polariton circuit design, Lightwave Logic’s chromophores or the combination of all three? At the Polariton booth, I asked about its future with Lightwave Logic. I got blank stares, and then they asked my opinion on the subject.

In any case, more testing and build-outs beyond current prototypes will determine if Lightwave’s technology can scale to high-volume manufacturing. For this scaling to happen, very substantial investment is needed in Lightwave Logic by a large commercial partner to prove the technology – a large semiconductor foundry will not likely work with Lightwave Logic in its current state.

Lightwave Logic, which today comprises a small research team making minimal investments, has not generated any revenue since its inception in 2008. Realistically, Lightwave Logic must either raise >$100M to compete with the plethora of other high-speed interconnect startups (which would be highly unlikely, and very dilutive), or be acquired by a large industrial company (in my opinion, large players will not be content to merely license Lightwave Logic’s technology).

The burning question: What is the strategic value of Lightwave Logic to a large player in the high-speed interconnect space? From my experience with private companies in Lightwave Logic’s niche and at the same level of development of a key enabling technology, Lightwave Logic would not receive more than $100M, likely less. The rub is that Lightwave Logic is currently valued at $742M, despite the stock falling 56% this year.

Other Companies Pursuing the High-Speed Interconnect Market

There’s no shortage of competitors, both private and public, vying for a piece of the high-speed interconnect market. On the private side, we have met/reviewed several private companies such as Luminous Computing, Icon Photonics, LightMatter, Lightelligence, Aeponyx, Ranovus, etc., with some having more substance than others. The current bear market will likely lead to bankruptcies or sales for some of these companies, depending on their investor base and commercial traction/prospects. The likely winners are those companies with access to sufficient capital despite the stock market crash, enabling them to forge ahead. While those companies will still need to execute, at least they will have a fighting chance.

Public companies are approaching the high-speed interconnect market from different angles. There are communication/FPGA chip companies like Broadcom (AVGO), Intel/Altera, AMD/Xilinx (AMD), Nvidia/Mellanox, and Marvell/Inphi. On the communication networking side, there are Cisco/Acacia/Luxtera, Nokia/Elenion (NOK), Ericsson (ERIC), and Arista (ANET). From the enterprise server side, there are HP (HPE) and IBM (IBM). On the data center customer side, Google (GOOG) (GOOGL) is best positioned to vertically integrate an in-house high-speed interconnect solution.

This is by no means an exhaustive list of public and private companies targeting the high-speed interconnect market. But it gives a flavor of the large number of entrants pursuing this huge market.

Currently, there are no public pure-play investments in the high-speed interconnect market aside from Lightwave Logic. Ayar Labs, Kandou, and possibly Ranovus will most likely be the first companies to go public, given the right market conditions.

I would like to thank William Colleran for his contribution to this article.

Be the first to comment

Leave a Reply

Your email address will not be published.


*


High-Speed Interconnect Opportunity (NASDAQ:COHR) | Seeking Alpha

Intels Silicon Laser Chip Promises Faster Data Transfer

Justin Sullivan/Getty Images News

Data traffic continues to grow at an unprecedented rate. According to Cisco (CSCO), data-byte traffic had a 25% compound annual growth rate over the last five years. Driven by artificial intelligence, video streaming, higher-capacity 5G networks, or the internet of things (IoT) – the amounts of bytes of data to be processed principally at large data centers and telecom networks are going through the roof.

chart of data-byte traffic

Cisco

Data centers are especially challenged to store and process these massive amounts of data in a fast and efficient manner. The limiting factor in data centers today is not chip processing speeds. It’s the limited data rate and attendant power consumption when driving data between chips. Part of this challenge can be addressed between memory and compute by various emerging in-memory processing technologies such as neuromorphic computing, where memory and processing reside together, similar to how the brain functions. The data center challenge also is addressed by high-speed interconnect solutions that are being driven by silicon photonics – silicon and optical integration. Both in-memory processing and silicon photonics technologies have been around for over a decade. Silicon photonics, however, should hit the mainstream over the next few years, creating a compelling investment opportunity for first movers.

All-optical networks for long-haul (>1000 km) communications have been deployed since the 1990s. The speed and data capacity of multiplexed light waves caused fiber to displace slower copper-wire networks. Over time, optical networks moved into metropolitan rings (~80 km) and within buildings (ˆ2 km). Shorter distances, however, require higher levels of integration between silicon and optics to realize the full potential of optical communications.

Data Interconnect Evolution

Semiconductor Acquisition Partners

The integration of specially-designed lasers and CMOS chips into compact, viable solutions will take several years to deploy. In the meantime, Mellanox (NVDA) – acquired by Nvidia for $6.9B in 2019 – has been the leading pure-play high-speed chip interconnect provider over the past decade, mainly using integrated solutions without silicon photonics – actually, Mellanox acquired silicon photonics startup Kotura for $82M in 2013, but subsequently wrote-off the investment. Other semiconductor companies like Intel (INTC) have been investing in silicon photonics for several years. Larger communication equipment companies like Cisco also are positioning themselves for the high-speed interconnect opportunity. Faster transmission of data at lower power is the ultimate goal of silicon photonics – that is, deliver the maximum bits per second for the minimum Joules per bit. To the victor goes the spoils – leadership in a $22B optical interconnect market by 2026, according to Mordor Intelligence.

Leading Innovators

I attended the European Conference for Optical Communications {ECOC} in Basel, Switzerland last month. At the trade show, Ayar Labs demonstrated a silicon photonics solution that can reach data rates of up to 2 terabits per second (Tbps). Their current product consists of the TeraPHY I/O (input/output) chiplet and the SuperNova light source, providing customers the first truly implementable silicon photonics solution. Furthermore, Intel announced that it’s supporting the CW WDM MSA rev 1.0 standard, of which Ayar Labs was an early adopter. Ayar Labs is the first mover, and with major industry players standardizing on the laser platform, customers will likely begin committing to the technology.

We’re still a few years from mass adoption of silicon photonics. For example, future versions of Ayar Labs’ product can integrate its laser and chiplet, creating even higher throughput and energy savings. Similarly, future versions of the CW WDM MSA standard will evolve to continually improve performance. But commercial adoption of silicon photonics is now visible.

In the meantime, several competitors with novel, but unproven, technologies are vying with Ayar Labs to become market leaders. Avicena is focusing on a low-power solution that can scale to very high throughput. Avicena uses arrays of gallium-nitride micro-LEDs that are connected via light-fiber bundles to silicon detectors on standard CMOS chips. This approach enables ultra-low power links of < 1pJ (picojoule) per bit with up to a 10-meter reach (at least 5x less power than what Ayar Labs claims, albeit at a shorter distance). When moving down the interconnect distance spectrum to within a system or a card, efficient power scaling becomes critical. Avicena is an earlier-stage company than Ayar Labs but has the potential to address not only data centers but everyday computers, LiDAR for autonomous driving, and other large markets. We observed heavy traffic at Avicena’s booth at ECOC.

data interconnect distance-efficiency tradeoff

DARPA

Enlightra (pka MicroR Systems), an early-stage startup that also attended ECOC, relies upon micro-resonator technology similar to that of Ayar Labs, but claims better power scaling. Enlightra, under-capitalized in a capital-intensive IT hardware business, is a long way off from commercialization but can be a good strategic fit for a larger player.

The big picture is that silicon photonics is inevitable because it’s much more efficient than copper beyond a certain distance, and that critical distance is shortening. That said, Kandou has a proven high-speed SerDes technology for electrical interconnects. SerDes stands for serializer/deserializer, and is chip-to-chip circuitry that converts a serial data stream to a parallel data stream, or vice versa. Kandou licensed its chip-to-chip interconnect technology to Marvell (MRVL) in 2016 and launched a commercial USB product in late 2021.

Kandou’s upcoming product for PCIe (Peripheral Component Interconnect express) for printed circuit boards (PCBs) will take fuller advantage of Kandou’s technology. Based on a novel modulation scheme, Kandou’s technology allows higher speeds mainly due to better design flexibility, allowing closer proximity of chips on PCBs (or even an integrated system on a chip), which in turn, leads to lower power. This PCB (or on-board) space also is being targeted by the aforementioned silicon photonic companies. These are not mutually exclusive solutions. As distances lengthen, Kandou’s advantage diminishes over silicon photonics solutions, and vice versa. So, both solutions can coexist.

Data Interconnect Players

Semiconductor Acquisition Partners

Coherent / II-VI

Coherent (NASDAQ:COHR), the merged II-VI – Coherent entity, had a large presence at ECOC. The company’s $3.2B acquisition of Finisar in 2019 combined with existing VCSEL capabilities makes Coherent a leading VCSEL optical components supplier. Other major optical component suppliers include Lumentum (LITE) (enabled by its NeoPhotonics and Oclaro acquisitions), Infinera (INFN), and IPG Photonics (IPGP). This newly-established Coherent entity has pro-forma CY 2023 consensus revenues of $5.7B (vs. $2.2B for Lumentum, $1.7B for Infinera, and $1.6B for IPG), making Coherent a force to be reckoned with. Potentially, Coherent can use its resources to move upstream from a component supplier, to a more integrated solutions provider. To do so, would require a major capital investment and likely acquisitions of key enabling technologies to leapfrog current innovators. Historically, II-VI/Coherent has thrived on growth by acquisitions with substantial experience integrating acquired companies into its product roadmap.

Lightwave Logic

At ECOC, I sat next to Michael Lebby, CEO of Lightwave Logic (NASDAQ:LWLG), at the IEEE/MSA presentation of the CW WDM MSA standard showcasing Ayar Labs. I met Michael over 20 years ago when he was the Co-founder/CEO of Ignis Optics, which was subsequently sold to Bookham Technologies at a fire-sale price of $15.5M in 2003 – Bookham subsequently merged with Avanex to from Oclaro in 2009, which was then acquired by Lumentum in 2018. He was quite positive on Lightwave Logic’s polymer-based electro-optic technology. Polymers (i.e., plastics) were previously thought to be too slow and unstable to compete in the optical arena. At ECOC, Lightwave Logic announced that it partnered with photonic integrated circuit startup Polariton Technologies to create a 250GHz high bandwidth electro-optical-electrical link. The demonstration at ETH Zurich used Polariton’s high-speed plasmonic modulators, Lightwave’s proprietary perkinamine chromophores and ETH Zurich’s high-speed graphene photodetectors.

While this news is encouraging, it’s unclear what contributed to the super high-speed link. Was it the graphene material (which is many years away from industrial production), the Polariton circuit design, Lightwave Logic’s chromophores or the combination of all three? At the Polariton booth, I asked about its future with Lightwave Logic. I got blank stares, and then they asked my opinion on the subject.

In any case, more testing and build-outs beyond current prototypes will determine if Lightwave’s technology can scale to high-volume manufacturing. For this scaling to happen, very substantial investment is needed in Lightwave Logic by a large commercial partner to prove the technology – a large semiconductor foundry will not likely work with Lightwave Logic in its current state.

Lightwave Logic, which today comprises a small research team making minimal investments, has not generated any revenue since its inception in 2008. Realistically, Lightwave Logic must either raise >$100M to compete with the plethora of other high-speed interconnect startups (which would be highly unlikely, and very dilutive), or be acquired by a large industrial company (in my opinion, large players will not be content to merely license Lightwave Logic’s technology).

The burning question: What is the strategic value of Lightwave Logic to a large player in the high-speed interconnect space? From my experience with private companies in Lightwave Logic’s niche and at the same level of development of a key enabling technology, Lightwave Logic would not receive more than $100M, likely less. The rub is that Lightwave Logic is currently valued at $742M, despite the stock falling 56% this year.

Other Companies Pursuing the High-Speed Interconnect Market

There’s no shortage of competitors, both private and public, vying for a piece of the high-speed interconnect market. On the private side, we have met/reviewed several private companies such as Luminous Computing, Icon Photonics, LightMatter, Lightelligence, Aeponyx, Ranovus, etc., with some having more substance than others. The current bear market will likely lead to bankruptcies or sales for some of these companies, depending on their investor base and commercial traction/prospects. The likely winners are those companies with access to sufficient capital despite the stock market crash, enabling them to forge ahead. While those companies will still need to execute, at least they will have a fighting chance.

Public companies are approaching the high-speed interconnect market from different angles. There are communication/FPGA chip companies like Broadcom (AVGO), Intel/Altera, AMD/Xilinx (AMD), Nvidia/Mellanox, and Marvell/Inphi. On the communication networking side, there are Cisco/Acacia/Luxtera, Nokia/Elenion (NOK), Ericsson (ERIC), and Arista (ANET). From the enterprise server side, there are HP (HPE) and IBM (IBM). On the data center customer side, Google (GOOG) (GOOGL) is best positioned to vertically integrate an in-house high-speed interconnect solution.

This is by no means an exhaustive list of public and private companies targeting the high-speed interconnect market. But it gives a flavor of the large number of entrants pursuing this huge market.

Currently, there are no public pure-play investments in the high-speed interconnect market aside from Lightwave Logic. Ayar Labs, Kandou, and possibly Ranovus will most likely be the first companies to go public, given the right market conditions.

I would like to thank William Colleran for his contribution to this article.

Be the first to comment

Leave a Reply

Your email address will not be published.


*