Listen to this articleBelow is our recent interview with Dr. Suresh Venkatesan, Chairman and Chief Executive Officer at POET Technologies:
Interviewer: Thank you for joining us, Dr. Venkatesan. POET Technologies is a hardware company with solutions for the AI and datacom industries. What are your current products?
Suresh Venkatesan: We have developed a portfolio of devices we call “optical engines” that are the primary working elements of optical transceivers that enable data (including streaming video, video conferencing, etc.) to be communicated from cloud data centers to users at light speed over fiber-optic networks. What makes POET unique is that our optical engines are fully integrated, chip-scale devices that are assembled like semiconductors hundreds at a time, rather than one at a time, which is how most optical engines are built today. We have met and overcome the challenge of scale. As such, we are able to make optical engines cost effectively at the size, volumes, and energy savings that are needed for high-speed, high-bandwidth applications like artificial intelligence.
Interviewer: Artificial Intelligence applications are relatively new to the public. How were you able to anticipate what hardware would be needed to address this emerging market?
Suresh Venkatesan: We didn’t. We were building optical engines for the cloud data center market. It just so happened that what we had developed was perfectly suited for the AI hardware market in two respects:
- We had developed a set of building blocks that have allowed us to design high-speed optical engines that are being built into modules. With our technology inside, those modules are then capable of specifically addressing the needs of data centers that specialize in AI. So, for example, while the majority of cloud data centers are equipped with fiber-optic networks operating at 100G or 400G, those centers that handle AI software have all been upgraded to 800G. POET is currently offering 800G receive engines and we will soon be introducing 800G transmit engines, and have designs for 1.6T engines in the works as well.
- AI hardware suppliers are also trying to solve the bandwidth, latency, and efficiency problems that exist at the chip level, that is, the communication links between the GPU chip and the memory chip. In order to solve this problem, these suppliers are looking for ways to make those chip-level data links using light rather than electrons traveling through copper. As it turns out, the POET Optical Interposer™, which is the platform technology that all of our products are built from, provides a cost and efficiency advantage over conventional approaches to this problem. It allows us to produce light source modules at a cost that can be up to 75% lower than other suppliers.
Interviewer: You have referred to POET’s approach as being disruptive, allowing the company to potentially achieve a sizable share in what looks like a very large market. Can you explain why you believe these statements to be true?
Suresh Venkatesan: Before coming to POET I had over 30 years of experience in technology development with some of the largest semiconductor companies in the world. My last role was Senior Vice President of Technology with GlobalFoundries, the second largest semiconductor manufacturer in the industry, so bringing new technologies to market is what I do.
As I was evaluating the technology that POET was developing when I joined, it became obvious to me that the photonics or optoelectronics industry (as it is called) was using outdated methods to construct optical engines. It was as if all of the progress that the semiconductor industry had made over the past 30 years to produce smaller, more powerful semiconductor devices was simply being ignored by optoelectronics developers. I adopted as the company’s primary mission to “semiconductorize” photonics, by applying proven semiconductor techniques and equipment to this field as a means to integrate components and package photonic devices. Just as integrated circuits led to a revolution in the size of computers from mainframes to cellphones, I believe that the “semiconductorization” of photonic and optoelectronic device fabrication would enable the production of smaller, more efficient optical engines. At the same time, being able to produce devices hundreds at a time on a single silicon wafer, rather than one at a time — which is the conventional approach — would result in the same economies of scale that we have seen in computing.
I firmly believe that the growth projections for data centers and AI cannot be met by incremental improvements to conventional production methods. New thinking and new approaches are required and POET has pioneered one such approach based on hybrid integration. Our method of packaging best-of-breed components onto a common platform that has unique features for the integration of both photonic and electronic devices is one of the most cost-effective solutions that is deployable now. That’s why I believe that the POET Optical Interposer is a disruptive technology.
As for the market and our ability to penetrate the market, I would make two observations:
- We are talking about very large markets. The market that we can serve today is about $1 billion in value and is expected to grow to almost $20 billion by 2029. Our revenue growth over the coming years comes from two sources: First, growth of the overall market for transceivers, as the number of new data centers being built by the Microsofts and Googles of the world is expanding rapidly; and second, by our own product roadmap in which we expect to launch leading-edge, high-speed optical engines over the course of the next year.
- Suppliers to the data center industry can rapidly lose market share as new companies come along with new technologies that offer lower cost and higher performance. This scenario has been witnessed over and over again in the past 20 years. Of the top 10 transceiver suppliers in 2010, only two companies were in the top 10 in 2022. That kind of rapid change demonstrates that incumbent technologies are constantly vulnerable to upstart competitors who have focused on novel ways to improve performance and efficiency. Such disruption will only increase as the market expands at the rate that is projected by third-party analysts. There will be plenty of room for newcomers like POET to take a meaningful, if not dominant share.
Interviewer: How extensive is your current product line and what does your product roadmap look like?
Suresh Venkatesan: In March 2022 we introduced our first optical engine, a 100G device, after about four years of platform technology development and about a year of product development. We invested a lot in the platform — the initial optical interposer — incorporating plenty of functionality that allows POET to easily substitute one component for another (such as different types of lasers, for example) and to rapidly design and prototype a variety of configurations.
Over the course of about 18 months, we have been able, as a result of our investment in the platform, to introduce roughly another dozen products. They are in various stages of sampling with our customers and in some cases are headed to production.
Interviewer: With all of these product introductions what does your revenue profile in the coming year or two look like?
Suresh Venkatesan: POET’s optical engines have been designed into modules that are currently being engineered by several module makers in China, which is where virtually all transceiver module companies are located. Once the module designs are completed, qualified and then sold by those companies to end users, we expect to see significant optical engine revenue. That revenue will go to our joint venture, Super Photonics Xiamen (SPX), in China that we formed back in 2020 with a subsidiary of San’an Optoelectronics. In return for completely funding the equipment and personnel for a state-of-the-art assembly facility, we gave Super Phonics the right to sell those products. However, POET currently owns a 70% stake in the JV, so we are building a company there with substantial current and future value. We expect that Super Photonics will continue to build and sell optical engines in China, which is the primary market for such devices.
POET is now focused on developing and selling modules, at leading-edge speeds and remote light sources, all directed at data centers that serve AI clients. The modules will require additional development, which requires additional expense, which is why we have been out raising capital for the company. We displayed mechanical samples of a prototype POET 800G transceiver module at the China International Optoelectronics Expo (CIOE) in Shenzhen in September.
The introduction of module making to our operations signals that we are ready to take the leap to selling directly to Tier 1 end users, which will speed our time to market. We are working with more than a half-dozen customers on module designs that incorporate POET Infinity chiplets and our 400G/800G receiver solutions for 400G FR4 and 800G 2xFR4 transceiver modules.
Secondly, we are working on a transmit optical engine design that will incorporate externally modulated lasers (EMLs), which have become an industry standard at 400G and 800G. These engines will operate at 100G per lane and, notably, 200G per lane, which is a significant advancement from existing solutions.
With those products, as well as our 100G and 200G products and a planned 8-channel packaged light source for the AI market, we expect 2025 to be a year of high growth and sales for POET.
Interviewer: How will the company fund its operations until the revenue you forecast in 2025 arrives?
Suresh Venkatesan: We are working hard to raise the capital needed to cross the finish line to achieve the revenue that we are confident awaits. Innovations like the ones POET is introducing to the marketplace have historically proven to be lucrative and we believe being in that class should attract the level of investment required. The current global equity capital market is extremely challenging for pre-revenue publicly traded microcap companies. That said, we have ongoing conversations with investors and industry partners who share our vision to bring our technology not only to the market but to the forefront of it.
Also, 2025 is not that far away. As our customers and our customers’ customers complete their sampling of our products, we expect to reach the next phase of our growth in the form of production orders and increased resources to meet demand.
As stated, our commercialization efforts have already started. At Super Photonics, orders for 100G and 200G products are being fulfilled for customers such as Fibertop Technology and BFYY.
In November, we delivered the first production units of custom 4x100G LR4 engines to Adtran, which is a large network infrastructure company based in North America. Aligning with Adtran’s schedule, we expect to ship production volumes of engines in the second half of 2024. Adtran’s unique pluggable device, with its MicroMux Quattro architecture, coupled with POET’s innovative optical engines, enables network operators to transform a 400GbE port into four 100GbE ports with no change in footprint.
Interviewer: Finally, what do you think POET’s outlook will be in one year from today?
Suresh Venkatesan: I foresee POET being in a position where it is well capitalized and, through Super Photonics, gaining broader adoption for our technology in the form of sales of optical engines, optical modules, and light sources. We’re moving forward and any given day now can provide the key moment that can launch the company’s fortunes higher. That makes being part of POET at this time tremendously exciting.
