Q&A: How POET Technologies' Optical Interposer Achieves ‘Universality’

April 7th, 2021

The landscape of technology is poised for significant change and POET Technologies is positioned to be one of the drivers of that evolution in optics and electronics.

Its hybrid solution to integrating optical devices with electronic components presents a massive savings in time and cost for an industry constantly seeking greater efficiency. By supplying improved performance at lower cost and with fewer engineering steps, the company’s optical interposer promises to be a vital component in the growth of autonomous-driving cars and other artificial intelligence solutions. First, its pending products will be implemented for 100G, 200G and 400G networking applications, bringing much-needed solutions for its customers.

But how might the company’s inventions impact day-to-day life and when will the world start to benefit from POET’s efforts? Those were among the questions posed to the company’s braintrust during a recent interview with journalist Adrian Brijbassi for a series of ongoing articles on the development of its commercialization plans. CEO Suresh Venkatesan touted the optical interposer platform as having the potential for “universality” and possessing a level of versatility for use “in a number of applications,” among other traits.

Q: What do you see the world looking like once the POET optical interposer is in use in a large number of datacenters and as a component in AI products and autonomous-driving cars?

Suresh Venkatesan: We are deploying the optical interposer immediately on 100G, 200G and 400G, and that fact means we are implementing the vision of hybrid integration.

Based on our past experiences collectively, a unifying packaging technology that can be deployed at scale and very low cost can be a game-changer because conventional thought processes of integration could be subverted using hybrid-integration techniques. This has been proven out in semiconductors.

There are many paths to integration and we believe what POET is doing is a particularly attractive path. With our unifying hybrid platform, the applications become somewhat endless. As long as photonics are involved and as long as photonics communicates to electronics then the applications that require that convergence all play into the space that the POET optical interposer can support. Whether we choose to support it or not depends on the business opportunity and the timing, and our ability to deliver solutions within the context of our resources.

But the fact of the matter is, regardless of whether it is sensing, communications, and other technologies where photonics and electronics need to co-exist then this platform comes into play. I think the platform is versatile enough that given enough resources it can be put to bear and be utilized in a number of applications. 

If you take common-place applications — autonomous driving could be one — then, yes, to the extent those applications drive the convergence of optics and electronics, then it makes sense for POET to play a role.

Microprocessors, high-speed communications, high-density communications between processors and memory — those are all areas where we can have an impact. When we talk about consumer products, people usually only think about consumer electronics. But microprocessors and data communications are also a big part of everyday life, they just happen to not be in the form of a consumer device. 

Q: I always think about data as a consumer product. I spend about $20 per month on data storage alone, plus well above that for my mobile-phone data. Will someone like me benefit from POET’s presence in the datacenters?

SV: We spend a lot of time talking about cloud and cloud infrastructure, and things of that nature. We are really talking about how quickly data can be moved as it relates to the information being generated by the switch.

Another really important aspect of data is storage. We don’t talk inside POET about storage-area networks, or SANs as they’re called. Storage-area networks have very similar requirements in terms of high-speed data communications between a storage-network node and a processor. In fact, they tend to use far more parallel communications than the standard data communications segments. 

So there are applications there that once we get our first set of products out become possible. There are well-known storage-area network protocols that also have reasonably high volume, not nearly as high as data communications, but they are notorious in terms of margins. So those types of applications do open up to us.

Again, it’s not something that a customer sees on a day in, day out basis, but a customer definitely experiences it on a day in, day out basis as they consume and generate data.

With regards to consumer-oriented devices, they are by definition on the lower end of the totem pole of technology. Phones and tablets and mobile devices that use sensors utilize VCSELs and there is not as much of a need for the kind of integration we provide in that space. As those sensors become more and more complex and perhaps need a bit more integration and form-factor benefits, is there a possibility down the road that those devices could use the capability we have? Perhaps. But it’s clearly not something that we would pursue right now. 

Today in sensing, the most critical high-volume application that we believe we can play a role in is LiDAR [light detection and ranging], only because LiDAR modules are huge and there is a push to move to solid-state for the next generation. It is not something we are chasing at the moment, but it is clearly something within the wheelhouse of the technology and its capabilities.

Vivek Rajgarhia: When I was making transceiver modules early in my career, we had the PC board, which would go through an SMT line, or surface-mount-technology line, and out would come the board with all of the electronics. Then 98% of the cost, time, and energy was taken up in putting the optics in place. The potential here, when it happens with the optical interposer, is powerful. Just imagine putting in an interposer, like you would an SMT line, and out comes an optical module, for use in whatever application. Think about it. 

That’s where the analogy comes in about this technology doing for the world what transistors previously did. Because of transistors and how they have proliferated you don’t even think about how a motherboard is manufactured. At the extreme, the optical interposer has the potential to be such a game-changing invention.

It’s not about assembly at all. We design it and it’s done just like semiconductors.

SV: The optical interposer enables an SMT-like assembly of photonics.

There is no inherent integration inside the PCB [printed circuit board]. What the interposer does is a balance between how much you integrate and how much you choose not to integrate. We have chosen to define the interposer as an integrated set of passive components with a non-integrated set of hybrid active components. In that context, we have taken the complexity of placing optical components and building that complexity into the interposer, so the rest of the components can be placed.

Since we take away a lot of the complexity of placement and alignment out of the equation, the optical interposer effectively enables an SMT-like capability onto photonics that has heretofore not existed.

It can be applied to anything. Any device you open will have a PCB that uses a surface-mount technology. Likewise, the interposer can be applied to anything that requires an assembly of photonics components. 

In an ideal world, we get to a point where you define different interposers for different applications and then it goes through an SMT-like module. You could almost think of it in that context. 

Let’s say we have a design kit available and we opened up the technology and basically licensed it to everybody and then everybody could make their own interposers for their specific application and create whatever assembled photonics components they choose to manufacture — that’s how it could work.

Q: Seeing how companies have gone to that licensing model in other industries, is that where POET is building towards for the future?

SV: That’s a different business model. I don’t think we’re quite there yet. POET as a company will not go after every application and it is possible that if we get to that point of universality with our development in the next year, year and a half, then the options and opportunities to supplement revenue grows. It would be similar to what Qualcomm does. They make their own devices but they also license the IP that other people use to make devices too. So you could work on that kind of model, where we have a set of products that we build and we license the underlying IP to someone else to build other products that may or may not compete.

That typically happens when you reach a level of market penetration where you have made your solution the de facto solution everywhere. I think we are a ways from there today. 

Q: What do you think will be the biggest change the optical interposer will have on the industry in the short term, the next year to two years?

VR: Short term, we are providing this engine In which the most difficult part of the integration has been done. Customers either make or buy lasers and photonics devices, and that part of the manufacturing process is very complicated. Then to use it in a module is another level up in complexity.

What impact we would have is in meeting the demand. We are in an industry that is constrained by supply. We can open that supply up by providing this rich optical engine. It has its cost savings that we’ve been through, of course, but the ease of manufacture is also a very attractive benefit. When we have multiple customers using and scaling with our engine that becomes significant.

Data is constrained by the optical supply, not what they can use. It’s too complicated for data companies in 100G, 200G and 400G to do themselves. We solve that complication by making it easier for module vendors to make a module using the engine.

Q: Something I don’t think the company has mentioned is whether POET’s solution can benefit the environment. Will POET be able to decrease the impact of data usage on the environment?

I think honestly most companies that market themselves as environmentally friendly will take some aspect of what they’re doing — like saying they’re using a thinner sheet of plastic than someone else — and call it a green solution. 

If you look at it in that perspective, then okay our solution basically is an interposer, wafer, and components. That’s it. The sheer reduction in the bill of materials could be parlayed in some manner as reducing consumption of materials and that can be considered green, but that’s a stretch, I feel.

VR: Where I think we are making a difference is in our ability to enable 400G to be mainstream. 

SV: Actually, there is another area that may be relevant. The drive to co-packaged optics is in reducing power in the switch ASIC [application-specific integrated circuit] console. You basically eliminate repeaters and a lot of the really power-hungry chips by bringing the optics closer and closer to the source of data. That’s something our technology does.

So, co-packaged optics can be a green technology because it is essential and it reduces power. But POET is not the only way to get to co-packaged optics. The question then is, does POET provide any incremental benefit to power reduction compared to some other co-packaged optics solution in terms of helping the environment? That’s a question we would have to dig into. What we are bringing in is a cost-effective way of doing co-packaged optics and perhaps a little more elegant way, too. 

Note: Author Adrian Brijbassi has previously written about POET Technologies for the Globe & Mail and has since become an investor in the company.

< Back to Blog