Barrier lowered for photonics innovation

In late January, Luxtera announced that it is opening its silicon photonics process and device-design library to the Optoelectronic Systems Integration in Silicon (OpSIS) foundry shuttle service for use by third parties. A shuttle service allows multiple low-volume customers to split the high cost of wafer runs. We believe this move represents an interesting opening, particularly for development of high-performance telecom components.

Another announcement the same week demonstrates the need for component innovation in advanced long-haul transport where volumes are initially low. 100G is challenged by cost, performance, and power consumption. Optelian has acquired Versawave, a maker of high-speed optical modulators. An investment into innovative optical components is a surprising move by any OEM, but particularly Optelian, which has maintained profitability through a tight focus on immediate business with tier-2 carriers. Optelian cites Versawave’s research interactions with OpSIS as part of the acquisition target’s attractiveness.

100G represents a game-changer in more ways than one

Optical integration proponents have been waiting for an opportunity like the current 100G opening since before the “killer app” cliché came and went. Much has already been written about how integration addresses 100G’s need for complex modulators and power reduction. We believe Optelian’s interest in 100G, as surprising to us as its response, speaks to a different angle. Traffic patterns and quantities are changing. 100G demand is springing up from multiple drivers, including data centers and core network pipes. The demand cannot be met by waiting for the trickle-down of technology developed for backbone networks. Aside from the acquisition, Optelian has also substantially increased its internal R&D budget. This is new behavior for an optical transport vendor that has prided itself on its profitability since its inception in 2002. It has expended resources only where there was direct customer pull. It skipped 40G but now says its customers, which have not been the type to invest ahead of demand, are asking for 100G.

Optelian’s interest in having Versawave’s technology in-house speaks to something like an ASIC model where a system OEM cannot find the sort of components it needs on the open market – at least initially. But total market unit volumes for transport are modest; for a company the size of Optelian to have its own component supply, there must be a cost-effective way to make low volumes.

Shuttle service goes beyond foundry to lower barriers to entry

The collaboration of Luxtera and OpSIS creates a model whereby even research volumes of optoelectronics can be fabricated cost-effectively but with commercial quality. This opens up the affordability of university and corporate R&D as well as a quick transfer to manufacturing in the sorts of volumes suitable for high-end telecom components.

Optics has historically relied on proprietary processes; hence, large vendors that can afford their own fabs have an advantage. The CMOS industry is oriented towards even higher volumes, but also has developed foundry and shuttle business models – such as the well-established MOSIS service for VLSI electronics – to deal with low-volume needs. The Luxtera-OpSIS partnership applies these models to silicon photonics.

Foundries eliminate the high initial cost of building a plant, but not the operating cost per run. Luxtera itself is fabless, relying on a commercial CMOS foundry. OpSIS, a nonprofit center run at a university, takes the cost reduction a step further by offering the shuttle service, which combines multiple customers’ designs into a single maskset and run. OpSIS estimates that in the best case, a customer could use as little as 1/100 of the wafer and hence incur a proportional cost. The minimum entry level charge is $15,000, on par with prototyping costs in VLSI. Ramp-up to modest product volumes is straightforward, as a single run can produce thousands of chips.

While OpSIS already had two other foundry processes in its toolkit, Luxtera’s is the only that integrates electronics with optics and is proven in the commercial market. It has developed, for its own use, the process and design library that allows optics and electronics to be combined in the same chip. It considers this toolkit mature, having shipped close to a million 10G channels. Through OpSIS, it will now license the use of its toolkit to third parties as a secondary revenue stream. It also hopes to develop an ecosystem of designers and tool developers to push its approach towards the mainstream.

OpSIS hopes to spur a wave of innovation such as has been seen in fabless electronics. Versawave is one such player. While its commercial modulators are in GaAs, it is interested in exploring silicon photonics for further integration, particularly with electronics. At high speeds, where Versawave plays, integrating the driver electronics with the modulator is particularly beneficial.