Lightwave Logic Inc (LWLG)

[KCCO7913]

Relevant...a copy of a recent Reddit comment of mine:

Great question. I’ve done a little bit of research on this the last few months.

Per Ashkan Seyedi at NVIDIA recently, today’s 800G transceivers consume 15-20 Watts of power. The industry is doing everything they can do to get that power consumption down. For 1.6T, the wattage increases further, and I believe I saw up to 50 Watts estimated for future 3.2T pluggable transceivers using today’s technology.

A lot of the focus today is on design changes, mostly LPO, which removes the DSP from the transceiver. This year it became a bit clearer that the industry appears to be hesitant to go the LPO route and now are talking about LRO – linear receive optics. Below is a quick Google link on the LPO vs LRO topic:

https://fast-photonics.com/linear-pluggable-optics-and-linear-receive-optics/

Andy Bechtolsheim from Arista has mentioned many times over the last 2 years that the other ways to reduce power consumption are to switch to better modulator technology and use lower power lasers. He has mentioned that better modulators can offer 20% or more in power savings versus LPO.

With that said and moving along…I am hoping/estimating we’ll see an EOP modulator based 800G transceiver operate under 10 Watts. I’d love to see 7-8 Watts.

Modulators themselves don’t consume a significant amount of the transceiver’s power. But these highly efficient modulators enable lower power lasers to be used and enable lower power DSPs and/or eliminate modulator driver chips altogether in the design.

Overall, it is strictly the use of these better modulators that offers drastic power savings because of the ancillary effects they have. They are going to enable LPO levels of power efficiency without having to go the LPO route.

Now to quantify this let’s look at the typical energy usage of a hyperscale datacenter. I used ChatGPT in some of my fact-finding FYI.

A typical hyperscale data center uses up to 50% of its total power for data processing and transmitting infrastructure.

Optical transceivers are a small portion of that and consume up to 5% of the total datacenter energy.

There are several categories of optical transceivers depending on distance needed. Initially and of course, not all optical transceivers inside a datacenter would use LWLG’s modulators. But it won’t be too long before much of the market is dominated by 800G+ transceivers in all categories which potentially can be served by LWLG’s EOP modulators. IMO…the sweet spot for LWLG is the 2k+ distance. But anything over 500m is likely the start of where LWLG could operate. Beyond 10k and up to 80k…the modulation is coherent and today is wildly power hungry and expensive. Tons of value to be unlocked there and EOP is already proven in that wavelength and modulation format.

Let’s say EOP based transceivers shave 2% off the total datacenter power consumption from optical transceivers. Gives us some wiggle room and assumes some types of transceivers will not use the ‘latest and greatest’ modulator technology.

The average hyperscale data center (50MW) has an average power bill of $72 million per year. ChatGPT says ‘it depends’ but an average of $6 million per month.

2% of that total is $1,440,000. Google operates 39 data centers and Amazon has 125 according to a quick search. At the end of 2023, there were almost 1,000 hyperscale datacenters worldwide with half of them in the USA. The number won't stop growing.

EOP modulators appear to have potential to save Google and Amazon a collective $236 million per year in energy costs. The future projections in a world of 1.6T/3.2T modules are significantly more.

Anyway…these datacenter upgrades to 800G and beyond are happening no matter what. LWLG has a very solid case to be enabling the most efficient and scalable modulator platform to power the future pluggable optics. EOP is far more power efficient than SiPh and certainly InP. It is also more efficient than TFLN. It is customizable for continued improvement. Above everything…it is the only emerging material platform that is totally compatible with silicon foundries.

It doesn’t stop at pluggables either. EOP will make its way into co-packaged optics and the power savings are amplified even further there. Lower power SerDes, cooler operation/less cooling requirements, etc. So it’ll chop off more of that 5% segment mentioned above plus into other sections of networking infrastructure.

Please anyone try to find some holes in my analysis here.