Suresh mentioned that it is easier to achieve market share in the early going. This makes sense, as most companies that need, say, and 800G module will do a search for an available supplier. They will want to check the price, the reputation of the supplier, the capacity to deliver in volume, and the viability of an ongoing relationship (a technology pathway) among other factors. 

It is a more open competition than just buying incrementally upgraded, non-differentiated parts. 

This is where the merits of a new product can shine. If the other factors (reputation, volume capability, tech pathway) align, and price, quality, and availability are exemplary, the market share can grow quickly. 

I looked at NVIDIA, the obvious leader in GPUs for AI. They are at 94% for AI applications, with an expectation of dropping to 75% as competition rises. This is the obvious example..

Qualcomm, a former upstart that has settled into a competitive environment in handset chips, is still 23%, behind MediaTek at 40%. Apple is 17% in this sector. 

What I expect from POET is higher than is publicly being stated, and many here don't expect, is because they have the advantages inherent in the OI. It is wafer based, not hand made; it is made for customizing, has better waveguides, thermal laser mounts, nifty fiber attach points, and the rest of the list we see over and over.

Given all this, how does the competition survive the clean sheet vendor selections for 800G? POET just filled in the size worry by working with Foxconn and Luxshare. MultiLane handles the customizing well. Suresh is working on a China+1 manufacturing partner, this is another kep point (which remains to be set right). 

I am hoping for some surprisingly high market share numbers as this develops. I also expect some royalty arrangements for the clever thing built into POET's products. There is a definite level of speed to market these data center improvements. This bodes well for POET, and its advantages. Even if the others can eventually compete, there is a premium on speed. There is no half a decade to waste on catching up. The penalties are too high for failure to keep up. The balance of system parts are really expensive, think NVIDIA processors at $40k each. There is no need to add a weak link to these systems and no time to catch up.