Translation: SIVE seems to play a dual role as both a "bottleneck" and a "bottleneck" in the CPO field next year. Continuously seeing information released by non-technical individuals, completely ignoring the differences in details. The reasons are as follows: 1、 The CW laser has a bottleneck, and LITE's financial report has sent a signal. The significant allocation of laser wafer fab capacity to EML is likely due to the previous contract with NVDA. Sumitomo/ancient river=bottleneck → Win Semi=bottleneck SIVE adopts fab lite mode, is it a bottleneck? Yes. SIVE is in a laser bottleneck because it has been controlling the final supply of CW lasers from Win Semi and other wafer fabs through allocation mechanisms since a very early stage (the CEO has stated that it is working with more production capacity partners). The most typical analogy is Armor/Disk: SNDK controls the output of NAND, thus gaining ultimate pricing power and becoming a bottleneck. The demand from Ayar, Jabil, and other plug-in solution providers, combined with the demand from the Nvidia NVLink CPO ecosystem, has exceeded the supply. The final supply of CW lasers is controlled by SIVE, which makes Sivers a bottleneck. MRVL Celestial and other large-scale cloud vendors are the main and even the only suppliers of ASIC/commercial CPO solutions. Therefore, it is impossible to bypass (Sivers' single channel CW laser cannot be hot swappable). 2、 SIVE is the bottleneck of the entire CPO ecosystem. LITE (the latter is likely to source CW laser production capacity from Japanese competitors). AVGO is likely to have also achieved vertical integration. However, the entire ecosystem surrounding CPO, including ASIC projects (such as Marvel and AlChip) and commercial projects (such as Ayar, Lightmatter, and Lightelligence), is highly likely to be designed around SIVE. Taking Ayar as an example: They attempted to procure multiple sources through MTSI/LITE around 2022, but these manufacturers' lasers may not meet Sivers' standards in array specifications (Lumentum and Macom have recently been removed from Ayar's supply chain page). If there is no alternative solution in at least the first generation of products (although they are clearly striving for multi-source), then SIVE is a structural bottleneck that cannot be bypassed in the laser link. Even the 1.6T LRO designed by JBL is likely to have a performance advantage based on the "deep moat" built exclusively from SIVE. AMD's first generation solution is likely to use Sivers and may also be combined with Ayar. If every major participant (excluding Nvidia/Roadcom) who has not yet achieved vertical integration relies on Sivers on CPO Then it's just a bottleneck. Just take a look at the partners in the entire NVDA NVLink CPO ecosystem: almost all of them are likely using Sivers. And they are also using GFS (Sivers is their default reference design). Therefore, when CPO is truly implemented on a large scale in the second half of 2027, SIVE will become both a bottleneck and a bottleneck - and this is precisely one of the most significant architectural transformations in history (according to GS research reports, TAM will jump from almost zero to $81 billion or even $91 billion in the next year and a half). This is precisely why I believe SIVE has the potential to become the next LITE with a market value of $75 billion in the coming years. All of this should start to be realized next year. And its current valuation is even lower than a company that purchases Sivers lasers for secondary packaging and holds a $50 million procurement agreement.