Regarding $AMZN RNG + OpenAI’s MRC and VCSELs. In what appears to be an indirect response on B. Riley's $AAOI sell report regarding optimizations that "flattens networks and cuts transceiver TAM 40-50%". Rosenblatt TLDR: They've already modeled for this, even though each XPU may use fewer transceivers. The number of XPUs is growing so quickly that total optical demand should still rise. On lasers architectures for CPO: VCSELs are useful as a credible bridge for NPO/short reach. Ideal architecture though for CPO is UHP CW over other lasers, and cites Broadcom for that statements. Which happens to supports my core $SIVE thesis. I feel like Rosenblatt and I share the same views on a lot of things, I like the way they roll.

@HayatiBTC @IChukwuwem46425 @Varshan_ts @btc314et @nfthree3 @Buivanp61273386 @luffyluffy54 @West86120064 @mitul_manish23 @ilkerduman90 @ash189596 @7h3r04d @OdinBHP @vlad_mrc @jmacc535 @CdukeyNft @mackharon59 @AdrianBrz_ @MaxKabanoff @ikigaiwif606 @anwarka36397203 @defipaca @Crypt0Vampyre @ootzs @ddaavviiddrr @Bigdawgg69 @tate77036 @limuelborjz1992 @0xRiver8 @AghiliHass11340 @Yashwardhan_G @MooreJared0 @Bouncer235 @chuchu42800416 @metababe_nft @gsamasol @swiniurainvests @davidno5uwu @slamsaint69 @shayan_far90294 @danktrader420 @ritaluv999 @AktieYoungblood @lahnovic16487 @1investimento @HippsyNFT @Scoobie366 @FelipeRibe64815 @cryp1999 @Humpolicious01

love when @OpenAI is open. incredible contribution to @OpenComputePrj 1. multiplane design for 800GbE. 2 tiers for 131k endpoints 2. custom MRC protocol that extends RoCE to support packet spraying 3. switch from dynamic to static routing with SRv6

Scaling AI isn’t just about compute—it’s about the network. Congratulations to the @OpenComputePrj community for the addition of the new Multipath Reliable Connection protocol! With MRC, Intel is helping to enable multi‑plane Ethernet fabrics that grow to massive scale using fewer switch tiers, lower power, and higher reliability. This is how frontier AI moves together. More:

🖼️ 24 HOURS OF ART 📊 August 26, 2026 🎨 Pictured: 24 Hour Highlights - 'die with the most likes (redistribution)' by @toadswiback - 'Grifter #236#' by @XCOPYART, collected by @F1rstclassJG - 'The First Sparks of Artificial Creativity (Digital Edition)' by @VanArman - 'GANdinsky 3D - Nothing Natural - Variant 01' by @Coldie - 'Argonauts' by @lphaCentauriKid, part of 1/1/9,999 digital + physical collection mint out with @MuseFacktory - 'HOLY #481#' by @wubbushi, collected by @mrc_arte, part of 500 artwork collection mint out with @verse_works - 'Everything, made simple' by @zancan, collected by @thisistolo - '9 Lives' by @money_alotta, collected by @nft_taz - 'Eden' by @hanrgb, collected by @zaphodok

At AI scale, raw bandwidth breaks down. What matters is resilience, recovery, and consistency under load. AMD in collaboration with Microsoft and @OpenAI defines a proven solution approach to AI networking with MRC. Learn more now: