Gambit kayan aikin AI: Wasan gani na Marvell da Bashi-Bashi na Silicon Valley

Gambit Kayan Kayan Aiki na AI: Wasan gani na Marvell da makomar Bashi na Silicon Valley

Kalanda ya juye zuwa Nuwamba 2026, kuma yanayin AI wani babban ci gaba ne na ƙima da sake kimanta kuɗi. Yayin da manyan nau'ikan harshe kamar Google's Gemini 4 Argon da Carbon ke tura iyakokin abin da zai yiwu Google Blog, tushen abubuwan more rayuwa da ke tallafawa wannan juyin yana fuskantar babban canji, sauyi biyu. Muna shaida dabarun haɗin gwiwa a cikin kayan aiki, wanda aka misalta ta hanyar fare na Marvell akan Lumentum, da kuma babban sauyi kan yadda Silicon Valley ke ba da kuɗin sha'awar ƙididdigewa. Waɗannan ba keɓantacce ba ne; alamu ne na masana'anta na lissafin duniya da aka sake gina su kuma an sake dawo da su a sikelin da ba a taɓa gani ba.

Marvell's Lumentum Bet: Kashin bayan gani na AI

Abubuwan da ake buƙata na lissafi na AI na zamani, musamman don horo da ƙima a sikelin, sun tura hanyoyin haɗin yanar gizo na gargajiya zuwa iyakar su. Kamar yadda samfuran AI ke girma da ƙarfi, haka ma buƙatar babban bandwidth, sadarwa mara ƙarancin latency tsakanin GPUs, DPUs, da wuraren ajiyar ajiya. Hanyoyin haɗin wutar lantarki, yayin da suke a ko'ina, suna fuskantar ƙaƙƙarfan matsi na zahiri: lalata mutuncin sigina akan nisa, ƙara yawan amfani da wutar lantarki tare da mafi girman ƙimar bayanai, da iyakoki masu yawa. Wannan shine inda na'urorin gani ba kawai fa'ida ba, amma mahimmanci.

Samun Marvell na Lumentum, motsi CNBC ya bayyana a matsayin "daya daga cikin manyan fare akan AI a cikin 2026" [CNBC] Marvell, wanda ya riga ya zama gidan wutar lantarki a cikin sassan sarrafa bayanai (DPUs), masu sauya hanyar sadarwa, da ASICs na al'ada, sun gane cewa mallakar ƙirar gani yana da mahimmanci don isar da aikin ƙarshe zuwa ƙarshe a cikin gungu na AI na gaba. Lumentum yana kawo babban fayil mai zurfi a cikin kayan aikin gani, gami da masu saurin sauri, ingantattun na'urorin gani, da fasahar laser na ci gaba masu mahimmanci ga silicon photonics.

Yi la'akari da tsarin gine-gine na zamani na AI supercluster. Dubban GPUs, kowane dabbar lissafi, dole ne suyi sadarwa tare da juna kuma tare da manyan bankunan ƙwaƙwalwar ajiya. Alamun jan ƙarfe na al'ada da igiyoyin gani masu aiki (AOCs) suna gwagwarmaya don yin ƙima da kyau fiye da ƴan mita a 800Gb/s ko 1.6Tb/s. Kasafin kuɗin wutar lantarki don siginar lantarki ya zama haramun, kuma sawun jiki na manyan igiyoyi da sauri yana cinye sararin tara mai ƙima.

Wannan shine inda fasaha kamar CXL (Compute Express Link) akan na'urorin gani da NVLink akan na'urorin gani suka zama masu canza wasa. Ta hanyar tsawaita waɗannan manyan haɗe-haɗe-haɗe-haɗe-haɗe-haɗe-haɗe-haɗe-haɗe-haɗe-haɗe-haɗe-haɗe-haɗe-haɗe-haɗe-haɗe ta amfani da filaye na gani,masu gine-ginen AI na iya gina haƙiƙanin rarrabuwa da mahalli mai ƙididdigewa. Ƙwarewar Lumentum a cikin EMLs (Laser-absorption Modulated Lasers) da DMLs (Masu Gyaran Laser Kai tsaye) don aikace-aikacen gajeriyar isar da saƙo, tare da madaidaitan DSPs don haɗin cibiyar bayanai mai tsayi, kai tsaye yana magance waɗannan ƙullun.

Haɗuwa da ƙarfin gani na Lumentum zuwa cikin Marvell's DPU da canza siliki zai ba da damar sabon nau'in haɗin kai sosai, mu'amalar hanyar sadarwa mai ƙarfi. Ka yi tunanin DPU tare da haɗe-haɗe na silicon photonics, kai tsaye tana jujjuya siginar lantarki zuwa na gani a cikin kunshin guntu, yana kawar da madaidaitan masu amfani da wutar lantarki da ke da alaƙa da latency sama. Wannan matakin haɗin kai yana da mahimmanci ga tsararraki na gaba na gungu na AI, inda kowane picosecond da milliwatt ke ƙidaya.

# Misali: Sauƙaƙe Tsarin Fabric Network don Tarin AI
# Nuna ƙayyadaddun tsarin haɗin kai na gani
network_fabric:
  suna: "AI_Supercluster_Fabric_01"
  topology: "fat_itace"
  spine_switches:
    - id: "kashin baya-01"
      samfurin: "Marvell_Teralynx_9_Optic"
      tashar jiragen ruwa:
        - irin: "QSFP-DD800"
          kafofin watsa labarai: "Optical"
          gudun: "800Gbps"
          link_to: "leaf-01/port-1"
          yarjejeniya: "CXL_over_Optics"
        - irin: "QSFP-DD800"
          kafofin watsa labarai: "Optical"
          gudun: "800Gbps"
          link_to: "leaf-02/port-1"
          yarjejeniya: "NVLink_over_Optics"
  leaf_switches:
    - id: "leaf-01"
      samfurin: "Marvell_Teralynx_9_Optic"
      tashar jiragen ruwa:
        - irin: "QSFP-DD800"
          kafofin watsa labarai: "Optical"
          gudun: "800Gbps"
          link_to: "gpu-server-001/nic-0"
          yarjejeniya: "CXL_over_Optics"
        - irin: "QSFP-DD800"
          kafofin watsa labarai: "Optical"
          gudun: "800Gbps"
          link_to: "gpu-server-002/nic-0"
          yarjejeniya: "CXL_over_Optics"

💡 Maɓallin Takeaway: Haɗin haɗin kai ba shine mafi kyawun mafita ba; suna da tushe don haɓaka kayan aikin AI yadda ya kamata. Sayen Lumentum na Marvell babban dabara ne don mallakar wannan mahimmin Layer, yana ba da damar mafi girman bandwidth, ƙarancin latency, da ingantaccen ingantaccen ƙarfin wutar lantarki don manyan gungun AI na gaba. Wannan kai tsaye yana tasiri rufin gine-gine don rikitaccen samfurin AI da sikelin horo.

Sabon Littafin Wasa na Silicon Valley: Tech Giants Tap Kasuwan Bashi don AI & Girman gajimare

Yayin da Marvell ke yin wasan kwaikwayo na kayan masarufi, tsarin gine-ginen kuɗi wanda ke ƙarƙashin duk haɓakar AI shima yana fuskantar canji mai ma'ana. Historically, Silicon Valley's tech giants, flush with cash from their highly profitable software and advertising businesses, largely self-funded their growth. Gudun kuɗaɗen cikin gida shine babban injin faɗaɗawa. Koyaya, kamar yadda kamfanin dillancin labarai na Reuters ya ruwaito a ranar 2 ga Nuwamba, 2026, manyan kamfanonin fasaha na duniya yanzu suna "taɓatar da kasuwannin basussuka da haɓaka daidaito don haɓaka abubuwan more rayuwa na AI" [Reuters] Wannan yana nuna alamar tashi mai mahimmanci.

Me yasa pivot? Amsar ta ta'allaka ne a cikin ɗimbin kuɗaɗen kuɗaɗen da ba a taɓa gani ba da ake buƙata don ci gaban AI mai iyaka. Horar da LLM na zamani guda ɗaya kamar Carbon Gemini 4 na Google ko haɓaka hadaddun wakilan AI suna buƙatar ba dubbai kawai ba, amma sau da yawa dubun ko ɗaruruwan GPUs masu girma, ƙwararrun masu haɓakawa, da nagartaccen hanyar sadarwa da kayan aikin sanyaya don tallafa musu. Kudin babban jigon AI guda ɗaya zai iya shiga cikin biliyoyin daloli cikin sauƙi.

Yi la'akari da ilimin tattalin arziki:

1. Hardware Refresh Cycles: AI accelerators suna tasowa cikin sauri. NVIDIA, AMD (duk da kwanan nan kasuwar PC jitters 247wallsts kowane watanni 8 masu haɓakawa. Tsayawa gasa na buƙatar ci gaba, babban saka hannun jari.
2. Yin Amfani da Makamashi: Yin iko da sanyaya waɗannan gungu yana da tsada ta fuskar taurari. Cibiyoyin bayanai suna zama masana'antar wutar lantarki.
3. Lokaci-zuwa-Kasuwa: tseren tura ayyukan AI masu gasa yana da zafi. Jiran tara tsabar kuɗi na ciki yana nufin rasa ƙasa ga abokan hamayya. Bashi yana bada alluran jari nan take.

Daga hangen nesa na injiniya, wannan jigon kuɗi yana da tasiri mai zurfi. Zamanin "kasafin kuɗi mara iyaka" don kayan aikin gwaji na AI ya ƙare. Lokacin da aka haɓaka haɓaka ta hanyar bashi, kowane dala da aka kashe yana zuwa tare da babban matakin bincike akan riba da dawowa kan saka hannun jari (ROI).

Wannan yana nufin:

• ** Haɓaka Haɓaka: *** Masu gine-ginen kayan aikin suna ƙarƙashin matsananciyar matsa lamba don ƙira don ƙimar inganci - ba kawai aiki ba, amma ingancin wutar lantarki, amfani da sararin samaniya, da ingancin sanyaya. Kowane watt da aka ajiye yana fassara kai tsaye zuwa rage kashe kuɗin aiki da ingantaccen ROI.
• ** Zane na Modular & Sikeli:** Ikon haɓaka abubuwan more rayuwa da ƙari da tsinkaya ya zama babba. Makullin mai siyarwa, yayin da koyaushe damuwa, ya zama haɗarin kuɗi mafi girma lokacin da jadawalin biyan bashi ya ƙare. Buɗe ƙa'idodi da ƙirar ƙira waɗanda ke ba da izinin haɓaka abubuwan haɓakawa ba tare da cikakken tsarin tsarin ba suna da fifiko.
• **Tsarin TCO Binciken: *** Ƙungiyoyin injiniya dole ne su zama masu ƙwarewa a cikin Jimlar Kudin Mallaka (TCO), ƙididdigewa ba kawai CAPEX (hardware, ginawa) ba har ma OPEX (ikon, sanyaya, kulawa, raguwa, da mahimmanci, biyan kuɗi akan bashi).

--------------------
|   Kasuwannin Bashi | --> |   Tech Giants' | --> | AI Kayayyakin Kayayyakin |
| (Bonds, Equity) |     |   Babban Shafi |     | (GPUs, Optics, |
+------------------+ +------------------+ |   Cibiyoyin Bayanai) |
        ^ |
        |                                                   v
        -----------------------------------------------------------------+
                                 |
                                 v
                       ---------------
                       |   AI Services & R&D |
                       | (LLMs, Agents, Cloud) |
                       ---------------
                                 |
                                 v
                       ---------------
                       |    Haraji & Riba |
                       | (Biyan kuɗi, Talla) |
                       ---------------
                                 |
                                 v
                       ---------------
                       |  Biyan Bashi & ROI |
                       ---------------

⚠️ ** Hatsarin Gine-gine:** Abubuwan da aka biya bashi suna buƙatar amfani da albarkatu masu inganci da dabarun samun kuɗi cikin sauri. Ƙungiyoyin injiniya dole ne su ba da fifikon ƙira waɗanda ke ba da fayyace, ROI masu iya aunawa da kuma rage yawan kashe kuɗi na aiki, yayin da ƙarfin kuɗin kuɗi yana ƙaruwa ga kowane yanke shawara na gine-gine. Kudin babban birnin yanzu shine shigarwa kai tsaye zuwa ƙirar kayan aikin.

Muhimmancin Injiniya

Haɗin gwiwar gambit na gani na Marvell da jigon Silicon Valley zuwa kasuwannin bashi yana ba da hoto sarai: ababen more rayuwa na AI suna shiga wani sabon lokaci, balagagge, da tsarin kuɗi. Kwanakin jefa lissafin kawai a cikin matsala suna ci gaba zuwa wani zamani inda kowane zaɓi na gine-gine, daga nau'in transceiver zuwa mafita mai sanyaya, dole ne ya zama barata ta hanyar gudummawar da yake bayarwa ga aiki, ingancin wutar lantarki, kuma a ƙarshe, komawa kan zuba jari.

Ga ƙungiyoyin injiniyanci, wannan yana nufin ƙarin mayar da hankali kan:

1. ** Zurfafa fahimtar Layer na zahiri: ** Rungumar hanyar sadarwar gani. Koyi game da silicon photonics, na'urorin gani masu daidaituwa, da haɗarsu cikin masana'anta. Matsalolin suna canzawa daga hawan CPU zuwa haɗin haɗin bandwidth da latency.
2. ** Jimlar Kudin Mallaka (TCO) a matsayin ƙaƙƙarfan ƙira na farko:** Matsa fiye da ma'aunin aikin ɗanyen aiki. Yi la'akari da mafita dangane da amfani da wutar lantarki kowane aiki, buƙatun sanyaya, da farashin aiki na dogon lokaci.
3. ** Modularity da Buɗaɗɗen Matsayi: ** Zane don sassauci. Guji kulle-kulle mai zurfi mai siyarwa a inda zai yiwu, saboda yana iyakance haɓakawa gaba da damar rage farashi.
4. Karancin Kudi: Fahimtar yadda ake tura jari da kuma matsalolin kuɗi akan ƙungiyar ku. Hukunce-hukuncen gine-ginen ku suna tasiri kai tsaye ga ikon kamfani na ba da bashi da kuma samar da riba.

Lokaci na gaba na ci gaban AI ba kawai zai kasance game da ci gaban algorithmic ba; za a ayyana shi ta ƙarfin aikin injiniya don ginawa da sarrafa abubuwan da ke cikin ƙasa tare da ingantaccen aiki wanda ba a taɓa ganin irinsa ba da kuma kula da kuɗi. Hannun jari sun fi girma, kuma ƙalubalen injiniya sun fi rikitarwa fiye da kowane lokaci.

Sources & Karin Karatu

• [Masu ɗaukan Marvell alama ce ta ɗaya daga cikin mahimman fare akan AI a cikin 2026]
• [Kamfanonin fasaha mafi girma a duniya suna yin kasuwancin basussuka da haɓaka daidaito don haɓaka ababen more rayuwa na AI]
• Sabuwar ci gaban AI na Google, gami da Gemini 4 Argon, Carbon, da Genie 4 - Yana nuna ma'aunin buƙatun abubuwan ci gaba na AI.
• AMD Slides 3% Gaban Samun Samun Kamar yadda PC ke damun PC; Intel Drops 5%, NVIDIA Edges Higher - Yana ba da mafi girman mahallin kasuwa don kashe kayan masarufi.
• Labarin AI A Yau Babban Labari: Buga Baƙo akan Terence Tao's Blog Pi Pitches 'Mathematics Egoless' azaman Amsar Ladabi ga Masu Haɗin gwiwar AI - Babban mahallin labarai na AI na Nuwamba 2026.