Kutsengwa kwe-Hugging Face yi-Nvidia nga-$12.9 bhiliyoni: Kusho kutsini emabhizinisini lakha etinhlelo tawo ku-AI levulekile

Imphendvulo kucala: mhla ziyi-26 nangeli-27 Ncwaba 2026, imibiko yetindzaba yatsi i-Nvidia beseyifinyelele esivumelwaneni sekutsenga i-Hugging Face cishe nga-$12.9 bhiliyoni. I-Hugging Face yinkundla labatfutfukisi emhlabeni wonkhe labayisebentisela kushicilela, kulanda nekuhlola ema-model e-AI lanetisindvo letivulekile. Lesivumelwano asikamenyetelwa ngalokusemtsetfweni, kute sivumelwano sekugcina lebesesisayiniwe ngesikhatsi kuphuma lemibiko, futsi kungenteka singaze siphotfulwe. Kodvwa kungenteka nje kwaloku kufanele kunakwe yinoma nguyiphi inkampani lesebentisa, noma lecabanga kusebentisa, ema-model e-AI etinhlelweni tayo kunekutsi itsembele ngalokuphelele kuma-API lavalekile.

Lokwenteke ngempela

I-The Information yabika ngelusuku lwaLesitsatfu kusihlwa, mhla ziyi-26 Ncwaba 2026, kutsi i-Nvidia beseyivumile kutsenga i-Hugging Face nga-$12.9 bhiliyoni. I-CNBC, i-Fortune, i-Forbes, i-TechCrunch ne-Bloomberg tayibika lendzaba. Lenani libeka linani lenkampani ngetulu kwa-$13 bhiliyoni, lokungayenta kube kutsenga lokukhulu kakhulu emlandvweni we-Nvidia nangabe kuphotfulwa.

Tinombolo letikhona ngenkampani lefunwako tichaza bukhulu balolugcuma. I-Hugging Face yacoca $235 miliyoni emzuliswaneni wekutfola timali nga-2023 ngelinani lelingu-$4.5 bhiliyoni, futsi imibiko yamuhla ibeka imali leyingenisako ngemnyaka cishe ku-$150 miliyoni, ikhuphuka isuka cishe ku-$100 miliyoni etinyangeni letimbili ngaphambilini. Ngako-ke intsengo lehlongotwako iphindzaphindza kakhulu imali leyingenako, lokukhombisa kutsi loko i-Nvidia lekutsengako akusiyo incwadzi yemali leyingenako, kodvwa yindzawo emakethe.

Kuyaphawuleka futsi kutsi ekupheleni kwa-2025 i-Nvidia yanikela kutsi itawutfutfukisa ngemali lengu-$500 miliyoni ngelinani lelingu-$7 bhiliyoni, i-Hugging Face yala kuze igweme kutsembela kakhulu kumtfutfukisi munye. Umehluko emkhatsini walemininikelo lemibili esikhatsini lesingaphansi kwemnyaka ukhombisa ngalokucacile kutsi ingcenye yema-model lavulekile ikhule ngekushesha kangakanani ekubalulekeni.

Siyeleliso lesidzingekako: imitfombo ichaza kutsi tingcoco atikaholeli esivumelwaneni lesisayiniwe, futsi iphawula kutsi lesivumelwano kusengenteka singaphumeleli. Ungasekeli sincumo sekusebenta esivumelwaneni lesingakamenyetelwa ngalokusemtsetfweni.

Kungani inkampani yema-chip ingakhokha lokungaka ngenkundla yema-model?

Umcondvo webhizinisi uconde kakhulu kunaloko lobonakala ngakhona. Ema-model lanetisindvo letivulekile avala ngekushesha sikhala lesiwehlukanisa netinhlelo letivalekile, futsi noma ngubani lolanda i-model levulekile udzinga i-hardware kuze ayisebentise — etimeni letinyenti kakhulu, ema-GPU e-Nvidia. Kuba ngumnikati wendzawo lapho luhambo lwemtfutfukisi lucala khona kusho kuba nemtselela esincumweni setobuchwepheshe singakentiwa.

Ngetulu kwaloko kunetinhlangotsi letimbili letengetiwe: kubuyela emakethe ye-cloud computing ngetinhlelo letikhona te-Hugging Face kanye nekukhona kutsengisa emandla ekubala langasetjentiswa, nekuvikela sikhundla se-Nvidia kubaphikisi labacale kudizayina ema-chip abo kuze banciphise kutsembela kuyo. Ngalombono, lokutsenga akusiko kwandziswa kwemkhicito; kucinisa umzila wekusabalalisa.

Yini lehlukile ngekwemsebenti ebhizinisini lakho?

Nangabe usebentisa sistimu letsembele ku-model levulekile lebekwe endzaweni yakho, kufanele uhlolisise tintfo letine:

  • Lokukhona lamuhla kuhlala kutfolakala. Tisindvo leshicilelwe ngaphansi kwemalayisensi lanjenge-Apache 2.0 noma i-MIT, ilayisensi yato ingeke yesulwe ngekubuyela emuva kunoma ngubani lose atilandile. Insiphela lonayo lamuhla ihlala isemtsetfweni kuyisebentisa kusasa, noma ngubani lonenkundla.
  • Kuphatfwa kungashintja ngekuhamba kwesikhatsi. Tintfo letibekwa phambili yinkundla lephetfwe ngumkhicitsi we-hardware akusho kutsi tifana naleto tenkundla lesimeme yodvwa: loko lokuvetwa embili, ngumaphi emathulusi latfutfukiswa kucala, nekutsi nguyiphi i-hardware architecture letfola kusekelwa lokuncono.
  • Tindleko tekusebenta tingehla ku-hardware letsite. Umsebenti wekusebenta kahle nekusheshisa ungenteka kakhulu kutsi utawulungiselelwa kucala i-stack ye-Nvidia — kuyinzuzo nawusebenta kuyo, kuyinhlawulo lencane nawuncike kulenye indlela.
  • Kuhlangana kwebahlinzeki kuyakhula. Kuba nema-chip, inkundla nerejista yema-model ngaphansi kwelufulela lunye kwengeta lelinye liphuzu linye lekwehluleka emzileni wakho wekuhlinzekwa kwetobuchwepheshe.

Tingoti lokufanele utibale

Ingoti yekucala kuboshelwa encenyeni yemathulusi kunasencenyeni ye-model. Emacembu lamanyenti abhala ikhodi leboshelwe ngco kuma-library nakuma-interface enkundla yinye, ngako kushintja umtfombo ngemuva kuba ngumsebenti wekubhala kabusha kunekushintja nje i-URL. Yesibili ingeyetemtsetfo: sivumelwano lesikhulu kangaka emakethe lebucayi sibhekana nekuhlolwa lokungasibambezela noma kusilolonge kabusha tinyanga. Yesitsatfu yimbi kakhulu kuto tonkhe — kwakha lusu lwemnyaka wonkhe endzabeni lengakaciniswa ngalokusemtsetfweni.

Tinyatselo letisihlanu letisebentako kulelisontfo

Kute naku sinye lesidzinga sabelomali, futsi kute naku sinye lesilindze umphumela walesivumelwano:

  • Bhala luhlu lwekutsembela kwakho kwangempela. Bhala ema-model, ema-library netinsita tangaphandle sistimu yakho letsembele kuto ngempela, nekutsi ngasinye sitsatfwa kuphi ngesikhatsi sekusebenta.
  • Gcina tinsiphela takho. Gcina insiphela yetisindvo tema-model lotsembele kuto ngekhatsi etinhlelweni takho, ne-checksum yalo lonkhe lifayela, futsi ungavumeli kusebenta kwe-production kutsi kutsembele ekulandeni lokubukhoma enkundleni yangaphandle.
  • Buketa emalayisensi bese ugcina umbhalo. Gcina umbhalo welayisensi lebewusebenta ngesikhatsi sekulanda eceleni kwayo yonkhe i-model loyisebentisako; lowo ngumtfombo wakho wetemtsetfo esikhatsini lesitako.
  • Hlukanisa ingcenye ye-inference. Yendlulisa konkhe kubitwa kwe-model ngendlela yinye yangekhatsi ekhodini yakho, kuze kushintja umhlinzeki kube kushintja kwekuhlelwa kunekulungisa emafayela lamashumi.
  • Linganisa tindleko tangempela. Catsanisa kusebentisa i-model ku-hardware yakho ne-API lekhokhelwa ngekusetjentiswa emsebentini wakho wangempela, hhayi lowucabangako.

Sifundvo lesibalulekile

Bunikati encenyeni ye-AI buhamba ngekushesha, futsi butawuchubeka buhamba. Inkampani leyakha tinhlelo tayo ngemuva kwe-interface yangekhatsi lehlantekile, kudatha lehlelekile, ingashintja i-model noma umhlinzeki ngemalanga nje. Inkampani leyabophela umkhicito wayo kumhlinzeki munye kuze kuyewufika phansi itfola tindleko ngelilanga umhlinzeki ashintja ngalo, hhayi ngelilanga layisayina ngalo. Umehluko akukhona kukhetsa i-model lefanele; kwakha sistimu lengaphuli nangabe i-model ishintja.

Imitfombo: umbiko we-The Information wamhla ziyi-26 Ncwaba 2026, nekubikwa ngu-CNBC, Fortune, TechCrunch na-Bloomberg. Ngesikhatsi kushicilelwa, lesivumelwano besingakamenyetelwa ngalokusemtsetfweni yi-Nvidia noma yi-Hugging Face, futsi tonkhe tinombolo letilapha tivela kulemibiko.