Intel Acquires Artificial Intelligence Chipmaker Habana Labs

Com­bi­na­ti­on Advan­ces Intel’s AI Stra­tegy, Streng­thens Port­fo­lio of AI Acce­le­ra­tors for the Data Center

SANTA CLARA, Calif.–(BUSINESS WIRE)–Intel Cor­po­ra­ti­on today announ­ced that it has acqui­red Haba­na Labs, an Isra­el-based deve­lo­per of pro­gramma­ble deep lear­ning acce­le­ra­tors for the data cen­ter for appro­xi­m­ate­ly $2 bil­li­on. The com­bi­na­ti­on streng­thens Intel’s arti­fi­ci­al intel­li­gence (AI) port­fo­lio and acce­le­ra­tes its efforts in the nas­cent, fast-gro­wing AI sili­con mar­ket, which Intel expects to be grea­ter than $25 bil­li­on by 20241.

This acqui­si­ti­on advan­ces our AI stra­tegy, which is to pro­vi­de cus­to­mers with solu­ti­ons to fit every per­for­mance need – from the intel­li­gent edge to the data cen­ter,” said Navin She­noy, exe­cu­ti­ve vice pre­si­dent and gene­ral mana­ger of the Data Plat­forms Group at Intel. “More spe­ci­fi­cal­ly, Haba­na tur­bo-char­ges our AI offe­rings for the data cen­ter with a high-per­for­mance trai­ning pro­ces­sor fami­ly and a stan­dards-based pro­gramming envi­ron­ment to address evol­ving AI workloads.”

Intel’s AI stra­tegy is groun­ded in the belief that harnes­sing the power of AI to impro­ve busi­ness out­co­mes requi­res a broad mix of tech­no­lo­gy – hard­ware and soft­ware – and full eco­sys­tem sup­port. Today, Intel AI solu­ti­ons are hel­ping cus­to­mers turn data into busi­ness value and dri­ving meaningful reve­nue for the com­pa­ny. In 2019, Intel expects to gene­ra­te over $3.5 bil­li­on in AI-dri­ven reve­nue, up more than 20 per­cent year-over-year. Tog­e­ther, Intel and Haba­na can acce­le­ra­te the deli­very of best-in-class AI pro­ducts for the data cen­ter, addres­sing cus­to­mers’ evol­ving needs.

She­noy con­tin­ued: “We know that cus­to­mers are loo­king for ease of pro­gramma­bi­li­ty with pur­po­se-built AI solu­ti­ons, as well as supe­ri­or, sca­lable per­for­mance on a wide varie­ty of workloads and neu­ral net­work topo­lo­gies. That’s why we’re thril­led to have an AI team of Habana’s cali­ber with a pro­ven track record of exe­cu­ti­on joi­ning Intel. Our com­bi­ned IP and exper­ti­se will deli­ver unmat­ched com­pu­ting per­for­mance and effi­ci­en­cy for AI workloads in the data center.”

Haba­na will remain an inde­pen­dent busi­ness unit and will con­ti­nue to be led by its cur­rent manage­ment team. Haba­na will report to Intel’s Data Plat­forms Group, home to Intel’s broad port­fo­lio of data cen­ter class AI tech­no­lo­gies. This com­bi­na­ti­on gives Haba­na access to Intel AI capa­bi­li­ties, inclu­ding signi­fi­cant resour­ces built over the last three years with deep exper­ti­se in AI soft­ware, algo­rith­ms and rese­arch that will help Haba­na sca­le and accelerate.

Haba­na chair­man Avigdor Wil­lenz has agreed to ser­ve as a seni­or advi­ser to the busi­ness unit as well as to Intel. Haba­na will con­ti­nue to be based in Isra­el whe­re Intel also has a signi­fi­cant pre­sence and long histo­ry of invest­ment. Pri­or to this tran­sac­tion, Intel Capi­tal was an inves­tor in Habana.

We have been for­t­u­na­te to get to know and col­la­bo­ra­te with Intel given its invest­ment in Haba­na, and we’re thril­led to be offi­ci­al­ly joi­ning the team,” said David Dahan, CEO of Haba­na. “Intel has crea­ted a world-class AI team and capa­bi­li­ty. We are exci­ted to part­ner with Intel to acce­le­ra­te and sca­le our busi­ness. Tog­e­ther, we will deli­ver our cus­to­mers more AI inno­va­ti­on, faster.”

Going for­ward, Intel plans to take full advan­ta­ge of its gro­wing port­fo­lio of AI tech­no­lo­gy and talent to deli­ver cus­to­mers unmat­ched com­pu­ting per­for­mance and effi­ci­en­cy for AI workloads.

Habana’s AI Trai­ning and Infe­rence Products

Habana’s Gau­di AI Trai­ning Pro­ces­sor is curr­ent­ly sam­pling with sel­ect hypers­ca­le cus­to­mers. Lar­ge-node trai­ning sys­tems based on Gau­di are expec­ted to deli­ver up to a 4x increase in through­put ver­sus sys­tems built with the equi­va­lent num­ber of GPUs. Gau­di is desi­gned for effi­ci­ent and fle­xi­ble sys­tem sca­le-up and scale-out.

Addi­tio­nal­ly, Habana’s Goya AI Infe­rence Pro­ces­sor, which is com­mer­ci­al­ly available, has demons­tra­ted excel­lent infe­rence per­for­mance inclu­ding through­put and real-time laten­cy in a high­ly com­pe­ti­ti­ve power enve­lo­pe. Gau­di for trai­ning and Goya for infe­rence offer a rich, easy-to-pro­gram deve­lo­p­ment envi­ron­ment to help cus­to­mers deploy and dif­fe­ren­tia­te their solu­ti­ons as AI workloads con­ti­nue to evol­ve with gro­wing demands on com­pu­te, memo­ry and connectivity.

About Intel

Intel (NASDAQ: INTC), a lea­der in the semi­con­duc­tor indus­try, is sha­ping the data-cen­tric future with com­pu­ting and com­mu­ni­ca­ti­ons tech­no­lo­gy that is the foun­da­ti­on of the world’s inno­va­tions. The company’s engi­nee­ring exper­ti­se is hel­ping address the world’s grea­test chal­lenges as well as hel­ping secu­re, power and con­nect bil­li­ons of devices and the infra­struc­tu­re of the smart, con­nec­ted world – from the cloud to the net­work to the edge and ever­y­thing in bet­ween. Find more infor­ma­ti­on about Intel at newsroom.intel.com and intel.com.

© Intel Cor­po­ra­ti­on. Intel, the Intel logo, and other Intel marks are trade­marks of Intel Cor­po­ra­ti­on or its sub­si­dia­ries. Other names and brands may be clai­med as the pro­per­ty of others.

1Intel esti­ma­tes the total addressa­ble mar­ket (TAM) for AI sili­con by 2024 will be grea­ter than $25 bil­li­on, and within that, AI sili­con in the data cen­ter is expec­ted to be grea­ter than $10 bil­li­on in the same timeframe.