Multiverse Computing and Qualcomm Collaborate to Bring Efficient AI Models to Data Centers
DONOSTIA-SAN SEBASTIÁN, Spai a d SAN DIEGO, Aug. 05, 2026 (GLOBE NEWSWIRE) — Multive se Computi g in the present day a ou ced a collabo atio with Qualcomm Tech ologies, I c. to b i g extremely efficie t AI fashions to information ce te s wo ldwide. Multive se Computi g’s AI fashions can be specialised fo Qualcomm D in the past fly™ AI200 a d AI250 accele ato s. The collabo atio focuses o Qualcomm Tech ologies’ AI accele atio ha dwa e a d Multive se Computi g’s mannequin optimizatio tech ology to e in a position information ce te ope ato s to u la ge-scale AI wo kloads with highe pe fo ma ce a d lowe powe co sumptio .
Fo information ce te ope ato s, the di ect be efit is mo e capability o the identical ha dwa e. By optimizi g AI fashions befo e they u o Qualcomm D in the past fly AI200 a d AI250 accele ato s, Multive se Computi g educes the compute a d memo y footp i t every mannequin equi es. That f ees up head oom o existi g deployme ts, letti g ope ato s se ve mo e i fe e ce equests, u mo e fashions co cu e tly, o scale thei AI se vices with out addi g ew accele ato s o expa di g information ce te i f ast uctu e.
The efficie cy gai s we e demo st ated reside at Cell Wo ld Co g ess i Ma ch 2026, whe e Multive se Computi g a d Qualcomm Tech ologies showcased a comp essed ope -sou ce la ge la guage mannequin u i g o a Qualcomm® Cloud AI100 Ult a accele ato . I a eal-time eme ge cy medical epo ti g use case, the comp essed mannequin demo delive ed as much as 93% faste espo se occasions a d as much as 44% highe th oughput tha the u comp essed base mannequin, whereas educi g memo y utilization by as much as 45% a d powe co sumptio by as much as 21%, with o loss i accu acy.
I a seco d demo st atio , a o -p emises Ret ieval-Augme ted Ge e atio (RAG) chatbot fo que yi g co fide tial fi a cial docume ts a as much as 35% faste a d delive ed as much as 54% highe th oughput, whereas cutti g memo y utilization by as much as 45% a d powe co sumptio by as much as 14%, agai with o loss i accu acy.
Applicatio s u i g o the ew Qualcomm D in the past fly AI200 a d AI250 accele ato s can be poised to delive eve bette be chma okay esults.
“Qualcomm Tech ologies has constructed its leade ship i segme ts reminiscent of cellular a d IoT, whe e maximizi g pe fo ma ce withi st ict powe a d efficie cy co st ai ts has at all times bee esse tial. Those self same p i ciples a e ow becomi g c itical i AI information ce te s. By combi i g Qualcomm Tech ologies’ extremely efficie t a d pe fo ma t AI accele ato s with Multive se Computi g’s mannequin comp essio a d optimizatio tech ology, we ca assist custome s obtain b eakth ough imp oveme ts i pe fo ma ce, price, a d e e gy efficie cy at scale.”
“As AI adoptio accele ates ac oss e te p ise a d cloud i f ast uctu e, custome s eed solutio s that delive excessive pe fo ma ce with g eate efficie cy. Qualcomm Tech ologies b i gs i mud y-leadi g AI accele atio , whereas Multive se Computi g wo ks intently with custome s, thei AI fashions, a d thei particular use instances to optimize eal-wo ld deployme ts. Togethe , we ca delive extremely efficie t AI solutio s tailo ed to the eeds of mode information ce te s.”
The collabo atio eflects a b oade shift i the information ce te i mud y towa d efficie cy-fi st AI i f ast uctu e, as ope ato s look to scale AI deployme ts with out a p opo tio al i c ease i ha dwa e, e e gy, a d price. By combi i g Qualcomm Tech ologies’ AI accele ato s with Multive se Computi g’s optimized AI fashions, the 2 compa ies p ovide information ce te ope ato s a d e te p ises with a p actical path to scalable AI.
Multive se Computi g is a leade i sove eig a d efficie t AI. The compa y develops quick, efficie t, a d extremely specialised AI fashions that e in a position o ga izatio s to deploy adva ced a tificial i tellige ce secu ely withi thei ow i f ast uctu e, e su i g full co t ol ove information, gove a ce, a d complia ce. Se vi g secto s whe e p ivacy, eliability, a d ope atio al efficie cy a e c itical, Multive se helps e te p ises u lock the worth of AI whereas mai tai i g sove eig ty ove thei most se sitive i fo matio . Headqua te ed i Do ostia-Sa Sebastiá , Spai , with workplaces i the U ited States, Ca ada, a d ac oss Eu ope, Multive se se ves mo e tha 100 world custome s, i cludi g Ibe d ola, Bosch, a d the Ba okay of Ca ada.
Qualcomm b a ded p oducts a e p oducts of Qualcomm Tech ologies, I c. a d/o its subsidia ies. Qualcomm a d Qualcomm D in the past fly a e t adema ks o egiste ed t adema ks of Qualcomm I co po ated.
