NVIDIA is firing on all cylin짯ders, achie짯ving record reve짯nues in Gam짯ing, Data Cen짯ter and over짯all, said Jen짯sen Huang, foun짯der and CEO of NVIDIA. 쏷he new NVIDIA GeForce RTX GPU pro짯vi짯des our lar짯gest-ever gene짯ra짯tio짯nal leap and demand is over짯whel짯ming. NVIDIA RTX has made ray tra짯cing the new stan짯dard in gaming.

We are con짯ti짯nuing to rai짯se the bar with NVIDIA AI. Our A100 com짯pu짯te plat짯form is ram짯ping fast, with the top cloud com짯pa짯nies deploy짯ing it glo짯bal짯ly. We swept the indus짯try AI infe짯rence bench짯mark, and our cus짯to짯mers are moving some of the world셲 most popu짯lar AI ser짯vices into pro짯duc짯tion, powered by NVIDIA technology.

We announ짯ced the NVIDIA DPU pro짯gramma짯ble data cen짯ter pro짯ces짯sor, and the plan짯ned acqui짯si짯ti짯on of Arm, crea짯tor of the world셲 most popu짯lar CPU. We are posi짯tio짯ning NVIDIA for the age of AI, when com짯pu짯ting will extend from the cloud to tril짯li짯ons of devices.

NVIDIA paid $99 mil짯li짯on in quar짯ter짯ly cash divi짯dends in the third quar짯ter. It will pay its next quar짯ter짯ly cash divi짯dend of $0.16 per share on Decem짯ber 29, 2020, to all share짯hol짯ders of record on Decem짯ber 4, 2020.

Q3 Fis짯cal 2021 Sum짯ma짯ry

($ in mil짯li짯ons, except
ear짯nings per share)
Q3 FY21 Q2 FY21 Q3 FY20 Q/Q Y/Y
Reve짯nue $4,726   $3,866   $3,014   Up 22% Up 57%
Gross mar짯gin   62.6 %   58.8 %   63.6 % Up 380 bps Down 100 bps
Ope짯ra짯ting expenses $1,562   $1,624   $989   Down 4% Up 58%
Ope짯ra짯ting income $1,398   $651   $927   Up 115% Up 51%
Net inco짯me $1,336   $622   $899   Up 115% Up 49%
Diluted ear짯nings per share $2.12   $0.99   $1.45   Up 114% Up 46%


($ in mil짯li짯ons, except
ear짯nings per share)
Q3 FY21 Q2 FY21 Q3 FY20 Q/Q Y/Y
Reve짯nue $4,726   $3,866   $3,014   Up 22% Up 57%
Gross mar짯gin   65.5 %   66.0 %   64.1 % Down 50 bps Up 140 bps
Ope짯ra짯ting expenses $1,101   $1,035   $774   Up 6% Up 42%
Ope짯ra짯ting income $1,993   $1,516   $1,156   Up 31% Up 72%
Net inco짯me $1,834   $1,366   $1,103   Up 34% Up 66%
Diluted ear짯nings per share $2.91   $2.18   $1.78   Up 33% Up 63%

NVIDIA셲 out짯look for the fourth quar짯ter of fis짯cal 2021 is as follows:

  • Reve짯nue is expec짯ted to be $4.80 bil짯li짯on, plus or minus 2 percent.
  • GAAP and non-GAAP gross mar짯gins are expec짯ted to be 62.8 per짯cent and 65.5 per짯cent, respec짯tively, plus or minus 50 basis points.
  • GAAP and non-GAAP ope짯ra짯ting expen짯ses are expec짯ted to be appro짯xi짯m짯ate짯ly $1.64 bil짯li짯on and $1.18 bil짯li짯on, respectively.
  • GAAP and non-GAAP other inco짯me and expen짯se are both expec짯ted to be an expen짯se of appro짯xi짯m짯ate짯ly $55 million.
  • GAAP and non-GAAP tax rates are both expec짯ted to be 8 per짯cent, plus or minus 1 per짯cent, exclu짯ding any dis짯crete items. GAAP dis짯crete items include excess tax bene짯fits or defi짯ci짯en짯ci짯es rela짯ted to stock-based com짯pen짯sa짯ti짯on, which are expec짯ted to gene짯ra짯te varia짯bi짯li짯ty on a quar짯ter-by-quar짯ter basis.


During the third quar짯ter, NVIDIA announ짯ced a defi짯ni짯ti짯ve agree짯ment to acqui짯re Arm Limi짯t짯ed from Soft짯Bank Capi짯tal Limi짯t짯ed and SVF Hold짯co (UK) Limi짯t짯ed in a tran짯sac짯tion valued at $40 bil짯li짯on. The tran짯sac짯tion will com짯bi짯ne NVIDIA셲 lea짯ding AI com짯pu짯ting plat짯form with Arm셲 vast eco짯sys짯tem to crea짯te the pre짯mier com짯pu짯ting com짯pa짯ny for the age of AI. The tran짯sac짯tion which is expec짯ted to be imme짯dia짯te짯ly accre짯ti짯ve to NVIDIA셲 non-GAAP gross mar짯gin and non-GAAP ear짯nings per share is expec짯ted to clo짯se in the first quar짯ter of calen짯dar 2022. 

NVIDIA also announ짯ced plans to build a world-class AI lab in Cam짯bridge, Eng짯land  inclu짯ding a powerful AI super짯com짯pu짯ter based on NVIDIA and Arm tech짯no짯lo짯gy and pro짯vi짯de rese짯arch fel짯low짯ships and part짯ner짯ships with local insti짯tu짯ti짯ons and AI trai짯ning cour짯ses. Sepa짯ra짯te짯ly, it plans to build Cambridge1, the U.K.셲 most powerful AI super짯com짯pu짯ter, based on an NVIDIA DGX Super짯POD꽓 sys짯tem and desi짯gned for AI rese짯arch in health짯ca짯re and drug discovery.

NVIDIA also achie짯ved pro짯gress sin짯ce its pre짯vious ear짯nings announce짯ment in the짯se areas:

Data Cen짯ter

  • Third-quar짯ter reve짯nue was a record $1.90 bil짯li짯on, up 8 per짯cent from the pre짯vious quar짯ter and up 162 per짯cent from a year earlier.
  • Shared news that Amazon Web Ser짯vices and Ora짯cle Cloud Infra짯struc짯tu짯re announ짯ced gene짯ral avai짯la짯bi짯li짯ty of cloud com짯pu짯ting ins짯tances based on the NVIDIA A100GPU, fol짯lo짯wing Goog짯le Cloud Plat짯form and Micro짯soft Azure.
  • Announ짯ced the NVIDIA DGX Super짯POD Solu짯ti짯on for Enter짯pri짯se  the world셲 first turn짯key AI infra짯struc짯tu짯re which is expec짯ted to be instal짯led by yea짯rend in Korea, the U.K., India and Sweden.
  • Announ짯ced that five super짯com짯pu짯ters backed by EuroHPC inclu짯ding 쏬eo짯nar짯do, the world셲 fas짯test AI super짯com짯pu짯ter built by the Ita짯li짯an inter-uni짯ver짯si짯ty con짯sor짯ti짯um CINECA will use NVIDIA셲 data cen짯ter acce짯le짯ra짯tors or net짯wor짯king.
  • Intro짯du짯ced the NVIDIA BlueField2 DPU (data pro짯ces짯sing unit) sup짯port짯ed by NVIDIA DOCA꽓, a novel data-cen짯ter-infra짯struc짯tu짯re-on-a-chip archi짯tec짯tu짯re to bring breakth짯rough net짯wor짯king, sto짯rage and secu짯ri짯ty per짯for짯mance to every data center.
  • Announ짯ced a broad part짯ner짯ship with VMware to crea짯te an end-to-end enter짯pri짯se plat짯form for AI and a new archi짯tec짯tu짯re for data cen짯ter, cloud and edge using NVIDIA DPUs, bene짯fiting 300,000-plus VMware customers.
  • Unvei짯led NVIDIA Maxi짯ne꽓, an AI video-strea짯ming plat짯form that enhan짯ces strea짯ming qua짯li짯ty and offers such AI-powered fea짯tures as gaze cor짯rec짯tion, super-reso짯lu짯ti짯on, noi짯se can짯cel짯la짯ti짯on and face relighting.
  • Intro짯du짯ced the NVIDIA RTX A6000 and NVIDIA A40 GPUs, built on the NVIDIA Ampere archi짯tec짯tu짯re and fea짯turing new RT Cores, Ten짯sor Cores and CUDA짰 cores.
  • Exten짯ded its lead on MLPerf per짯for짯mance bench짯marks for infe짯rence, win짯ning every test across all six appli짯ca짯ti짯on are짯as for data cen짯ter and edge com짯pu짯ting systems.
  • Announ짯ced a part짯ner짯ship with GSK to inte짯gra짯te com짯pu짯ting plat짯forms for ima짯ging, geno짯mics and AI into the drug and vac짯ci짯ne dis짯co짯very process.
  • Intro짯du짯ced at SC20, three powerful advan짯ces in AI tech짯no짯lo짯gy: the NVIDIAA100 80GB GPU, powe짯ring the NVIDIA HGXAI super짯com짯pu짯ting plat짯form with twice the memo짯ry of its pre짯de짯ces짯sor; the NVIDIA DGX Sta짯ti짯on꽓 A100, the world셲 only petas짯ca짯le work짯group ser짯ver, for machi짯ne lear짯ning and data sci짯ence workloads; and the next gene짯ra짯ti짯on of NVIDIA Mel짯lan짯ox Infi짯ni짯Band, for the fas짯test net짯wor짯king performance.