Nvidia CEO Jensen Huang: AI is going to fundamentally change how we compute everything (YouTube Video Transcript)

(00:00:00) Your YouTube transcript will appear here (00:00:01) THE ROAD. (00:00:02) >> YEAH, I WAS GOING TO SAY YOU (00:00:03) COULDN'T ACTUALLY WALK OVER (00:00:05) HERE, BUT YOU'RE HERE. AND YOU (00:00:06) SAID YOU'VE NEVER BEEN HERE (00:00:07) BEFORE. (00:00:08) >> WELL, I'VE BEEN NOT LIKE (00:00:09) THIS. NOT LIKE THIS. LOOK AT (00:00:10) ALL THESE TENTS. IT'S (00:00:12) INCREDIBLE. IT'S GOING TO BE A (00:00:13) FESTIVAL. (00:00:14) >> SO IT'S AN INTERESTING TIME (00:00:15) TO HAVE YOU, OBVIOUSLY, WHICH (00:00:16) ANY TIME IS, BUT ESPECIALLY (00:00:17) GIVEN WHAT'S BEEN HAPPENING (00:00:18) WITHIN THE MARKET TODAY, LOOKS (00:00:20) A LOT DIFFERENT THAN THE LAST (00:00:21) COUPLE OF DAYS. HAVE THE MARKET (00:00:22) OBVIOUSLY CONCERNED WITH WHAT'S (00:00:24) HAPPENING ON THE SPENDING SIDE (00:00:25) OF THE THE HYPERSCALERS AND (00:00:26) THIS BIG BUILD OUT IN AI? (00:00:27) WHAT'S YOUR TAKE FROM THAT? THE (00:00:29) WAY THE MARKET CONTINUES TO (00:00:30) REACT TO THIS. (00:00:31) >> DEMAND IS SKY HIGH. AND (00:00:32) THERE'S A FUNDAMENTAL REASON (00:00:34) FOR THAT. WE'RE IN THE ONCE IN (00:00:35) A GENERATION INFRASTRUCTURE (00:00:37) BUILD OUT. THIS IS THE LARGEST (00:00:38) INFRASTRUCTURE BUILD OUT IN (00:00:38) HUMAN HISTORY. AND THERE'S A (00:00:40) FUNDAMENTAL REASON FOR THAT. (00:00:41) ARTIFICIAL INTELLIGENCE IS (00:00:42) GOING TO FUNDAMENTALLY CHANGE (00:00:44) HOW WE COMPUTE EVERYTHING, (00:00:46) EVERYTHING FROM DATABASE (00:00:47) PROCESSING, THE WAY WE DO (00:00:48) SEARCH, THE WAY WE DO (00:00:49) RECOMMENDER SYSTEMS, YOU SHOP, (00:00:51) THE WAY YOU WATCH MOVIES, AND (00:00:52) OF COURSE, THESE NEW SYSTEMS (00:00:54) THAT ARE BEING THAT ARE BEING (00:00:56) DEVELOPED AND EVOLVED. LAST (00:00:57) YEAR, THIS LAST YEAR, WE SAW AN (00:01:00) INFLECTION POINT IN AI. AI (00:01:02) BECAME SUPER USEFUL, NO LONGER (00:01:05) HALLUCINATING. IT'S GENERATING (00:01:07) INFORMED CONTENT. IT'S (00:01:09) REASONING, IT'S THINKING, IT'S (00:01:10) DOING RESEARCH. IT'S ABLE TO (00:01:12) USE TOOLS. ALL OF A SUDDEN, AI (00:01:14) OVER THE LAST COUPLE OF YEARS (00:01:15) WENT FROM BEING CURIOUS TO (00:01:18) SUPER USEFUL. THE INFLECTION (00:01:19) POINT ALSO CAME WITH IT (00:01:21) PROFITABLE TOKENS. ANTHROPIC IS (00:01:23) MAKING GREAT MONEY. OPENAI IS (00:01:25) MAKING GREAT MONEY. IF THEY (00:01:27) COULD HAVE TWICE AS MUCH (00:01:28) COMPUTE, THE REVENUES WOULD GO (00:01:30) UP FOUR TIMES AS MUCH. I MEAN, (00:01:31) LITERALLY THESE GUYS ARE SO (00:01:34) DEMAND, SO COMPUTE CONSTRAINED (00:01:36) AND THE DEMAND IS SO INCREDIBLY (00:01:38) GREAT. THE NUMBER OF ENTERPRISE (00:01:40) USERS, THE NUMBER OF CONSUMER (00:01:42) USERS, THE NUMBER OF STARTUPS (00:01:43) THAT HAVE BEEN BUILT ON TOP OF (00:01:45) THESE COMPANIES. IT'S JUST (00:01:46) GOING THROUGH THE ROOF. (00:01:47) >> SO WHEN YOU SEE NUMBERS THAT (00:01:48) ARE FRANKLY REMARKABLE, THAT (00:01:51) $660 BILLION THIS YEAR WILL BE (00:01:53) SPENT BY THE HYPERSCALERS AS WE (00:01:55) CONTINUE TO LEARN LIKE WE DID (00:01:56) FROM AMAZON LAST NIGHT, YOU SIT (00:01:59) BACK AND THINK TO YOURSELF, (00:02:01) THAT'S COMPLETELY JUSTIFIED (00:02:02) BASED ON WHAT YOU SEE. (00:02:03) >> IT IS APPROPRIATE AND (00:02:06) SUSTAINABLE. AND THE REASON FOR (00:02:07) THAT IS BECAUSE ALL OF THESE (00:02:09) COMPANIES CASH FLOWS ARE GOING (00:02:10) TO START RISING. YOU KNOW, (00:02:12) PEOPLE ARE COMPARING IT TO CASH (00:02:13) FLOWS. ONE OF THOSE NUMBERS ARE (00:02:15) WRONG. IT'S JUST THE CASH FLOW (00:02:16) IS WRONG. WE ARE ADDRESSING THE (00:02:19) LARGEST SOFTWARE OPPORTUNITY IN (00:02:21) HISTORY FOR THE VERY FIRST TIME. (00:02:24) SOFTWARE IS NOT JUST A TOOL. A (00:02:25) TOOL IS LIKE EXCEL. NOW (00:02:28) SOFTWARE USES TOOLS. SO THESE (00:02:30) AI'S USE EXCEL. AND SO I THINK (00:02:33) THE THE THE OPPORTUNITY TO TO (00:02:35) FOR THIS NEW NEW ERA OF (00:02:36) SOFTWARE IS INCREDIBLE. AND (00:02:37) WE'RE SEEING IT ALREADY MOVING (00:02:41) THE EARNINGS OF META. NOBODY (00:02:43) USES AI BETTER THAN META. AND (00:02:45) SO IF YOU LOOK AT THE WAY THAT (00:02:47) THEY'RE USING AI, AI WENT FROM (00:02:48) A, A CLASSICAL RECOMMENDER (00:02:50) SYSTEM RUNNING ON CPUS TO NOW A (00:02:52) GENERATIVE AI AGENTIC SYSTEM (00:02:55) THAT IS MAKING RECOMMENDATIONS, (00:02:58) EVERYTHING FROM THE WAY SOCIAL (00:02:59) MEDIA WORKS AND THE WAY THEY (00:03:00) THEY RECOMMEND ADS AND HELP (00:03:03) ADVERTISERS CREATE CONTENT HAS (00:03:05) FUNDAMENTALLY BEEN CHANGED AND (00:03:06) THEIR EARNINGS SHOW IT. AND (00:03:08) THAT'S THE REASON WHY THEY'RE (00:03:09) INVESTING SO HARD. THEY SEE (00:03:10) JUST A MUCH LARGER FUTURE (00:03:11) POTENTIAL FOR IT AND NOT JUST (00:03:13) ONE COMPANY. THIS IS GOING TO (00:03:14) AFFECT AWS SHOPPING AND THE WAY (00:03:16) THEY RECOMMEND GOODS. THIS IS (00:03:18) GOING TO AFFECT HOW MICROSOFT'S (00:03:19) ENTERPRISE SOFTWARE WORKS. (00:03:20) EVERY SINGLE COMPANY IS SEES (00:03:23) THE SAME INFLECTION POINT. AND (00:03:24) THAT'S WHY EVERYBODY LEANING IN (00:03:25) SO HARD. (00:03:26) >> I FEEL LIKE THAT'S THE POINT (00:03:28) THAT YOU WERE. IT'S AN (00:03:29) INTERESTING POINT. IT SORT OF (00:03:30) TURNS THE WAY YOU NEED TO THINK (00:03:32) ABOUT THESE BUSINESSES ON ITS (00:03:33) HEAD, THAT IF YOU'RE JUST (00:03:35) FIXATED ON THE SPENDING SIDE, (00:03:37) AS YOU WERE TALKING EARLIER, (00:03:38) ABOUT, OH, THEY'RE SPENDING (00:03:39) SUCH A HUGE PORTION OF THEIR (00:03:40) FREE CASH FLOW, YOU'RE KIND OF (00:03:42) MISSING THE STORY. THE MARKET (00:03:43) NEEDS TO GET ITS ARMS AROUND (00:03:45) THAT. YOU AS WELL AS AN (00:03:46) INVESTOR. (00:03:47) >> YES. IF YOU TURN THE CLOCK (00:03:49) BACK SCOTT, TO 2008 2009. (00:03:51) >> RIGHT. AMAZON COULD HAVE (00:03:52) TAKEN THEIR PROFITS AND SENT (00:03:54) THEM BACK TO US BY WAY OF (00:03:55) DIVIDENDS. INSTEAD, JEFF BEZOS (00:03:56) SAID, I'M GOING TO INVEST IN (00:03:58) THIS THING CALLED AWS. THINK OF (00:04:00) IT AS DIGGING A GOLD MINE. YOU (00:04:01) HAVE TO SPEND A LOT OF MONEY TO (00:04:03) DIG THE GOLD MINE BEFORE YOU (00:04:05) GET THE MONEY, BEFORE YOU GET (00:04:06) THE GOLD OUT, THESE GUYS ARE (00:04:08) DIGGING THE BIGGEST GOLD MINE (00:04:09) IN THE HISTORY OF SOFTWARE, (00:04:10) RIGHT? BUT IT COSTS SOMETHING (00:04:12) UP FRONT. NOW, THE QUESTION IS, (00:04:14) DO YOU BELIEVE ANDY JASSY AND (00:04:15) MARK, YOU KNOW, MARK ZUCKERBERG (00:04:18) AND SUNDAR AND JENSEN, THIS (00:04:20) OPPORTUNITY IS THAT BIG OR DO (00:04:22) YOU HAVE A SPECIAL INSIGHT THAT (00:04:23) CAUSES YOU TO UNDERSTAND (00:04:25) THERE'S NO GOLD IN THE BOTTOM (00:04:26) OF THE GOLD MINE. AS AN (00:04:27) INVESTOR, I'LL JUST TELL YOU (00:04:28) THIS. I WANT MY PERSONAL NET (00:04:30) WORTH. I WANT MY FUNDS NET (00:04:31) WORTH LEVERED AGAINST AI (00:04:34) BECAUSE ALL HUMAN PROGRESS IS (00:04:36) GOING TO BE DERIVED FROM (00:04:37) MACHINES HELPING HUMANS THINK (00:04:39) AND AUGMENT HUMAN THINKING. AND (00:04:41) SO WHETHER IT'S OWNING NVIDIA, (00:04:42) WHICH IS OUR LARGEST PUBLIC (00:04:43) POSITION OR WHETHER IT'S (00:04:45) ANTHROPIC OR OPENAI, ALL OF (00:04:46) THESE COMPANIES ARE GOING TO BE (00:04:48) TREMENDOUS BENEFICIARIES AND (00:04:49) THEY'RE DOING IT AT SCALE. (00:04:51) THERE ARE INCREASING ADVANTAGES (00:04:53) TO SCALE. RIGHT. AND SO I THINK (00:04:55) IT'S HARD FOR PEOPLE TO GET (00:04:57) THEIR ARMS AROUND NOW, JUST (00:04:58) LIKE IT WAS IN 2008 AND 2009. (00:05:00) BUT THANK GOD JEFF BEZOS DID (00:05:01) THAT. NOW IT'S $140 BILLION (00:05:03) BUSINESS, GENERATING 30 BILLION (00:05:05) A YEAR IN ANNUAL PROFITS THAT (00:05:07) PEOPLE CRITICIZED HIM FOR (00:05:08) INVESTING IN BACK THEN. (00:05:09) >> HOW CAN JUST. (00:05:11) >> IT'S INCREDIBLE THAT OPENAI (00:05:13) AND ANTHROPIC ARE $20 BILLION (00:05:15) RUN RATE COMPANIES PROFIT (00:05:18) GENERATING PROFITABLE REVENUES (00:05:20) AND ACCELERATING GROWTH ALL AT (00:05:22) THE SAME TIME? THESE ARE (00:05:24) EXTRAORDINARY THINGS AND IT (00:05:25) MAKES PERFECT SENSE FROM WHERE (00:05:26) WE SIT. (00:05:27) >> HOW CAN WE BE SO CERTAIN (00:05:29) THAT THE AMOUNT OF COMPUTE WE (00:05:31) THINK WE'LL NEED, AND THAT WE (00:05:33) CONTINUE TO HEAR FROM CEOS LIKE (00:05:36) YOU THAT WE WILL NEED, IS (00:05:38) ACTUALLY GOING TO COME TO (00:05:39) FRUITION? ISN'T IT INEVITABLE (00:05:40) THAT WE'RE GOING TO, AT SOME (00:05:42) POINT OVERDO IT? BUT WE'RE NOT (00:05:44) GOING TO KNOW THAT UNTIL IT'S (00:05:46) TOO LATE? (00:05:46) >> IT'S NOT LIKE ROADS, ROADS. (00:05:48) ONCE YOU LAY IT DOWN, IT STAYS (00:05:50) FAIRLY USEFUL FOR A LONG TIME. (00:05:52) THE WAY COMPUTERS WORK IS YOU (00:05:53) BUILD UP THE INFRASTRUCTURE, (00:05:56) BUT THEN YOU START REPLACING IT (00:05:57) FIVE, SIX, SEVEN YEARS LATER. (00:05:59) AND SO I THINK IT'S GOING TO (00:06:01) TAKE US SOME SEVEN YEARS, EIGHT (00:06:02) YEARS TO BUILD UP TO THE LEVEL (00:06:05) THAT WE HAVE TO SUSTAIN. AND (00:06:07) AFTER THAT WE'RE GOING TO BE (00:06:08) REFRESHING AND SLIGHTLY GROWING. (00:06:09) AND SO WE'VE GOT SEVERAL YEARS (00:06:11) OF BUILD OUT AHEAD OF US. THE (00:06:12) IMPORTANT THING IS THIS YOU (00:06:14) HAVE TO GO BACK TO FIRST (00:06:15) PRINCIPLES. COMPUTING HAS (00:06:16) FUNDAMENTALLY CHANGED. IT USED (00:06:18) TO BE PRERECORDED EXCEL AND (00:06:21) POWERPOINT AND ALL THESE TOOLS (00:06:22) AND ALL THIS SOFTWARE IS (00:06:24) PRE-COMPILED. THEY COMPILED IT (00:06:25) AND THEY SEND IT TO US. THE WAY (00:06:26) SOFTWARE WORKS NOW, BECAUSE (00:06:28) IT'S CONTEXTUALLY AWARE EVERY (00:06:30) SINGLE TIME YOU USE IT, IT (00:06:32) TAKES INTO CONSIDERATION THE (00:06:33) CONTEXT, WHO YOU ARE, WHAT (00:06:34) YOU'RE ASKING ABOUT, WHAT'S (00:06:36) HAPPENING TO THE WORLD, WHAT (00:06:37) OTHER INFORMATION YOU GIVE IT. (00:06:39) THE CONTEXT AND EVERY SINGLE (00:06:41) CONTEXT IS DIFFERENT AND EVERY (00:06:42) SINGLE RESPONSE IS DIFFERENT. (00:06:44) SO FROM NOW ON, EVERY SINGLE (00:06:47) PIXEL, EVERY SINGLE BEEP OF A (00:06:48) SOUND, EVERY VIDEO THAT COMES (00:06:50) OUT OF THE COMPUTER IS (00:06:51) GENERATED IN REAL TIME. AND (00:06:54) THAT'S THE REASON WHY WE NEED (00:06:56) COMPUTERS TO OPERATE SOFTWARE (00:06:57) AT SUCH A LARGE SCALE GOING (00:06:59) FORWARD. AND ALL THESE TOKENS (00:07:00) ARE WHAT WE CALL INTELLIGENCE. (00:07:02) AND FOR THE FIRST TIME WE HAVE, (00:07:04) WE'RE GENERATING SOMETHING OF (00:07:06) TREMENDOUS VALUE INTELLIGENCE (00:07:09) AS NUMBERS. AND THESE THINGS, (00:07:10) WE AS I MENTIONED EARLIER THIS (00:07:11) LAST YEAR, WE WENT THROUGH AN (00:07:13) INFLECTION POINT BECAUSE THESE (00:07:14) TOKENS BECAME PROFITABLE. AND (00:07:16) SO WE JUST HAVE TO MAKE MORE OF (00:07:17) IT. (00:07:18) >> CAN YOU BLAME IN SOME WAYS, (00:07:20) INVESTORS WHO HAVE PTSD FROM (00:07:23) THE LAST TIME TECH WENT THROUGH (00:07:25) THIS REVOLUTION AND EVOLUTION (00:07:28) AT THE SAME TIME, FOR SEEING (00:07:30) CERTAIN THINGS THAT REMIND THEM (00:07:32) OF THEN AND WORRY THAT THEY'LL (00:07:34) HAPPEN AGAIN. (00:07:35) >> IT'S GOOD TO ALWAYS REFLECT (00:07:36) ON HISTORY. HISTORY INFORMS US, (00:07:39) BUT HISTORY DOESN'T REPEAT. AND (00:07:41) SO YOU HAVE TO TAKE INTO (00:07:42) CONSIDERATION WHAT'S ACTUALLY (00:07:43) HAPPENING. ALWAYS GO BACK TO (00:07:44) FIRST PRINCIPLES AND THINK (00:07:45) ABOUT WHAT IS ACTUALLY (00:07:46) HAPPENING RIGHT NOW. THE (00:07:48) FUNDAMENTAL DIFFERENCE, A HUGE (00:07:49) DIFFERENCE BETWEEN NOW AND SOME (00:07:52) THINGS THAT HAPPENED IN THE (00:07:53) PAST WITH THE INTERNET. THERE (00:07:55) WAS A TON OF DARK FIBER. THERE (00:07:58) IS NO THERE ARE NO DARK GPUS. (00:08:00) 100% OF THE GPUS ARE RENTED. IN (00:08:03) FACT, GPUS THAT WE SOLD SIX (00:08:05) YEARS AGO, THE PRICES ARE GOING (00:08:07) UP. IT'S NOT LIKE IT'S ANTIQUE. (00:08:09) I MEAN, IT'S INCREDIBLE. IT'S (00:08:10) LIKE A FINE WINE. THE DEMAND IS (00:08:12) SO HIGH THAT GPUS I SOLD SIX (00:08:14) YEARS AGO ARE GOING UP IN PRICE. (00:08:16) AND SO I THINK THE DEMAND IS (00:08:17) JUST INCREDIBLY HIGH FOR THE (00:08:19) REASONS THAT I JUST SAID ONE. (00:08:21) WE WENT THROUGH AN INFLECTION (00:08:23) POINT. AI HAS BECOME USEFUL AND (00:08:25) VERY, VERY CAPABLE TO THE (00:08:28) ADOPTION OF IT IS INCREDIBLY (00:08:29) HIGH. AND THREE, BECAUSE THESE (00:08:31) AIS ARE THINKING THE AMOUNT OF (00:08:33) COMPUTATION IT NEEDS IS GREAT. (00:08:35) AN