โ†

๐Ÿ“„ Earnings Call Transcript ๋ฒˆ์—ญ ๊ฒฐ๊ณผ

๐Ÿ“Š Presentation

Original Translation
Call Start: 12:45 January 1, 0000 1:26 PM ET

NVIDIA Corporation (NASDAQ:NVDA)
JPMorgan Healthcare Conference
January 13, 2025, 12:45 PM ET

Company Participants

Kimberly Powell - VP of Healthcare
Toy Jensen - CEO Avatar

Conference Call Participants

Harlan Sur - JPMorgan

Harlan Sur

All right. Good morning, and welcome to JPMorgan's 43rd Annual Healthcare Conference here in San Francisco. My name is Harlan Sur. I'm the Semiconductor Analyst for the firm. And for the sixth time in seven years, we have the team from NVIDIA presenting.
์ „ํ™” ์‹œ์ž‘: 12:45 January 1, 0000 ์˜คํ›„ 1:26 ET

์—”๋น„๋””์•„ ์ฝ”ํผ๋ ˆ์ด์…˜ (NASDAQ:NVDA)
JP๋ชจ๊ฑด ํ—ฌ์Šค์ผ€์–ด ์ปจํผ๋Ÿฐ์Šค
2025๋…„ 1์›” 13์ผ, ์˜คํ›„ 12:45 ET

ํšŒ์‚ฌ ์ฐธ์„์ž

ํ‚ด๋ฒŒ๋ฆฌ ํŒŒ์›” - ํ—ฌ์Šค์ผ€์–ด ๋ถ€์‚ฌ์žฅ
ํ† ์ด ์  ์Šจ - CEO ์•„๋ฐ”ํƒ€

์ปจํผ๋Ÿฐ์Šค ์ฝœ ์ฐธ์„์ž

ํ• ๋ž€ ์„œ - JP๋ชจ๊ฑด

ํ• ๋ž€ ์„œ

์ข‹์€ ์•„์นจ์ž…๋‹ˆ๋‹ค. ์ƒŒํ”„๋ž€์‹œ์Šค์ฝ”์—์„œ ์—ด๋ฆฌ๋Š” JP๋ชจ๊ฑด ์ œ43ํšŒ ์—ฐ๋ก€ ํ—ฌ์Šค์ผ€์–ด ์ปจํผ๋Ÿฐ์Šค์— ์˜ค์‹  ๊ฒƒ์„ ํ™˜์˜ํ•ฉ๋‹ˆ๋‹ค. ์ €๋Š” ํ• ๋ž€ ์„œ์ด๋ฉฐ, ๋‹น์‚ฌ์˜ ๋ฐ˜๋„์ฒด ์• ๋„๋ฆฌ์ŠคํŠธ์ž…๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  7๋…„ ์ค‘ 6๋ฒˆ์งธ๋กœ ์—”๋น„๋””์•„ ํŒ€์ด ๋ฐœํ‘œ๋ฅผ ํ•˜๊ฒŒ ๋˜์—ˆ์Šต๋‹ˆ๋‹ค.
NVIDIA, as most of you know, leader in accelerated compute and AI semiconductors, software and systems, powering some of the world's most powerful AI compute clusters, high-performance computing systems, and driving compute innovation for cloud hyperscalers as well as large vertical markets like healthcare and life sciences. Here with us today from NVIDIA is Kimberly Powell, Vice President of Healthcare at NVIDIA. She's responsible for the company's worldwide healthcare business, including hardware and software platforms for accelerated computing, AI, visualization that power the ecosystems of imaging, genomics, life sciences, drug discovery, and healthcare analytics.๋Œ€๋ถ€๋ถ„ ์•„์‹œ๋‹ค์‹œํ”ผ NVIDIA๋Š” ๊ฐ€์† ์ปดํ“จํŒ… ๋ฐ AI ๋ฐ˜๋„์ฒด, ์†Œํ”„ํŠธ์›จ์–ด, ์‹œ์Šคํ…œ ๋ถ„์•ผ์˜ ์„ ๋„๊ธฐ์—…์œผ๋กœ, ์„ธ๊ณ„์—์„œ ๊ฐ€์žฅ ๊ฐ•๋ ฅํ•œ AI ์ปดํ“จํŒ… ํด๋Ÿฌ์Šคํ„ฐ์™€ ๊ณ ์„ฑ๋Šฅ ์ปดํ“จํŒ… ์‹œ์Šคํ…œ์„ ๊ตฌ๋™ํ•˜๋ฉฐ, ํด๋ผ์šฐ๋“œ ํ•˜์ดํผ์Šค์ผ€์ผ๋Ÿฌ๋Š” ๋ฌผ๋ก  ํ—ฌ์Šค์ผ€์–ด ๋ฐ ์ƒ๋ช…๊ณผํ•™๊ณผ ๊ฐ™์€ ๋Œ€๊ทœ๋ชจ ์ˆ˜์ง ์‹œ์žฅ์„ ์œ„ํ•œ ์ปดํ“จํŒ… ํ˜์‹ ์„ ์ฃผ๋„ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์˜ค๋Š˜ NVIDIA์—์„œ ์ฐธ์„ํ•ด ์ฃผ์‹  ๋ถ„์€ ํ‚ด๋ฒŒ๋ฆฌ ํŒŒ์›”(Kimberly Powell) NVIDIA ํ—ฌ์Šค์ผ€์–ด ๋ถ€์‚ฌ์žฅ์ž…๋‹ˆ๋‹ค. ๊ทธ๋…€๋Š” NVIDIA์˜ ์ „ ์„ธ๊ณ„ ํ—ฌ์Šค์ผ€์–ด ์‚ฌ์—…์„ ๋‹ด๋‹นํ•˜๊ณ  ์žˆ์œผ๋ฉฐ, ์—ฌ๊ธฐ์—๋Š” ์ด๋ฏธ์ง•, ์œ ์ „์ฒดํ•™, ์ƒ๋ช…๊ณผํ•™, ์‹ ์•ฝ ๊ฐœ๋ฐœ, ํ—ฌ์Šค์ผ€์–ด ๋ถ„์„ ์ƒํƒœ๊ณ„๋ฅผ ๊ตฌ๋™ํ•˜๋Š” ๊ฐ€์† ์ปดํ“จํŒ…, AI, ์‹œ๊ฐํ™”๋ฅผ ์œ„ํ•œ ํ•˜๋“œ์›จ์–ด ๋ฐ ์†Œํ”„ํŠธ์›จ์–ด ํ”Œ๋žซํผ์ด ํฌํ•จ๋ฉ๋‹ˆ๋‹ค.
So Kimberly, thank you for joining us today, and let me turn it over to you. Kimberly Powell

Thank you so much, Harlan. Thank you. Wow, packed house. Amazing. Welcome to JPMorgan 2025. Here we are. Appreciate so much the invitation to be here. This is a wonderful honor not only to talk about what the future of healthcare is going to look like, but also celebrate the amazing achievements that we've been able to accomplish with it, an amazing ecosystem of partners. So if I could just take a minute and give you a look at our safe harbor. Please take a minute. You are very familiar with this.
ํ‚ด๋ฒŒ๋ฆฌ, ์˜ค๋Š˜ ํ•จ๊ป˜ํ•ด ์ฃผ์…”์„œ ๊ฐ์‚ฌํ•ฉ๋‹ˆ๋‹ค. ์ด์ œ ๋ง์”€์„ ๋ถ€ํƒ๋“œ๋ฆฌ๊ฒ ์Šต๋‹ˆ๋‹ค.

ํ‚ด๋ฒŒ๋ฆฌ ํŒŒ์›”

์ •๋ง ๊ฐ์‚ฌํ•ฉ๋‹ˆ๋‹ค, ํ• ๋ž€. ๊ฐ์‚ฌํ•ฉ๋‹ˆ๋‹ค. ์™€, ์ •๋ง ๋งŽ์€ ๋ถ„๋“ค์ด ์˜ค์…จ๋„ค์š”. ๋†€๋ž์Šต๋‹ˆ๋‹ค. JP๋ชจ๊ฑด 2025์— ์˜ค์‹  ๊ฒƒ์„ ํ™˜์˜ํ•ฉ๋‹ˆ๋‹ค. ์—ฌ๊ธฐ ์™”์Šต๋‹ˆ๋‹ค. ์ด ์ž๋ฆฌ์— ์ดˆ๋Œ€ํ•ด ์ฃผ์…”์„œ ์ •๋ง ๊ฐ์‚ฌํ•ฉ๋‹ˆ๋‹ค. ํ—ฌ์Šค์ผ€์–ด์˜ ๋ฏธ๋ž˜๊ฐ€ ์–ด๋–ค ๋ชจ์Šต์ผ์ง€์— ๋Œ€ํ•ด ์ด์•ผ๊ธฐํ•  ๋ฟ๋งŒ ์•„๋‹ˆ๋ผ, ๋†€๋ผ์šด ํŒŒํŠธ๋„ˆ ์ƒํƒœ๊ณ„์™€ ํ•จ๊ป˜ ์ด๋ฃฐ ์ˆ˜ ์žˆ์—ˆ๋˜ ๋†€๋ผ์šด ์„ฑ๊ณผ๋“ค์„ ์ถ•ํ•˜ํ•  ์ˆ˜ ์žˆ๋Š” ๊ฒƒ์€ ์ •๋ง ํฐ ์˜๊ด‘์ž…๋‹ˆ๋‹ค. ์ž ์‹œ ์‹œ๊ฐ„์„ ๋‚ด์–ด ์•ˆ์ „ ์กฐํ•ญ(safe harbor)์„ ๋ณด์—ฌ๋“œ๋ฆฌ๊ฒ ์Šต๋‹ˆ๋‹ค. ์ž ์‹œ ์‹œ๊ฐ„์„ ๋‚ด์–ด ํ™•์ธํ•ด ์ฃผ์„ธ์š”. ์—ฌ๋Ÿฌ๋ถ„๊ป˜์„œ๋Š” ์ด๋ฏธ ์ž˜ ์•„์‹œ๋Š” ๋‚ด์šฉ์ž…๋‹ˆ๋‹ค.
The presentation contains forward-looking statements, and investors are advised to read our reports filed with the SEC for information related to risks and uncertainties. Okay. Now we can get into the good stuff. This clicker is not working, which would be wonderful if it did. Perhaps you could give it a go while I continue. I would like to introduce you to this amazing computing platform shift. Give it one second and we are -- here we go. This AI revolution that we are living in is a function of two simultaneous computing platform shifts, artificial intelligence and accelerating computing, and one cannot happen without the other.์ด ๋ฐœํ‘œ์—๋Š” ๋ฏธ๋ž˜ ์ „๋ง ์ง„์ˆ ์ด ํฌํ•จ๋˜์–ด ์žˆ์œผ๋ฉฐ, ํˆฌ์ž์ž๋“ค๊ป˜์„œ๋Š” ์œ„ํ—˜ ์š”์†Œ์™€ ๋ถˆํ™•์‹ค์„ฑ์— ๊ด€ํ•œ ์ •๋ณด๋ฅผ ์œ„ํ•ด SEC์— ์ œ์ถœ๋œ ๋‹น์‚ฌ์˜ ๋ณด๊ณ ์„œ๋ฅผ ์ฝ์–ด๋ณด์‹œ๊ธฐ ๋ฐ”๋ž๋‹ˆ๋‹ค. ์ข‹์Šต๋‹ˆ๋‹ค. ์ด์ œ ๋ณธ๊ฒฉ์ ์ธ ๋‚ด์šฉ์œผ๋กœ ๋“ค์–ด๊ฐ€๊ฒ ์Šต๋‹ˆ๋‹ค. ์ด ํด๋ฆฌ์ปค๊ฐ€ ์ž‘๋™ํ•˜์ง€ ์•Š๋„ค์š”. ์ž‘๋™ํ–ˆ๋‹ค๋ฉด ์ข‹์•˜์„ ํ…๋ฐ์š”. ์ œ๊ฐ€ ๊ณ„์† ์ง„ํ–‰ํ•˜๋Š” ๋™์•ˆ ํ•œ๋ฒˆ ์‹œ๋„ํ•ด ๋ณด์‹œ๊ฒ ์Šต๋‹ˆ๊นŒ? ์ด ๋†€๋ผ์šด ์ปดํ“จํŒ… ํ”Œ๋žซํผ ์ „ํ™˜์— ๋Œ€ํ•ด ์†Œ๊ฐœํ•ด๋“œ๋ฆฌ๊ณ  ์‹ถ์Šต๋‹ˆ๋‹ค. ์ž ์‹œ๋งŒ์š”, ์šฐ๋ฆฌ๋Š” -- ๋์Šต๋‹ˆ๋‹ค. ์šฐ๋ฆฌ๊ฐ€ ์‚ด๊ณ  ์žˆ๋Š” ์ด AI ํ˜๋ช…์€ ๋‘ ๊ฐ€์ง€ ๋™์‹œ์ ์ธ ์ปดํ“จํŒ… ํ”Œ๋žซํผ ์ „ํ™˜, ์ฆ‰ ์ธ๊ณต์ง€๋Šฅ๊ณผ ๊ฐ€์† ์ปดํ“จํŒ…์˜ ๊ฒฐ๊ณผ์ด๋ฉฐ, ํ•˜๋‚˜๋Š” ๋‹ค๋ฅธ ํ•˜๋‚˜ ์—†์ด๋Š” ์ผ์–ด๋‚  ์ˆ˜ ์—†์Šต๋‹ˆ๋‹ค.
It's a really important concept to understand that the last many decades of computing was software being written by humans in what humans can understand, which is a series of instructions that would execute on CPUs. Today, the modern operating system is large language models, and what happens is you introduce the computer with data. These large language models learn, they do machine learning and then they can assemble software and out of the computer are tokens. These tokens essentially represent digital intelligence. This digital intelligence is the entire industry's future products and services.์ง€๋‚œ ์ˆ˜์‹ญ ๋…„๊ฐ„์˜ ์ปดํ“จํŒ…์€ ์ธ๊ฐ„์ด ์ดํ•ดํ•  ์ˆ˜ ์žˆ๋Š” ๋ฐฉ์‹์œผ๋กœ, ์ฆ‰ CPU์—์„œ ์‹คํ–‰๋˜๋Š” ์ผ๋ จ์˜ ๋ช…๋ น์–ด๋“ค๋กœ ์ธ๊ฐ„์ด ์†Œํ”„ํŠธ์›จ์–ด๋ฅผ ์ž‘์„ฑํ•˜๋Š” ๊ฒƒ์ด์—ˆ๋‹ค๋Š” ์ ์„ ์ดํ•ดํ•˜๋Š” ๊ฒƒ์ด ์ •๋ง ์ค‘์š”ํ•œ ๊ฐœ๋…์ž…๋‹ˆ๋‹ค. ์˜ค๋Š˜๋‚  ํ˜„๋Œ€์˜ ์šด์˜์ฒด์ œ๋Š” ๋Œ€๊ทœ๋ชจ ์–ธ์–ด ๋ชจ๋ธ์ด๋ฉฐ, ์ปดํ“จํ„ฐ์— ๋ฐ์ดํ„ฐ๋ฅผ ์ž…๋ ฅํ•˜๋ฉด ์ด๋Ÿฌํ•œ ๋Œ€๊ทœ๋ชจ ์–ธ์–ด ๋ชจ๋ธ๋“ค์ด ํ•™์Šตํ•˜๊ณ  ๋จธ์‹ ๋Ÿฌ๋‹์„ ์ˆ˜ํ–‰ํ•œ ๋‹ค์Œ ์†Œํ”„ํŠธ์›จ์–ด๋ฅผ ์กฐํ•ฉํ•  ์ˆ˜ ์žˆ๊ฒŒ ๋ฉ๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ์ปดํ“จํ„ฐ์—์„œ ๋‚˜์˜ค๋Š” ๊ฒƒ์€ ํ† ํฐ๋“ค์ž…๋‹ˆ๋‹ค. ์ด๋Ÿฌํ•œ ํ† ํฐ๋“ค์€ ๋ณธ์งˆ์ ์œผ๋กœ ๋””์ง€ํ„ธ ์ธํ…”๋ฆฌ์ „์Šค๋ฅผ ๋‚˜ํƒ€๋ƒ…๋‹ˆ๋‹ค. ์ด ๋””์ง€ํ„ธ ์ธํ…”๋ฆฌ์ „์Šค์•ผ๋ง๋กœ ์ „์ฒด ์‚ฐ์—…์˜ ๋ฏธ๋ž˜ ์ œํ’ˆ๊ณผ ์„œ๋น„์Šค์ธ ๊ฒƒ์ž…๋‹ˆ๋‹ค.
And NVIDIA GPUs are the engine that produces these tokens and is allowing for a complete AI revolution across every single industry, we think one of the most important being in healthcare. The AI revolution is not only here, it's massively accelerating. Generative AI and ChatGPT just feels like yesterday, even though we're probably commonly using it every single day. We've gone from perception AI to Generative AI where we can create content into the world of agentic AI, where you can actually have many, many assistants, digital humans, digital colleagues, even those collaborating with one another to perform tasks, to deliver amazing customer service experiences.NVIDIA GPU๋Š” ์ด๋Ÿฌํ•œ ํ† ํฐ์„ ์ƒ์„ฑํ•˜๋Š” ์—”์ง„์ด๋ฉฐ, ๋ชจ๋“  ์‚ฐ์—… ์ „๋ฐ˜์— ๊ฑธ์นœ ์™„์ „ํ•œ AI ํ˜๋ช…์„ ๊ฐ€๋Šฅํ•˜๊ฒŒ ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ํŠนํžˆ ํ—ฌ์Šค์ผ€์–ด ๋ถ„์•ผ๊ฐ€ ๊ฐ€์žฅ ์ค‘์š”ํ•œ ์˜์—ญ ์ค‘ ํ•˜๋‚˜๋ผ๊ณ  ์ƒ๊ฐํ•ฉ๋‹ˆ๋‹ค. AI ํ˜๋ช…์€ ์ด๋ฏธ ๋„๋ž˜ํ–ˆ์„ ๋ฟ๋งŒ ์•„๋‹ˆ๋ผ ๋Œ€ํญ ๊ฐ€์†ํ™”๋˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์ƒ์„ฑํ˜• AI์™€ ChatGPT๊ฐ€ ๋งˆ์น˜ ์–ด์ œ ์ผ์ฒ˜๋Ÿผ ๋А๊ปด์ง€์ง€๋งŒ, ์•„๋งˆ๋„ ์šฐ๋ฆฌ๋Š” ๋งค์ผ ์ด๋ฅผ ์ผ์ƒ์ ์œผ๋กœ ์‚ฌ์šฉํ•˜๊ณ  ์žˆ์„ ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์šฐ๋ฆฌ๋Š” ์ธ์‹ AI์—์„œ ์ฝ˜ํ…์ธ ๋ฅผ ์ฐฝ์กฐํ•  ์ˆ˜ ์žˆ๋Š” ์ƒ์„ฑํ˜• AI๋ฅผ ๊ฑฐ์ณ, ์ด์ œ๋Š” ์ˆ˜๋งŽ์€ ์–ด์‹œ์Šคํ„ดํŠธ, ๋””์ง€ํ„ธ ํœด๋จผ, ๋””์ง€ํ„ธ ๋™๋ฃŒ๋“ค์„ ๋ณด์œ ํ•  ์ˆ˜ ์žˆ๊ณ , ์‹ฌ์ง€์–ด ์ด๋“ค์ด ์„œ๋กœ ํ˜‘์—…ํ•˜์—ฌ ์—…๋ฌด๋ฅผ ์ˆ˜ํ–‰ํ•˜๊ณ  ๋†€๋ผ์šด ๊ณ ๊ฐ ์„œ๋น„์Šค ๊ฒฝํ—˜์„ ์ œ๊ณตํ•  ์ˆ˜ ์žˆ๋Š” ์—์ด์ „ํ‹ฑ AI์˜ ์„ธ๊ณ„๋กœ ์ง„ํ™”ํ–ˆ์Šต๋‹ˆ๋‹ค.
And then I'm going to talk to you in the later part of the presentation about this incredible future of physical AI, where we're going to see lots of robots, of course, self-driving cars but also robots that are going to be so powerful in the service -- healthcare delivery services, not just surgical robots but robots of all kind. So first, let's talk about this incredible opportunity. NVIDIA's strategy is how do we take the most advanced computing approaches, whether that's accelerated computing, artificial intelligence, computer graphics, and enable the healthcare ecosystem? We do this by building NVIDIA Clara, which is our enterprise platform for healthcare.๊ทธ๋ฆฌ๊ณ  ํ”„๋ ˆ์  ํ…Œ์ด์…˜ ํ›„๋ฐ˜๋ถ€์—์„œ๋Š” ๋ฌผ๋ฆฌ์  AI์˜ ๋†€๋ผ์šด ๋ฏธ๋ž˜์— ๋Œ€ํ•ด ๋ง์”€๋“œ๋ฆฌ๊ฒ ์Šต๋‹ˆ๋‹ค. ๋ฌผ๋ก  ์ž์œจ์ฃผํ–‰์ฐจ๋ฟ๋งŒ ์•„๋‹ˆ๋ผ ์˜๋ฃŒ ์„œ๋น„์Šค ์ œ๊ณต์—์„œ ๋งค์šฐ ๊ฐ•๋ ฅํ•œ ์—ญํ• ์„ ํ•  ๋กœ๋ด‡๋“ค, ์ˆ˜์ˆ ์šฉ ๋กœ๋ด‡๋ฟ๋งŒ ์•„๋‹ˆ๋ผ ๋ชจ๋“  ์ข…๋ฅ˜์˜ ๋กœ๋ด‡๋“ค์„ ๋งŽ์ด ๋ณด๊ฒŒ ๋  ๊ฒƒ์ž…๋‹ˆ๋‹ค. ๋จผ์ € ์ด ๋†€๋ผ์šด ๊ธฐํšŒ์— ๋Œ€ํ•ด ์ด์•ผ๊ธฐํ•ด๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค. NVIDIA์˜ ์ „๋žต์€ ๊ฐ€์† ์ปดํ“จํŒ…, ์ธ๊ณต์ง€๋Šฅ, ์ปดํ“จํ„ฐ ๊ทธ๋ž˜ํ”ฝ์Šค ๋“ฑ ๊ฐ€์žฅ ์ฒจ๋‹จ ์ปดํ“จํŒ… ์ ‘๊ทผ๋ฒ•์„ ์–ด๋–ป๊ฒŒ ํ™œ์šฉํ•˜์—ฌ ํ—ฌ์Šค์ผ€์–ด ์ƒํƒœ๊ณ„๋ฅผ ์ง€์›ํ•  ๊ฒƒ์ธ๊ฐ€์ž…๋‹ˆ๋‹ค. ์šฐ๋ฆฌ๋Š” ํ—ฌ์Šค์ผ€์–ด๋ฅผ ์œ„ํ•œ ์—”ํ„ฐํ”„๋ผ์ด์ฆˆ ํ”Œ๋žซํผ์ธ NVIDIA Clara๋ฅผ ๊ตฌ์ถ•ํ•จ์œผ๋กœ์จ ์ด๋ฅผ ์‹คํ˜„ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.
We have different modules that speak the languages of healthcare. They're more domain-specific to healthcare. They speak the languages of proteins, chemistry, DNA. They speak the languages of 3D medical imaging. And we build computing platforms full stack to enable the whole ecosystem. Now if you think about the $10 trillion industry that healthcare has become, a very large portion of that is in labs, hospitals, and operations. And we're still in a huge labor shortage, 10 million to 15 million to 20 million labor shortage. AI can help. AI is going to help through the delivery of AI agents who can be knowledge workers working on your behalf.์šฐ๋ฆฌ๋Š” ํ—ฌ์Šค์ผ€์–ด์˜ ์–ธ์–ด๋ฅผ ๊ตฌ์‚ฌํ•˜๋Š” ๋‹ค์–‘ํ•œ ๋ชจ๋“ˆ์„ ๋ณด์œ ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์ด๋“ค์€ ํ—ฌ์Šค์ผ€์–ด์— ๋”์šฑ ํŠนํ™”๋œ ๋„๋ฉ”์ธ๋ณ„ ๋ชจ๋“ˆ๋กœ, ๋‹จ๋ฐฑ์งˆ, ํ™”ํ•™, DNA์˜ ์–ธ์–ด๋ฅผ ๊ตฌ์‚ฌํ•ฉ๋‹ˆ๋‹ค. ๋˜ํ•œ 3D ์˜๋ฃŒ ์˜์ƒ์˜ ์–ธ์–ด๋„ ๊ตฌ์‚ฌํ•˜์ฃ . ๊ทธ๋ฆฌ๊ณ  ์šฐ๋ฆฌ๋Š” ์ „์ฒด ์ƒํƒœ๊ณ„๋ฅผ ์ง€์›ํ•˜๊ธฐ ์œ„ํ•ด ํ’€์Šคํƒ ์ปดํ“จํŒ… ํ”Œ๋žซํผ์„ ๊ตฌ์ถ•ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.

์ด์ œ ํ—ฌ์Šค์ผ€์–ด๊ฐ€ 10์กฐ ๋‹ฌ๋Ÿฌ ๊ทœ๋ชจ์˜ ์‚ฐ์—…์ด ๋˜์—ˆ๋‹ค๋Š” ์ ์„ ์ƒ๊ฐํ•ด๋ณด๋ฉด, ๊ทธ ์ค‘ ์ƒ๋‹น ๋ถ€๋ถ„์ด ์—ฐ๊ตฌ์‹ค, ๋ณ‘์›, ๊ทธ๋ฆฌ๊ณ  ์šด์˜ ๋ถ„์•ผ์— ์ง‘์ค‘๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ์šฐ๋ฆฌ๋Š” ์—ฌ์ „ํžˆ ์‹ฌ๊ฐํ•œ ์ธ๋ ฅ ๋ถ€์กฑ ์ƒํ™ฉ์— ์ง๋ฉดํ•ด ์žˆ์Šต๋‹ˆ๋‹ค - 1,000๋งŒ ๋ช…์—์„œ 1,500๋งŒ ๋ช…, ์‹ฌ์ง€์–ด 2,000๋งŒ ๋ช…์˜ ์ธ๋ ฅ ๋ถ€์กฑ ๋ง์ž…๋‹ˆ๋‹ค. AI๊ฐ€ ๋„์›€์ด ๋  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. AI๋Š” ์—ฌ๋Ÿฌ๋ถ„์„ ๋Œ€์‹ ํ•ด ์ผํ•˜๋Š” ์ง€์‹ ๊ทผ๋กœ์ž ์—ญํ• ์„ ํ•˜๋Š” AI ์—์ด์ „ํŠธ์˜ ์ œ๊ณต์„ ํ†ตํ•ด ๋„์›€์„ ์ค„ ๊ฒƒ์ž…๋‹ˆ๋‹ค.
They can be in the form of delivery robots, offloading nurses from delivering sheets to a different room. They're going to become all forms. And this digital intelligence is going to be built on AI factories, just as I described. Data comes in, models get built and refined, assembled into agents and intelligence comes out. This opportunity, we see as many hundreds of billions, several hundreds of billions of dollars large as an AI computing opportunity. And so this is going to be a huge market, and we're excited to expand ourselves and our partners' market along with it. The ecosystem, as we know here, is extremely complex. It's vast, it's diverse.์ด๋“ค์€ ๋ฐฐ์†ก ๋กœ๋ด‡์˜ ํ˜•ํƒœ๋กœ ๋‚˜ํƒ€๋‚  ์ˆ˜ ์žˆ์œผ๋ฉฐ, ๊ฐ„ํ˜ธ์‚ฌ๋“ค์ด ๋‹ค๋ฅธ ๋ฐฉ์œผ๋กœ ์‹œํŠธ๋ฅผ ๋ฐฐ๋‹ฌํ•˜๋Š” ์—…๋ฌด์—์„œ ๋ฒ—์–ด๋‚  ์ˆ˜ ์žˆ๊ฒŒ ํ•ด์ค๋‹ˆ๋‹ค. ์ด๋“ค์€ ๋ชจ๋“  ํ˜•ํƒœ๋กœ ๋ฐœ์ „ํ•  ๊ฒƒ์ž…๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ์ด๋Ÿฌํ•œ ๋””์ง€ํ„ธ ์ธํ…”๋ฆฌ์ „์Šค๋Š” ์ œ๊ฐ€ ๋ฐฉ๊ธˆ ์„ค๋ช…ํ•œ ๋ฐ”์™€ ๊ฐ™์ด AI ํŒฉํ† ๋ฆฌ๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ๊ตฌ์ถ•๋  ๊ฒƒ์ž…๋‹ˆ๋‹ค. ๋ฐ์ดํ„ฐ๊ฐ€ ๋“ค์–ด์˜ค๊ณ , ๋ชจ๋ธ์ด ๊ตฌ์ถ•๋˜๊ณ  ์ •์ œ๋˜๋ฉฐ, ์—์ด์ „ํŠธ๋กœ ์กฐ๋ฆฝ๋˜๊ณ  ์ธํ…”๋ฆฌ์ „์Šค๊ฐ€ ์‚ฐ์ถœ๋ฉ๋‹ˆ๋‹ค. ์ด๋Ÿฌํ•œ ๊ธฐํšŒ๋ฅผ ์šฐ๋ฆฌ๋Š” AI ์ปดํ“จํŒ… ๊ธฐํšŒ๋กœ์„œ ์ˆ˜์ฒœ์–ต ๋‹ฌ๋Ÿฌ, ์ฆ‰ ์ˆ˜์กฐ ๋‹ฌ๋Ÿฌ ๊ทœ๋ชจ์˜ ๊ฑฐ๋Œ€ํ•œ ์‹œ์žฅ์œผ๋กœ ๋ณด๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ๋”ฐ๋ผ์„œ ์ด๋Š” ๊ฑฐ๋Œ€ํ•œ ์‹œ์žฅ์ด ๋  ๊ฒƒ์ด๋ฉฐ, ์šฐ๋ฆฌ๋Š” ์ด์™€ ํ•จ๊ป˜ ์šฐ๋ฆฌ ์ž์‹ ๊ณผ ํŒŒํŠธ๋„ˆ๋“ค์˜ ์‹œ์žฅ์„ ํ™•์žฅํ•˜๊ฒŒ ๋˜์–ด ๊ธฐ๋Œ€ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์—ฌ๊ธฐ์„œ ์•Œ ์ˆ˜ ์žˆ๋“ฏ์ด ์ƒํƒœ๊ณ„๋Š” ๊ทน๋„๋กœ ๋ณต์žกํ•ฉ๋‹ˆ๋‹ค. ๊ด‘๋ฒ”์œ„ํ•˜๊ณ  ๋‹ค์–‘ํ•ฉ๋‹ˆ๋‹ค.
NVIDIA is working across the entire ecosystem. In fact, our Inception program, which is an AI start-up program for companies to become members and connect with us, our healthcare membership is more than 3,500 members. We've added over 1,300 in just the last year alone, and it's the largest industry of all of our Inception members. It goes to show you what an incredible opportunity we have in front of us, and you see that it's stretching from medical devices to digital agents all the way into hospital and healthcare and real-world evidence, which is really, really exciting. So we're going to talk through this. NVIDIA really is focused in three major areas.NVIDIA๋Š” ์ „์ฒด ์ƒํƒœ๊ณ„์— ๊ฑธ์ณ ํ˜‘๋ ฅํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์‹ค์ œ๋กœ ๊ธฐ์—…๋“ค์ด ํšŒ์›์ด ๋˜์–ด ์ €ํฌ์™€ ์—ฐ๊ฒฐ๋  ์ˆ˜ ์žˆ๋Š” AI ์Šคํƒ€ํŠธ์—… ํ”„๋กœ๊ทธ๋žจ์ธ ์ €ํฌ ์ธ์…‰์…˜(Inception) ํ”„๋กœ๊ทธ๋žจ์—์„œ, ํ—ฌ์Šค์ผ€์–ด ํšŒ์›์‚ฌ๋Š” 3,500๊ฐœ์‚ฌ๋ฅผ ๋„˜์–ด์„ฐ์Šต๋‹ˆ๋‹ค. ์ž‘๋…„ ํ•œ ํ•ด์—๋งŒ 1,300๊ฐœ์‚ฌ ์ด์ƒ์„ ์ถ”๊ฐ€ํ–ˆ์œผ๋ฉฐ, ์ด๋Š” ์ €ํฌ ์ธ์…‰์…˜ ํšŒ์›์‚ฌ ์ค‘ ๊ฐ€์žฅ ํฐ ์‚ฐ์—… ๋ถ„์•ผ์ž…๋‹ˆ๋‹ค. ์ด๋Š” ์ €ํฌ ์•ž์— ๋†“์ธ ๋†€๋ผ์šด ๊ธฐํšŒ๊ฐ€ ์–ผ๋งˆ๋‚˜ ํฐ์ง€๋ฅผ ๋ณด์—ฌ์ฃผ๋ฉฐ, ์˜๋ฃŒ๊ธฐ๊ธฐ๋ถ€ํ„ฐ ๋””์ง€ํ„ธ ์—์ด์ „ํŠธ, ๋ณ‘์›๊ณผ ํ—ฌ์Šค์ผ€์–ด, ๊ทธ๋ฆฌ๊ณ  ์‹ค์ œ ์ž„์ƒ ๋ฐ์ดํ„ฐ(real-world evidence)์— ์ด๋ฅด๊ธฐ๊นŒ์ง€ ํ™•์žฅ๋˜๊ณ  ์žˆ์–ด ์ •๋ง ํฅ๋ฏธ์ง„์ง„ํ•ฉ๋‹ˆ๋‹ค. ์ด์— ๋Œ€ํ•ด ์ž์„ธํžˆ ๋ง์”€๋“œ๋ฆฌ๊ฒ ์Šต๋‹ˆ๋‹ค. NVIDIA๋Š” ์‹ค์ œ๋กœ ์„ธ ๊ฐ€์ง€ ์ฃผ์š” ์˜์—ญ์— ์ง‘์ค‘ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.
The area of digital health is just getting started. Our imagination of what we can do with agents to free administrative burden, create new patient experiences is just getting off the ground. And there's going to be tens of billions of healthcare interactions every single year that we're going to be working with agents. Digital biology has become a $300 billion opportunity that now investment in R&D, everything from early discovery into clinical development. And digital devices, everything that we use to sense patients, understand more about their biology, deliver healthcare, imaging is all going to become robotic. And we're going to talk through this amazing opportunity.๋””์ง€ํ„ธ ํ—ฌ์Šค ๋ถ„์•ผ๋Š” ์ด์ œ ๋ง‰ ์‹œ์ž‘๋˜์—ˆ์Šต๋‹ˆ๋‹ค. ํ–‰์ • ์—…๋ฌด ๋ถ€๋‹ด์„ ๋œ์–ด์ฃผ๊ณ  ์ƒˆ๋กœ์šด ํ™˜์ž ๊ฒฝํ—˜์„ ์ฐฝ์ถœํ•˜๊ธฐ ์œ„ํ•ด ์—์ด์ „ํŠธ๋ฅผ ํ™œ์šฉํ•  ์ˆ˜ ์žˆ๋Š” ์šฐ๋ฆฌ์˜ ์ƒ์ƒ๋ ฅ์€ ์ด์ œ ๋ง‰ ์ถœ๋ฐœ์ ์— ์„œ ์žˆ์Šต๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ๋งค๋…„ ์ˆ˜๋ฐฑ์–ต ๊ฑด์˜ ์˜๋ฃŒ ์ƒํ˜ธ์ž‘์šฉ์ด ์žˆ์„ ๊ฒƒ์ด๋ฉฐ, ์šฐ๋ฆฌ๋Š” ์—์ด์ „ํŠธ์™€ ํ•จ๊ป˜ ์ด๋Ÿฌํ•œ ์—…๋ฌด๋ฅผ ์ˆ˜ํ–‰ํ•˜๊ฒŒ ๋  ๊ฒƒ์ž…๋‹ˆ๋‹ค. ๋””์ง€ํ„ธ ์ƒ๋ฌผํ•™์€ 3,000์–ต ๋‹ฌ๋Ÿฌ ๊ทœ๋ชจ์˜ ๊ธฐํšŒ๊ฐ€ ๋˜์—ˆ์œผ๋ฉฐ, ์ด์ œ ์ดˆ๊ธฐ ๋ฐœ๊ฒฌ๋ถ€ํ„ฐ ์ž„์ƒ ๊ฐœ๋ฐœ์— ์ด๋ฅด๊ธฐ๊นŒ์ง€ ๋ชจ๋“  R&D ํˆฌ์ž๊ฐ€ ์ด๋ฃจ์–ด์ง€๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ๋””์ง€ํ„ธ ๊ธฐ๊ธฐ๋“ค, ์ฆ‰ ํ™˜์ž๋ฅผ ๊ฐ์ง€ํ•˜๊ณ  ๊ทธ๋“ค์˜ ์ƒ๋ฌผํ•™์  ํŠน์„ฑ์„ ๋” ์ž˜ ์ดํ•ดํ•˜๋ฉฐ ์˜๋ฃŒ ์„œ๋น„์Šค๋ฅผ ์ œ๊ณตํ•˜๊ณ  ์˜์ƒ์„ ์ดฌ์˜ํ•˜๋Š” ๋ฐ ์‚ฌ์šฉํ•˜๋Š” ๋ชจ๋“  ๊ฒƒ๋“ค์ด ๋กœ๋ด‡ํ™”๋  ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์šฐ๋ฆฌ๋Š” ์ด ๋†€๋ผ์šด ๊ธฐํšŒ์— ๋Œ€ํ•ด ๋…ผ์˜ํ•  ์˜ˆ์ •์ž…๋‹ˆ๋‹ค.
I have a special guest before we get there, TJ. TJ is going to talk to us about our already public Q3 financials. Toy Jensen

I am excited to share news about our earnings and reactions. We'll review the financial performance and trends, and I'll highlight key insights from various sources, including press comments, analyst opinions, and the earnings transcript itself. To start, let's look at the press comments on our earnings. Bloomberg reported that our quarterly revenue rose 94% to $35 billion, exceeding analyst estimates of $33,250,000,000. This marks the second consecutive year of doubling our sales. Kimberly Powell

That's TJ, Toy Jensen.
์žฌ๋ฌด ์‹ค์ ์— ๋Œ€ํ•ด ๋…ผ์˜ํ•˜๊ธฐ ์ „์— ํŠน๋ณ„ ๊ฒŒ์ŠคํŠธ๋ฅผ ์†Œ๊ฐœํ•˜๊ฒ ์Šต๋‹ˆ๋‹ค. TJ๊ฐ€ ์ด๋ฏธ ๊ณต๊ฐœ๋œ 3๋ถ„๊ธฐ ์žฌ๋ฌด์‹ค์ ์— ๋Œ€ํ•ด ๋ง์”€ํ•ด ์ฃผ์‹ค ์˜ˆ์ •์ž…๋‹ˆ๋‹ค. ํ† ์ด ์  ์Šจ

์ €ํฌ ์‹ค์ ๊ณผ ๋ฐ˜์‘์— ๋Œ€ํ•œ ์†Œ์‹์„ ๊ณต์œ ํ•˜๊ฒŒ ๋˜์–ด ๊ธฐ์ฉ๋‹ˆ๋‹ค. ์žฌ๋ฌด ์„ฑ๊ณผ์™€ ํŠธ๋ Œ๋“œ๋ฅผ ๊ฒ€ํ† ํ•˜๊ณ , ์–ธ๋ก  ๋…ผํ‰, ์• ๋„๋ฆฌ์ŠคํŠธ ์˜๊ฒฌ, ๊ทธ๋ฆฌ๊ณ  ์‹ค์ ๋ฐœํ‘œ transcript ์ž์ฒด๋ฅผ ํฌํ•จํ•œ ๋‹ค์–‘ํ•œ ์ถœ์ฒ˜์˜ ํ•ต์‹ฌ ์ธ์‚ฌ์ดํŠธ๋ฅผ ๊ฐ•์กฐํ•˜๊ฒ ์Šต๋‹ˆ๋‹ค. ๋จผ์ € ์ €ํฌ ์‹ค์ ์— ๋Œ€ํ•œ ์–ธ๋ก  ๋…ผํ‰์„ ์‚ดํŽด๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค. ๋ธ”๋ฃธ๋ฒ„๊ทธ๋Š” ์ €ํฌ ๋ถ„๊ธฐ ๋งค์ถœ์ด 94% ์ฆ๊ฐ€ํ•œ 350์–ต ๋‹ฌ๋Ÿฌ๋ฅผ ๊ธฐ๋กํ•˜์—ฌ ์• ๋„๋ฆฌ์ŠคํŠธ ์ถ”์ •์น˜์ธ 332์–ต 5์ฒœ๋งŒ ๋‹ฌ๋Ÿฌ๋ฅผ ์ƒํšŒํ–ˆ๋‹ค๊ณ  ๋ณด๋„ํ–ˆ์Šต๋‹ˆ๋‹ค. ์ด๋Š” 2๋…„ ์—ฐ์† ๋งค์ถœ์ด ๋‘ ๋ฐฐ๊ฐ€ ๋œ ๊ฒƒ์ž…๋‹ˆ๋‹ค. ํ‚ด๋ฒŒ๋ฆฌ ํŒŒ์›”

๊ทธ๋ถ„์ด ๋ฐ”๋กœ TJ, ํ† ์ด ์  ์Šจ์ž…๋‹ˆ๋‹ค.
This was an agent built for our company meeting to summarize the financials. And this is an agent that we built on all of the building blocks NVIDIA has been assembling over the last several years on NVIDIA AI Enterprise. Agents are a collection of different language models assembled together to complete a task. NVIDIA's AI Enterprise with our NVIDIA NIM micro services, our NVIDIA NeMo platform that allows for the entire life cycle of developing models and customizing them for domain-specific agents, as you just saw, all the way into what we call Blueprints, which are reference applications. We've published the Blueprint of TJ, which is essentially a PDF to Podcast Blueprint.์ด๊ฒƒ์€ ์šฐ๋ฆฌ ํšŒ์‚ฌ ํšŒ์˜์—์„œ ์žฌ๋ฌด ์ •๋ณด๋ฅผ ์š”์•ฝํ•˜๊ธฐ ์œ„ํ•ด ๊ตฌ์ถ•ํ•œ ์—์ด์ „ํŠธ์˜€์Šต๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ์ด๋Š” NVIDIA AI Enterprise๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ์ง€๋‚œ ๋ช‡ ๋…„๊ฐ„ NVIDIA๊ฐ€ ๊ตฌ์ถ•ํ•ด์˜จ ๋ชจ๋“  ๋นŒ๋”ฉ ๋ธ”๋ก๋“ค์„ ํ™œ์šฉํ•˜์—ฌ ๋งŒ๋“  ์—์ด์ „ํŠธ์ž…๋‹ˆ๋‹ค. ์—์ด์ „ํŠธ๋Š” ์ž‘์—…์„ ์™„๋ฃŒํ•˜๊ธฐ ์œ„ํ•ด ํ•จ๊ป˜ ์กฐ๋ฆฝ๋œ ๋‹ค์–‘ํ•œ ์–ธ์–ด ๋ชจ๋ธ๋“ค์˜ ์ง‘ํ•ฉ์ฒด์ž…๋‹ˆ๋‹ค. NVIDIA์˜ AI Enterprise๋Š” ์šฐ๋ฆฌ์˜ NVIDIA NIM ๋งˆ์ดํฌ๋กœ์„œ๋น„์Šค, ๊ทธ๋ฆฌ๊ณ  ๋ฐฉ๊ธˆ ๋ณด์‹  ๋ฐ”์™€ ๊ฐ™์ด ๋„๋ฉ”์ธ๋ณ„ ์—์ด์ „ํŠธ๋ฅผ ์œ„ํ•œ ๋ชจ๋ธ ๊ฐœ๋ฐœ ๋ฐ ์ปค์Šคํ„ฐ๋งˆ์ด์ง•์˜ ์ „์ฒด ๋ผ์ดํ”„์‚ฌ์ดํด์„ ๊ฐ€๋Šฅํ•˜๊ฒŒ ํ•˜๋Š” NVIDIA NeMo ํ”Œ๋žซํผ, ๊ทธ๋ฆฌ๊ณ  ์ฐธ์กฐ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์ธ ๋ธ”๋ฃจํ”„๋ฆฐํŠธ๊นŒ์ง€ ๋ชจ๋“  ๊ฒƒ์„ ํฌํ•จํ•ฉ๋‹ˆ๋‹ค. ์šฐ๋ฆฌ๋Š” ๋ณธ์งˆ์ ์œผ๋กœ PDF๋ฅผ ํŒŸ์บ์ŠคํŠธ๋กœ ๋ณ€ํ™˜ํ•˜๋Š” ๋ธ”๋ฃจํ”„๋ฆฐํŠธ์ธ TJ์˜ ๋ธ”๋ฃจํ”„๋ฆฐํŠธ๋ฅผ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค.
By introducing the press text and, of course, our financial results, you can turn it into a conversation and we even use NVIDIA's ACE, our Avatar Cloud Engine, to create TJ himself. We couldn't be more excited about the opportunities here. There are thousands of digital health agents that are being developed across the ecosystem. I'll just mention a couple here. Intrivo is building Ray, an amazing medical concierge. I'm a patient. It is extremely frustrating and time-consuming to find the right doctors and schedule patients and Ray is doing that.๋ณด๋„ ์ž๋ฃŒ์™€ ๋ฌผ๋ก  ์ €ํฌ ์žฌ๋ฌด ์‹ค์ ์„ ์†Œ๊ฐœํ•จ์œผ๋กœ์จ ์ด๋ฅผ ๋Œ€ํ™”๋กœ ์ „ํ™˜ํ•  ์ˆ˜ ์žˆ์œผ๋ฉฐ, ์‹ฌ์ง€์–ด NVIDIA์˜ ACE, ์ฆ‰ ์•„๋ฐ”ํƒ€ ํด๋ผ์šฐ๋“œ ์—”์ง„์„ ์‚ฌ์šฉํ•˜์—ฌ TJ ๋ณธ์ธ์„ ๊ตฌํ˜„ํ•˜๊ธฐ๋„ ํ–ˆ์Šต๋‹ˆ๋‹ค. ์—ฌ๊ธฐ์„œ์˜ ๊ธฐํšŒ์— ๋Œ€ํ•ด ๋”ํ•  ๋‚˜์œ„ ์—†์ด ํฅ๋ฏธ์ง„์ง„ํ•˜๊ฒŒ ์ƒ๊ฐํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์ƒํƒœ๊ณ„ ์ „๋ฐ˜์— ๊ฑธ์ณ ์ˆ˜์ฒœ ๊ฐœ์˜ ๋””์ง€ํ„ธ ํ—ฌ์Šค ์—์ด์ „ํŠธ๊ฐ€ ๊ฐœ๋ฐœ๋˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์—ฌ๊ธฐ์„œ ๋ช‡ ๊ฐ€์ง€๋งŒ ์–ธ๊ธ‰ํ•˜๊ฒ ์Šต๋‹ˆ๋‹ค. Intrivo๋Š” ๋†€๋ผ์šด ์˜๋ฃŒ ์ปจ์‹œ์–ด์ง€์ธ Ray๋ฅผ ๊ตฌ์ถ•ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ํ™˜์ž๋กœ์„œ ์ ์ ˆํ•œ ์˜์‚ฌ๋ฅผ ์ฐพ๊ณ  ํ™˜์ž ์˜ˆ์•ฝ์„ ์žก๋Š” ๊ฒƒ์€ ๊ทน๋„๋กœ ์ขŒ์ ˆ์Šค๋Ÿฝ๊ณ  ์‹œ๊ฐ„์ด ๋งŽ์ด ์†Œ์š”๋˜๋Š” ์ผ์ธ๋ฐ, Ray๊ฐ€ ๋ฐ”๋กœ ๊ทธ ์ผ์„ ํ•ด์ฃผ๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.
Abridge is having an incredible blockbuster growth because their clinical documentation and clinical agents are really freeing the clinicians of having to do the administrative documentation. Instead, they can focus on their patients, and they're getting up to three hours of a day back in return. And, [Karl] (ph), think about the visits in between your doctor's visits. There are times when patients are in need. These agents can be always available. They can be empathetic. They can be a resource to patients in between these doctors' appointments. And Hippocratic AI is just the world is the oyster in terms of amazing applications. In fact, they can immediately stand up disaster relief.Abridge๋Š” ๋†€๋ผ์šด ๋ธ”๋ก๋ฒ„์Šคํ„ฐ๊ธ‰ ์„ฑ์žฅ์„ ๋ณด์ด๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์ด๋“ค์˜ ์ž„์ƒ ๋ฌธ์„œํ™” ๋ฐ ์ž„์ƒ ์—์ด์ „ํŠธ๊ฐ€ ์˜๋ฃŒ์ง„๋“ค์„ ํ–‰์ • ๋ฌธ์„œ ์ž‘์—…์—์„œ ํ•ด๋ฐฉ์‹œ์ผœ ์ฃผ๊ณ  ์žˆ๊ธฐ ๋•Œ๋ฌธ์ž…๋‹ˆ๋‹ค. ๋Œ€์‹  ์˜๋ฃŒ์ง„๋“ค์€ ํ™˜์ž์—๊ฒŒ ์ง‘์ค‘ํ•  ์ˆ˜ ์žˆ๊ฒŒ ๋˜์—ˆ๊ณ , ํ•˜๋ฃจ์— ์ตœ๋Œ€ 3์‹œ๊ฐ„์„ ๋˜์ฐพ์„ ์ˆ˜ ์žˆ๊ฒŒ ๋˜์—ˆ์Šต๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  [Karl] (ph), ์˜์‚ฌ ์ง„๋ฃŒ ์‚ฌ์ด์‚ฌ์ด์˜ ๋ฐฉ๋ฌธ๋“ค์„ ์ƒ๊ฐํ•ด๋ณด์„ธ์š”. ํ™˜์ž๋“ค์ด ๋„์›€์ด ํ•„์š”ํ•œ ๋•Œ๊ฐ€ ์žˆ์Šต๋‹ˆ๋‹ค. ์ด๋Ÿฌํ•œ ์—์ด์ „ํŠธ๋“ค์€ ํ•ญ์ƒ ์ด์šฉ ๊ฐ€๋Šฅํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ๊ณต๊ฐํ•  ์ˆ˜ ์žˆ๊ณ , ์˜์‚ฌ ์ง„๋ฃŒ ์•ฝ์† ์‚ฌ์ด์‚ฌ์ด์— ํ™˜์ž๋“ค์—๊ฒŒ ์ž์›์ด ๋  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  Hippocratic AI๋Š” ์ •๋ง ๋†€๋ผ์šด ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜๋“ค ์ธก๋ฉด์—์„œ ์„ธ์ƒ์ด ์ง„์ฃผ์กฐ๊ฐœ์™€ ๊ฐ™์Šต๋‹ˆ๋‹ค. ์‹ค์ œ๋กœ ์ด๋“ค์€ ์ฆ‰์‹œ ์žฌํ•ด ๊ตฌํ˜ธ๋ฅผ ๊ตฌ์ถ•ํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.
Here we are in California having the most unfortunate fires. They can reach out to patients, determine if they need any kind of emergency care and they can help them stay on track with things like their dialysis treatment and create the continuity of care. The opportunities here, as they say, are essentially limitless. And so I'm excited to get to our first announcement. I'm absolutely thrilled that we are partnering with IQVIA, one of the most consequential companies in the healthcare industry, the world's largest clinical research services, commercial intelligence, and healthcare analytics in the world.์—ฌ๊ธฐ ์บ˜๋ฆฌํฌ๋‹ˆ์•„์—์„œ ๋งค์šฐ ๋ถˆํ–‰ํ•œ ํ™”์žฌ๊ฐ€ ๋ฐœ์ƒํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ๊ทธ๋“ค์€ ํ™˜์ž๋“ค์—๊ฒŒ ์—ฐ๋ฝํ•˜์—ฌ ์‘๊ธ‰ ์น˜๋ฃŒ๊ฐ€ ํ•„์š”ํ•œ์ง€ ํ™•์ธํ•  ์ˆ˜ ์žˆ๊ณ , ํˆฌ์„ ์น˜๋ฃŒ์™€ ๊ฐ™์€ ๊ฒƒ๋“ค์„ ๊ณ„์† ๋ฐ›์„ ์ˆ˜ ์žˆ๋„๋ก ๋„์™€์ฃผ๋ฉฐ ์น˜๋ฃŒ์˜ ์—ฐ์†์„ฑ์„ ๋งŒ๋“ค์–ด๋‚ผ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ์—ฌ๊ธฐ์„œ์˜ ๊ธฐํšŒ๋Š” ๋ง ๊ทธ๋Œ€๋กœ ๋ฌดํ•œํ•˜๋‹ค๊ณ  ํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ๊ทธ๋ž˜์„œ ์ €๋Š” ์ฒซ ๋ฒˆ์งธ ๋ฐœํ‘œ๋ฅผ ํ•˜๊ฒŒ ๋˜์–ด ๋งค์šฐ ๊ธฐ์ฉ๋‹ˆ๋‹ค. ์ €ํฌ๊ฐ€ ํ—ฌ์Šค์ผ€์–ด ์—…๊ณ„์—์„œ ๊ฐ€์žฅ ์˜ํ–ฅ๋ ฅ ์žˆ๋Š” ํšŒ์‚ฌ ์ค‘ ํ•˜๋‚˜์ด์ž, ์„ธ๊ณ„ ์ตœ๋Œ€์˜ ์ž„์ƒ์—ฐ๊ตฌ ์„œ๋น„์Šค, ์ƒ์—… ์ธํ…”๋ฆฌ์ „์Šค, ๊ทธ๋ฆฌ๊ณ  ํ—ฌ์Šค์ผ€์–ด ๋ถ„์„ ํšŒ์‚ฌ์ธ IQVIA์™€ ํŒŒํŠธ๋„ˆ์‹ญ์„ ๋งบ๊ฒŒ ๋˜์–ด ์ •๋ง ๊ธฐ์ฉ๋‹ˆ๋‹ค.
IQVIA and NVIDIA are going to be working together to take the amazing health data network that IQVIA has built over the last couple of decades. They've got a unique database where they can put that to work and build custom proprietary AI models using NVIDIA's AI foundry services on our DGX Cloud capacity, all the way through to the entire AI life cycle, just like I showed you with TJ, and assemble agentic platforms and workflows to address some of the most critical areas of work. In fact, some of the areas that we're going to work on first is in clinical trials. How can we accelerate the pace of clinical trials? How can we use this to reduce the administrative burden of clinical trials?IQVIA์™€ NVIDIA๋Š” ํ˜‘๋ ฅํ•˜์—ฌ IQVIA๊ฐ€ ์ง€๋‚œ ์ˆ˜์‹ญ ๋…„๊ฐ„ ๊ตฌ์ถ•ํ•ด์˜จ ๋†€๋ผ์šด ํ—ฌ์Šค ๋ฐ์ดํ„ฐ ๋„คํŠธ์›Œํฌ๋ฅผ ํ™œ์šฉํ•  ์˜ˆ์ •์ž…๋‹ˆ๋‹ค. IQVIA๋Š” ๊ณ ์œ ํ•œ ๋ฐ์ดํ„ฐ๋ฒ ์ด์Šค๋ฅผ ๋ณด์œ ํ•˜๊ณ  ์žˆ์œผ๋ฉฐ, ์ด๋ฅผ ํ™œ์šฉํ•˜์—ฌ NVIDIA์˜ DGX Cloud ์šฉ๋Ÿ‰์—์„œ ์ œ๊ณต๋˜๋Š” AI ํŒŒ์šด๋“œ๋ฆฌ ์„œ๋น„์Šค๋ฅผ ์ด์šฉํ•ด ๋งž์ถคํ˜• ๋…์  AI ๋ชจ๋ธ์„ ๊ตฌ์ถ•ํ•  ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์ด๋Š” ์ œ๊ฐ€ TJ์™€ ํ•จ๊ป˜ ๋ณด์—ฌ๋“œ๋ฆฐ ๊ฒƒ์ฒ˜๋Ÿผ ์ „์ฒด AI ๋ผ์ดํ”„์‚ฌ์ดํด์— ๊ฑธ์ณ ์ง„ํ–‰๋˜๋ฉฐ, ๊ฐ€์žฅ ์ค‘์š”ํ•œ ์—…๋ฌด ์˜์—ญ๋“ค์„ ๋‹ค๋ฃจ๊ธฐ ์œ„ํ•œ ์—์ด์ „ํ‹ฑ ํ”Œ๋žซํผ๊ณผ ์›Œํฌํ”Œ๋กœ์šฐ๋ฅผ ๊ตฌ์„ฑํ•˜๊ฒŒ ๋ฉ๋‹ˆ๋‹ค. ์‹ค์ œ๋กœ ์šฐ๋ฆฌ๊ฐ€ ๋จผ์ € ์ž‘์—…ํ•  ์˜์—ญ ์ค‘ ์ผ๋ถ€๋Š” ์ž„์ƒ์‹œํ—˜ ๋ถ„์•ผ์ž…๋‹ˆ๋‹ค. ์–ด๋–ป๊ฒŒ ํ•˜๋ฉด ์ž„์ƒ์‹œํ—˜์˜ ์†๋„๋ฅผ ๊ฐ€์†ํ™”ํ•  ์ˆ˜ ์žˆ์„๊นŒ์š”? ์–ด๋–ป๊ฒŒ ํ•˜๋ฉด ์ด๋ฅผ ํ™œ์šฉํ•˜์—ฌ ์ž„์ƒ์‹œํ—˜์˜ ํ–‰์ •์  ๋ถ€๋‹ด์„ ์ค„์ผ ์ˆ˜ ์žˆ์„๊นŒ์š”?
All of this in order to get medicines to patients faster and really imagine how we're going to see a different clinical trial process as well as the growth and the ability to get medicines to market. This is an incredibly exciting partnership and Bhavik is here. Thank you for flying in. We're delighted to embark on building on top of their IQVIA Connected Intelligence, through all of the NVIDIA AI Enterprise from foundry services, NeMo services, agentic services to deploy agents of all kinds through to their healthcare grade AI solutions and services to bring to their over 10,000 customers across 100 countries, incredible. Okay. Let's switch gears to computer-aided drug discovery.์ด ๋ชจ๋“  ๊ฒƒ์€ ํ™˜์ž๋“ค์—๊ฒŒ ๋” ๋น ๋ฅด๊ฒŒ ์˜์•ฝํ’ˆ์„ ์ „๋‹ฌํ•˜๊ธฐ ์œ„ํ•œ ๊ฒƒ์ด๋ฉฐ, ์ž„์ƒ์‹œํ—˜ ๊ณผ์ •์ด ์–ด๋–ป๊ฒŒ ๋‹ฌ๋ผ์งˆ์ง€, ๊ทธ๋ฆฌ๊ณ  ์˜์•ฝํ’ˆ์„ ์‹œ์žฅ์— ์ถœ์‹œํ•˜๋Š” ์„ฑ์žฅ๊ณผ ๋Šฅ๋ ฅ์„ ์–ด๋–ป๊ฒŒ ๋ณด๊ฒŒ ๋ ์ง€ ์ •๋ง ์ƒ์ƒํ•ด๋ณด์‹œ๊ธฐ ๋ฐ”๋ž๋‹ˆ๋‹ค. ์ด๊ฒƒ์€ ๋งค์šฐ ํฅ๋ฏธ์ง„์ง„ํ•œ ํŒŒํŠธ๋„ˆ์‹ญ์ด๋ฉฐ ๋ฐ”๋น…์ด ์—ฌ๊ธฐ ์™€ ์žˆ์Šต๋‹ˆ๋‹ค. ๋น„ํ–‰๊ธฐ๋กœ ์™€์ฃผ์…”์„œ ๊ฐ์‚ฌํ•ฉ๋‹ˆ๋‹ค. ์šฐ๋ฆฌ๋Š” ๊ทธ๋“ค์˜ IQVIA Connected Intelligence๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ๊ตฌ์ถ•ํ•˜๊ฒŒ ๋˜์–ด ๊ธฐ์ฉ๋‹ˆ๋‹ค. ํŒŒ์šด๋“œ๋ฆฌ ์„œ๋น„์Šค, NeMo ์„œ๋น„์Šค, ๋ชจ๋“  ์ข…๋ฅ˜์˜ ์—์ด์ „ํŠธ๋ฅผ ๋ฐฐํฌํ•˜๋Š” ์—์ด์ „ํ‹ฑ ์„œ๋น„์Šค๋ถ€ํ„ฐ ํ—ฌ์Šค์ผ€์–ด ๋“ฑ๊ธ‰ AI ์†”๋ฃจ์…˜ ๋ฐ ์„œ๋น„์Šค์— ์ด๋ฅด๊ธฐ๊นŒ์ง€ ๋ชจ๋“  NVIDIA AI Enterprise๋ฅผ ํ†ตํ•ด 100๊ฐœ๊ตญ์— ๊ฑธ์นœ 10,000๋ช… ์ด์ƒ์˜ ๊ณ ๊ฐ๋“ค์—๊ฒŒ ์ œ๊ณตํ•  ์˜ˆ์ •์ž…๋‹ˆ๋‹ค. ์ •๋ง ๋†€๋ž์Šต๋‹ˆ๋‹ค. ์ž, ์ด์ œ ์ปดํ“จํ„ฐ ์ง€์› ์‹ ์•ฝ ๊ฐœ๋ฐœ๋กœ ํ™”์ œ๋ฅผ ๋ฐ”๊ฟ”๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค.
We talked last year how we would love to turn the word discovery into design. I think last year's Nobel Prize, congratulations to David Baker, Demis Hassabis and John Jumper of the Nobel Prize in Chemistry for AlphaFold and its profound impact in protein structure. And this is a testament to say how important this capability is going to be for drug discovery and biology understanding in the future. It is on a tear. We continue to see new models of all kinds. Our fantastic partners are building models like Immix, Neuroplexor3, MIT, the Geneo Clinic and Genesis Therapeutics with their Bolt 1, Evolutionary Scale and their ESM3, these are multimodal.์ž‘๋…„์— ์šฐ๋ฆฌ๊ฐ€ '๋ฐœ๊ฒฌ(discovery)'์ด๋ผ๋Š” ๋‹จ์–ด๋ฅผ '์„ค๊ณ„(design)'๋กœ ๋ฐ”๊พธ๊ณ  ์‹ถ๋‹ค๊ณ  ๋ง์”€๋“œ๋ ธ์Šต๋‹ˆ๋‹ค. ์ž‘๋…„ ๋…ธ๋ฒจ์ƒ, AlphaFold์™€ ๋‹จ๋ฐฑ์งˆ ๊ตฌ์กฐ์— ๋ฏธ์นœ ๊นŠ์€ ์˜ํ–ฅ์œผ๋กœ ๋…ธ๋ฒจ ํ™”ํ•™์ƒ์„ ์ˆ˜์ƒํ•œ David Baker, Demis Hassabis, John Jumper์—๊ฒŒ ์ถ•ํ•˜๋ฅผ ๋“œ๋ฆฝ๋‹ˆ๋‹ค. ์ด๋Š” ์ด๋Ÿฌํ•œ ์—ญ๋Ÿ‰์ด ํ–ฅํ›„ ์‹ ์•ฝ ๋ฐœ๊ฒฌ๊ณผ ์ƒ๋ฌผํ•™์  ์ดํ•ด์— ์–ผ๋งˆ๋‚˜ ์ค‘์š”ํ• ์ง€๋ฅผ ๋ณด์—ฌ์ฃผ๋Š” ์ฆ๊ฑฐ์ž…๋‹ˆ๋‹ค. ์ด ๋ถ„์•ผ๋Š” ๊ธ‰์†๋„๋กœ ๋ฐœ์ „ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์šฐ๋ฆฌ๋Š” ๊ณ„์†ํ•ด์„œ ๋ชจ๋“  ์ข…๋ฅ˜์˜ ์ƒˆ๋กœ์šด ๋ชจ๋ธ๋“ค์„ ๋ณด๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์šฐ๋ฆฌ์˜ ํ›Œ๋ฅญํ•œ ํŒŒํŠธ๋„ˆ๋“ค์ด Immix, Neuroplexor3, MIT, Geneo Clinic, ๊ทธ๋ฆฌ๊ณ  Bolt 1์„ ๊ฐœ๋ฐœํ•œ Genesis Therapeutics, ESM3๋ฅผ ๊ฐœ๋ฐœํ•œ Evolutionary Scale๊ณผ ๊ฐ™์€ ๋ชจ๋ธ๋“ค์„ ๊ตฌ์ถ•ํ•˜๊ณ  ์žˆ์œผ๋ฉฐ, ์ด๋“ค์€ ๋ชจ๋‘ ๋ฉ€ํ‹ฐ๋ชจ๋‹ฌ ๋ชจ๋ธ์ž…๋‹ˆ๋‹ค.
They can work with DNA, RNA, proteins, ligands, really understanding much more about their interactions and transforming virtual screening in incredible new ways. The NVIDIA BioNeMo platform is a way of democratizing this. We have an end-to-end platform for, again, the entire life cycle, starting from the training of models. Every single tech bio, biotech, pharma company has wet labs running approximately 24 hours a day, generating very, very rich data. That data needs to be built into intelligence. And those are NVIDIA NIMs and micro services. We've also curated a whole bunch, over a dozen now of micro services that can be used as APIs. They're as simple as an API call.๊ทธ๋“ค์€ DNA, RNA, ๋‹จ๋ฐฑ์งˆ, ๋ฆฌ๊ฐ„๋“œ๋ฅผ ๋‹ค๋ฃฐ ์ˆ˜ ์žˆ์œผ๋ฉฐ, ์ด๋“ค ๊ฐ„์˜ ์ƒํ˜ธ์ž‘์šฉ์„ ํ›จ์”ฌ ๋” ๊นŠ์ด ์ดํ•ดํ•˜๊ณ  ๊ฐ€์ƒ ์Šคํฌ๋ฆฌ๋‹์„ ๋†€๋ผ์šด ์ƒˆ๋กœ์šด ๋ฐฉ์‹์œผ๋กœ ๋ณ€ํ™”์‹œํ‚ฌ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. NVIDIA BioNeMo ํ”Œ๋žซํผ์€ ์ด๋ฅผ ๋ฏผ์ฃผํ™”ํ•˜๋Š” ๋ฐฉ๋ฒ•์ž…๋‹ˆ๋‹ค. ์ €ํฌ๋Š” ๋ชจ๋ธ ํ›ˆ๋ จ๋ถ€ํ„ฐ ์‹œ์ž‘ํ•˜์—ฌ ์ „์ฒด ์ƒ๋ช…์ฃผ๊ธฐ๋ฅผ ์œ„ํ•œ ์—”๋“œํˆฌ์—”๋“œ ํ”Œ๋žซํผ์„ ๋ณด์œ ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ๋ชจ๋“  ํ…Œํฌ๋ฐ”์ด์˜ค, ๋ฐ”์ด์˜คํ…Œํฌ, ์ œ์•ฝํšŒ์‚ฌ๋“ค์€ ํ•˜๋ฃจ ์•ฝ 24์‹œ๊ฐ„ ๋™์•ˆ ์›ป๋žฉ์„ ์šด์˜ํ•˜๋ฉฐ ๋งค์šฐ ํ’๋ถ€ํ•œ ๋ฐ์ดํ„ฐ๋ฅผ ์ƒ์„ฑํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์ด๋Ÿฌํ•œ ๋ฐ์ดํ„ฐ๋Š” ์ธํ…”๋ฆฌ์ „์Šค๋กœ ๊ตฌ์ถ•๋˜์–ด์•ผ ํ•ฉ๋‹ˆ๋‹ค. ๊ทธ๊ฒƒ์ด ๋ฐ”๋กœ NVIDIA NIM๊ณผ ๋งˆ์ดํฌ๋กœ์„œ๋น„์Šค์ž…๋‹ˆ๋‹ค. ๋˜ํ•œ ์ €ํฌ๋Š” API๋กœ ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ๋Š” ๋งˆ์ดํฌ๋กœ์„œ๋น„์Šค๋ฅผ ํ˜„์žฌ 12๊ฐœ ์ด์ƒ ํ๋ ˆ์ด์…˜ํ–ˆ์Šต๋‹ˆ๋‹ค. ์ด๋“ค์€ API ํ˜ธ์ถœ๋งŒํผ ๊ฐ„๋‹จํ•ฉ๋‹ˆ๋‹ค.
Introduce a protein sequence, outcomes a 3D structure. But then the other part of the -- new part of the platform is NVIDIA BioNeMo blueprints. Again, these are reference applications. This is an example of how can you do virtual screening by assembling three different generative AI applications to do full generative AI virtual screening at unprecedented scale. We are introducing here today and it's available on ai.nvidia.com, one of our own foundation models called, GenMol. This is an evolution. We started in 2020, I think it was, with AstraZeneca and our molecular generation, MolMIM last year and next generation here now with GenMol, which is goal-directed molecular generation.๋‹จ๋ฐฑ์งˆ ์„œ์—ด์„ ์ž…๋ ฅํ•˜๋ฉด 3D ๊ตฌ์กฐ๋ฅผ ์ถœ๋ ฅํ•ฉ๋‹ˆ๋‹ค. ๊ทธ๋Ÿฐ๋ฐ ํ”Œ๋žซํผ์˜ ๋˜ ๋‹ค๋ฅธ ์ƒˆ๋กœ์šด ๋ถ€๋ถ„์€ NVIDIA BioNeMo ๋ธ”๋ฃจํ”„๋ฆฐํŠธ์ž…๋‹ˆ๋‹ค. ๋‹ค์‹œ ๋ง์”€๋“œ๋ฆฌ์ง€๋งŒ, ์ด๊ฒƒ๋“ค์€ ์ฐธ์กฐ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์ž…๋‹ˆ๋‹ค. ์ด๋Š” ์„ธ ๊ฐ€์ง€ ์„œ๋กœ ๋‹ค๋ฅธ ์ƒ์„ฑํ˜• AI ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์„ ์กฐํ•ฉํ•˜์—ฌ ์ „๋ก€ ์—†๋Š” ๊ทœ๋ชจ๋กœ ์™„์ „ํ•œ ์ƒ์„ฑํ˜• AI ๊ฐ€์ƒ ์Šคํฌ๋ฆฌ๋‹์„ ์ˆ˜ํ–‰ํ•˜๋Š” ๋ฐฉ๋ฒ•์˜ ์˜ˆ์‹œ์ž…๋‹ˆ๋‹ค. ์˜ค๋Š˜ ์—ฌ๊ธฐ์„œ ์†Œ๊ฐœํ•˜๊ณ  ai.nvidia.com์—์„œ ์ด์šฉ ๊ฐ€๋Šฅํ•œ ๊ฒƒ์€ GenMol์ด๋ผ๊ณ  ๋ถˆ๋ฆฌ๋Š” ์ €ํฌ ์ž์ฒด ํŒŒ์šด๋ฐ์ด์…˜ ๋ชจ๋ธ ์ค‘ ํ•˜๋‚˜์ž…๋‹ˆ๋‹ค. ์ด๋Š” ์ง„ํ™”์˜ ๊ฒฐ๊ณผ๋ฌผ์ž…๋‹ˆ๋‹ค. 2020๋…„๊ฒฝ๋ถ€ํ„ฐ AstraZeneca์™€ ํ•จ๊ป˜ ๋ถ„์ž ์ƒ์„ฑ ์ž‘์—…์„ ์‹œ์ž‘ํ–ˆ๊ณ , ์ž‘๋…„์—๋Š” MolMIM์„, ๊ทธ๋ฆฌ๊ณ  ์ง€๊ธˆ ์—ฌ๊ธฐ์„œ ์ฐจ์„ธ๋Œ€ GenMol์„ ์„ ๋ณด์ด๋Š”๋ฐ, ์ด๋Š” ๋ชฉํ‘œ ์ง€ํ–ฅ์  ๋ถ„์ž ์ƒ์„ฑ ๋ชจ๋ธ์ž…๋‹ˆ๋‹ค.
So this is where you can introduce your own objectives and it can generate molecules. And it generalizes across really critical workflows from hit to lead workflows, as you can see here with generative de novo molecule design and Motif extension upon many more. Our NVIDIA BioNeMo platform continues to expand. And as I said, these Generative AI models have applicability from beginning to end, all the way from target discovery all the way through to clinical trials. And we continue to build these micro services as easy ways to integrate this capability into processes and drug discovery platforms of all kinds, right next to the wet lab serving as your dry lab.๋”ฐ๋ผ์„œ ์—ฌ๊ธฐ์„œ ๊ณ ์œ ํ•œ ๋ชฉํ‘œ๋ฅผ ๋„์ž…ํ•  ์ˆ˜ ์žˆ์œผ๋ฉฐ, ์ด๋ฅผ ํ†ตํ•ด ๋ถ„์ž๋ฅผ ์ƒ์„ฑํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ์ด๋Š” ์—ฌ๊ธฐ์„œ ๋ณด์‹œ๋Š” ๋ฐ”์™€ ๊ฐ™์ด ์ƒ์„ฑํ˜• ๋“œ ๋…ธ๋ณด ๋ถ„์ž ์„ค๊ณ„(generative de novo molecule design)์™€ ๋ชจํ‹ฐํ”„ ํ™•์žฅ(Motif extension) ๋“ฑ์„ ํฌํ•จํ•œ ํžˆํŠธ ํˆฌ ๋ฆฌ๋“œ(hit to lead) ์›Œํฌํ”Œ๋กœ์šฐ์™€ ๊ฐ™์€ ๋งค์šฐ ์ค‘์š”ํ•œ ์›Œํฌํ”Œ๋กœ์šฐ ์ „๋ฐ˜์— ๊ฑธ์ณ ์ผ๋ฐ˜ํ™”๋ฉ๋‹ˆ๋‹ค.

์ €ํฌ NVIDIA BioNeMo ํ”Œ๋žซํผ์€ ์ง€์†์ ์œผ๋กœ ํ™•์žฅ๋˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ์ œ๊ฐ€ ๋ง์”€๋“œ๋ฆฐ ๋ฐ”์™€ ๊ฐ™์ด, ์ด๋Ÿฌํ•œ ์ƒ์„ฑํ˜• AI ๋ชจ๋ธ๋“ค์€ ํƒ€๊ฒŸ ๋ฐœ๊ฒฌ๋ถ€ํ„ฐ ์ž„์ƒ์‹œํ—˜์— ์ด๋ฅด๊ธฐ๊นŒ์ง€ ์ฒ˜์Œ๋ถ€ํ„ฐ ๋๊นŒ์ง€ ์ ์šฉ ๊ฐ€๋Šฅ์„ฑ์„ ๊ฐ€์ง€๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์ €ํฌ๋Š” ์ด๋Ÿฌํ•œ ๋งˆ์ดํฌ๋กœ์„œ๋น„์Šค๋“ค์„ ๋ชจ๋“  ์ข…๋ฅ˜์˜ ํ”„๋กœ์„ธ์Šค์™€ ์‹ ์•ฝ ๊ฐœ๋ฐœ ํ”Œ๋žซํผ์— ์‰ฝ๊ฒŒ ํ†ตํ•ฉํ•  ์ˆ˜ ์žˆ๋Š” ๋ฐฉ๋ฒ•์œผ๋กœ ๊ณ„์† ๊ตฌ์ถ•ํ•˜๊ณ  ์žˆ์œผ๋ฉฐ, ์ด๋Š” ์Šต์‹ ์‹คํ—˜์‹ค(wet lab) ๋ฐ”๋กœ ์˜†์—์„œ ๊ฑด์‹ ์‹คํ—˜์‹ค(dry lab) ์—ญํ• ์„ ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.
And we're building these blueprints, which you can essentially think of as drug discovery assistants. The blueprints, we -- last year in 2024, we introduced the virtual screening blueprint. We're continuing to evolve it, make it smarter and make it faster. And then we, today, are introducing our protein design assistant for protein binder design. So we have small molecule on the left, large molecule on the right, and this is so that we can essentially de novo develop new proteins. Let me show you how it works. [Video Presentation]

Incredible stuff. I'm excited to announce our second announcement.
๊ทธ๋ฆฌ๊ณ  ์ €ํฌ๋Š” ๋ณธ์งˆ์ ์œผ๋กœ ์‹ ์•ฝ ๋ฐœ๊ฒฌ ์–ด์‹œ์Šคํ„ดํŠธ๋ผ๊ณ  ์ƒ๊ฐํ•˜์‹ค ์ˆ˜ ์žˆ๋Š” ์ด๋Ÿฌํ•œ ๋ธ”๋ฃจํ”„๋ฆฐํŠธ๋“ค์„ ๊ตฌ์ถ•ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ๋ธ”๋ฃจํ”„๋ฆฐํŠธ์™€ ๊ด€๋ จํ•ด์„œ๋Š” -- ์ž‘๋…„ 2024๋…„์— ๊ฐ€์ƒ ์Šคํฌ๋ฆฌ๋‹ ๋ธ”๋ฃจํ”„๋ฆฐํŠธ๋ฅผ ๋„์ž…ํ–ˆ์Šต๋‹ˆ๋‹ค. ์ €ํฌ๋Š” ์ด๋ฅผ ์ง€์†์ ์œผ๋กœ ๋ฐœ์ „์‹œ์ผœ ๋”์šฑ ์Šค๋งˆํŠธํ•˜๊ณ  ๋น ๋ฅด๊ฒŒ ๋งŒ๋“ค๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ์˜ค๋Š˜ ์ €ํฌ๋Š” ๋‹จ๋ฐฑ์งˆ ๊ฒฐํ•ฉ์ฒด ์„ค๊ณ„๋ฅผ ์œ„ํ•œ ๋‹จ๋ฐฑ์งˆ ์„ค๊ณ„ ์–ด์‹œ์Šคํ„ดํŠธ๋ฅผ ์†Œ๊ฐœํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์™ผ์ชฝ์—๋Š” ์ €๋ถ„์ž, ์˜ค๋ฅธ์ชฝ์—๋Š” ๊ณ ๋ถ„์ž๊ฐ€ ์žˆ์œผ๋ฉฐ, ์ด๋ฅผ ํ†ตํ•ด ๋ณธ์งˆ์ ์œผ๋กœ ์ƒˆ๋กœ์šด ๋‹จ๋ฐฑ์งˆ์„ ์™„์ „ํžˆ ์ƒˆ๋กญ๊ฒŒ ๊ฐœ๋ฐœํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ์–ด๋–ป๊ฒŒ ์ž‘๋™ํ•˜๋Š”์ง€ ๋ณด์—ฌ๋“œ๋ฆฌ๊ฒ ์Šต๋‹ˆ๋‹ค. [๋น„๋””์˜ค ํ”„๋ ˆ์  ํ…Œ์ด์…˜]

์ •๋ง ๋†€๋ผ์šด ๋‚ด์šฉ์ž…๋‹ˆ๋‹ค. ๋‘ ๋ฒˆ์งธ ๋ฐœํ‘œ๋ฅผ ํ•˜๊ฒŒ ๋˜์–ด ๊ธฐ์ฉ๋‹ˆ๋‹ค.
We are working with the Arc Institute on the development of true foundation models for biology. As you know, the foundation models in language are hitting the trillions of parameters. In biology, we're sitting in the tens of billions, approaching 100 billion parameters. There's a lot of work to do. In fact, the Arc Institute has the charter, the mission of wanting to understand disease at a much deeper level. And where else to do that than in the source code of biology itself and DNA, RNA, and proteins. And so we're working together to build very large-scale foundation models. We're going to -- we're using the NVIDIA DGX Cloud infrastructure, BioNeMo.์ €ํฌ๋Š” Arc Institute์™€ ํ•จ๊ป˜ ์ƒ๋ฌผํ•™์„ ์œ„ํ•œ ์ง„์ •ํ•œ ํŒŒ์šด๋ฐ์ด์…˜ ๋ชจ๋ธ ๊ฐœ๋ฐœ์— ํ˜‘๋ ฅํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์•„์‹œ๋‹ค์‹œํ”ผ, ์–ธ์–ด ๋ถ„์•ผ์˜ ํŒŒ์šด๋ฐ์ด์…˜ ๋ชจ๋ธ๋“ค์€ ์ˆ˜์กฐ ๊ฐœ์˜ ๋งค๊ฐœ๋ณ€์ˆ˜์— ๋„๋‹ฌํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์ƒ๋ฌผํ•™ ๋ถ„์•ผ์—์„œ๋Š” ํ˜„์žฌ ์ˆ˜๋ฐฑ์–ต ๊ฐœ ์ˆ˜์ค€์—์„œ 1,000์–ต ๊ฐœ ๋งค๊ฐœ๋ณ€์ˆ˜์— ๊ทผ์ ‘ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์•„์ง ํ•ด์•ผ ํ•  ์ผ์ด ๋งŽ์Šต๋‹ˆ๋‹ค. ์‹ค์ œ๋กœ Arc Institute๋Š” ์งˆ๋ณ‘์„ ํ›จ์”ฌ ๋” ๊นŠ์€ ์ˆ˜์ค€์—์„œ ์ดํ•ดํ•˜๊ณ ์ž ํ•˜๋Š” ์‚ฌ๋ช…์„ ๊ฐ€์ง€๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ์ƒ๋ฌผํ•™์˜ ์†Œ์Šค ์ฝ”๋“œ ์ž์ฒด์ธ DNA, RNA, ๊ทธ๋ฆฌ๊ณ  ๋‹จ๋ฐฑ์งˆ๋ณด๋‹ค ๋” ์ ํ•ฉํ•œ ๊ณณ์ด ์–ด๋”” ์žˆ๊ฒ ์Šต๋‹ˆ๊นŒ. ๋”ฐ๋ผ์„œ ์ €ํฌ๋Š” ํ•จ๊ป˜ ๋งค์šฐ ๋Œ€๊ทœ๋ชจ์˜ ํŒŒ์šด๋ฐ์ด์…˜ ๋ชจ๋ธ์„ ๊ตฌ์ถ•ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์ €ํฌ๋Š” NVIDIA DGX Cloud ์ธํ”„๋ผ์™€ BioNeMo๋ฅผ ์‚ฌ์šฉํ•  ์˜ˆ์ •์ž…๋‹ˆ๋‹ค.
All of the models that we develop are -- they're going to be built upon last year's Arc Institute's breakthrough with their EVO model, which, again, is a multimodal model for biology. And the Arc Institute is a really unique place. They have really invested in the funding of people versus projects. And I think there's a really profound thing that needs to be considered in science. In order for science to scale, we know we need machine learning. In order to do machine learning, you need to have machines. And so this is really a shift that needs to happen in the scientific community. We're working with Arc to deeply understand that.์ €ํฌ๊ฐ€ ๊ฐœ๋ฐœํ•˜๋Š” ๋ชจ๋“  ๋ชจ๋ธ๋“ค์€ ์ž‘๋…„ Arc Institute์˜ EVO ๋ชจ๋ธ ํ˜์‹ ์„ ๊ธฐ๋ฐ˜์œผ๋กœ ๊ตฌ์ถ•๋  ์˜ˆ์ •์ž…๋‹ˆ๋‹ค. ๋‹ค์‹œ ๋ง์”€๋“œ๋ฆฌ๋ฉด, ์ด๋Š” ์ƒ๋ฌผํ•™์„ ์œ„ํ•œ ๋ฉ€ํ‹ฐ๋ชจ๋‹ฌ ๋ชจ๋ธ์ž…๋‹ˆ๋‹ค. Arc Institute๋Š” ์ •๋ง ๋…ํŠนํ•œ ๊ณณ์ž…๋‹ˆ๋‹ค. ๊ทธ๋“ค์€ ํ”„๋กœ์ ํŠธ๊ฐ€ ์•„๋‹Œ ์ธ์žฌ์— ๋Œ€ํ•œ ํˆฌ์ž์— ์ง„์ •์œผ๋กœ ์ง‘์ค‘ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ๊ณผํ•™ ๋ถ„์•ผ์—์„œ ์ •๋ง ์‹ฌ๋„ ์žˆ๊ฒŒ ๊ณ ๋ คํ•ด์•ผ ํ•  ์ค‘์š”ํ•œ ์ ์ด ์žˆ๋‹ค๊ณ  ์ƒ๊ฐํ•ฉ๋‹ˆ๋‹ค. ๊ณผํ•™์ด ํ™•์žฅ๋˜๊ธฐ ์œ„ํ•ด์„œ๋Š” ๋จธ์‹ ๋Ÿฌ๋‹์ด ํ•„์š”ํ•˜๋‹ค๋Š” ๊ฒƒ์„ ์•Œ๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ๋จธ์‹ ๋Ÿฌ๋‹์„ ์ˆ˜ํ–‰ํ•˜๊ธฐ ์œ„ํ•ด์„œ๋Š” ๊ธฐ๊ณ„๊ฐ€ ์žˆ์–ด์•ผ ํ•ฉ๋‹ˆ๋‹ค. ๋”ฐ๋ผ์„œ ์ด๋Š” ๊ณผํ•™๊ณ„์—์„œ ์ผ์–ด๋‚˜์•ผ ํ•˜๋Š” ์ง„์ •ํ•œ ํŒจ๋Ÿฌ๋‹ค์ž„ ์ „ํ™˜์ž…๋‹ˆ๋‹ค. ์ €ํฌ๋Š” Arc์™€ ํ˜‘๋ ฅํ•˜์—ฌ ์ด๋ฅผ ๊นŠ์ด ์ดํ•ดํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.
Everything that we build, we will be putting back into the open source into BioNeMo. And you're going to see a lot of that work come to fruition over the next several months and throughout 2025. So keep a lookout for future announcements with the Arc Institute and NVIDIA. As I say, the source code of humans and biology is in genomics. We've been working with amazing partners in the genomics industry, helping to build these regional population-specific genomics data sets. And the world now is really rushing to determine their sovereign AI strategies. We've really helped to lay the foundation, giving the capabilities to generate this data.์šฐ๋ฆฌ๊ฐ€ ๊ตฌ์ถ•ํ•˜๋Š” ๋ชจ๋“  ๊ฒƒ์€ BioNeMo๋ฅผ ํ†ตํ•ด ์˜คํ”ˆ ์†Œ์Šค๋กœ ๋‹ค์‹œ ์ œ๊ณต๋  ๊ฒƒ์ž…๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ์•ž์œผ๋กœ ๋ช‡ ๋‹ฌ๊ฐ„ ๊ทธ๋ฆฌ๊ณ  2025๋…„ ๋‚ด๋‚ด ์ด๋Ÿฌํ•œ ์ž‘์—…์˜ ๋งŽ์€ ์„ฑ๊ณผ๋ฅผ ๋ณด์‹œ๊ฒŒ ๋  ๊ฒƒ์ž…๋‹ˆ๋‹ค. Arc Institute์™€ NVIDIA์˜ ํ–ฅํ›„ ๋ฐœํ‘œ๋ฅผ ์ฃผ๋ชฉํ•ด ์ฃผ์‹œ๊ธฐ ๋ฐ”๋ž๋‹ˆ๋‹ค. ์ œ๊ฐ€ ๋ง์”€๋“œ๋ ธ๋“ฏ์ด, ์ธ๊ฐ„๊ณผ ์ƒ๋ฌผํ•™์˜ ์†Œ์Šค ์ฝ”๋“œ๋Š” ์œ ์ „์ฒดํ•™์— ์žˆ์Šต๋‹ˆ๋‹ค. ์šฐ๋ฆฌ๋Š” ์œ ์ „์ฒดํ•™ ์—…๊ณ„์˜ ๋†€๋ผ์šด ํŒŒํŠธ๋„ˆ๋“ค๊ณผ ํ˜‘๋ ฅํ•˜์—ฌ ์ด๋Ÿฌํ•œ ์ง€์—ญ๋ณ„, ์ธ๊ตฌ๋ณ„ ํŠนํ™” ์œ ์ „์ฒด ๋ฐ์ดํ„ฐ์…‹ ๊ตฌ์ถ•์„ ์ง€์›ํ•ด์™”์Šต๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ํ˜„์žฌ ์ „ ์„ธ๊ณ„๋Š” ์ž๊ตญ์˜ ์ฃผ๊ถŒ AI ์ „๋žต์„ ์ˆ˜๋ฆฝํ•˜๊ธฐ ์œ„ํ•ด ์ •๋ง ์„œ๋‘๋ฅด๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์šฐ๋ฆฌ๋Š” ์ด๋Ÿฌํ•œ ๋ฐ์ดํ„ฐ๋ฅผ ์ƒ์„ฑํ•  ์ˆ˜ ์žˆ๋Š” ์—ญ๋Ÿ‰์„ ์ œ๊ณตํ•จ์œผ๋กœ์จ ๊ธฐ๋ฐ˜์„ ๋งˆ๋ จํ•˜๋Š” ๋ฐ ์‹ค์งˆ์ ์œผ๋กœ ๋„์›€์„ ์ฃผ์—ˆ์Šต๋‹ˆ๋‹ค.
But we need to, by working with the Arc Institute and building new foundation models, we need to give the world not only the ability to generate the data but to derive insights from this data. And so I'm extremely pleased to announce our third announcement for the day, which is our partnership with Illumina. Illumina is a world leader in genomics. We're going to be working together by combining the Illumina sequencing platforms, NVIDIA's NVIDIA Clara tool set, everything from our multiomics BioNeMo for foundation models in biology, RAPIDS for data science acceleration, MONAI for cell imaging and other workflows into the Illumina connected analytics.ํ•˜์ง€๋งŒ ์šฐ๋ฆฌ๋Š” Arc Institute์™€์˜ ํ˜‘๋ ฅ์„ ํ†ตํ•ด ์ƒˆ๋กœ์šด ํŒŒ์šด๋ฐ์ด์…˜ ๋ชจ๋ธ์„ ๊ตฌ์ถ•ํ•จ์œผ๋กœ์จ, ์„ธ์ƒ์— ๋ฐ์ดํ„ฐ๋ฅผ ์ƒ์„ฑํ•˜๋Š” ๋Šฅ๋ ฅ๋ฟ๋งŒ ์•„๋‹ˆ๋ผ ์ด ๋ฐ์ดํ„ฐ๋กœ๋ถ€ํ„ฐ ์ธ์‚ฌ์ดํŠธ๋ฅผ ๋„์ถœํ•˜๋Š” ๋Šฅ๋ ฅ์„ ์ œ๊ณตํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค. ๊ทธ๋ž˜์„œ ์˜ค๋Š˜ ์„ธ ๋ฒˆ์งธ ๋ฐœํ‘œ๋ฅผ ํ•˜๊ฒŒ ๋˜์–ด ๋งค์šฐ ๊ธฐ์ฉ๋‹ˆ๋‹ค. ๋ฐ”๋กœ Illumina์™€์˜ ํŒŒํŠธ๋„ˆ์‹ญ์ž…๋‹ˆ๋‹ค. Illumina๋Š” ์œ ์ „์ฒดํ•™ ๋ถ„์•ผ์˜ ์„ธ๊ณ„์  ์„ ๋„๊ธฐ์—…์ž…๋‹ˆ๋‹ค. ์šฐ๋ฆฌ๋Š” Illumina ์‹œํ€€์‹ฑ ํ”Œ๋žซํผ๊ณผ NVIDIA์˜ NVIDIA Clara ํˆด์…‹์„ ๊ฒฐํ•ฉํ•˜์—ฌ ํ˜‘๋ ฅํ•  ์˜ˆ์ •์ž…๋‹ˆ๋‹ค. ์ƒ๋ฌผํ•™ ๋ถ„์•ผ ํŒŒ์šด๋ฐ์ด์…˜ ๋ชจ๋ธ์„ ์œ„ํ•œ ๋ฉ€ํ‹ฐ์˜ค๋ฏน์Šค BioNeMo๋ถ€ํ„ฐ ๋ฐ์ดํ„ฐ ์‚ฌ์ด์–ธ์Šค ๊ฐ€์†ํ™”๋ฅผ ์œ„ํ•œ RAPIDS, ์„ธํฌ ์ด๋ฏธ์ง•์„ ์œ„ํ•œ MONAI ๋ฐ ๊ธฐํƒ€ ์›Œํฌํ”Œ๋กœ์šฐ๊นŒ์ง€ ๋ชจ๋“  ๊ฒƒ์„ Illumina ์ปค๋„ฅํ‹ฐ๋“œ ์• ๋„๋ฆฌํ‹ฑ์Šค์— ํ†ตํ•ฉํ•  ๊ฒƒ์ž…๋‹ˆ๋‹ค.
This partnership is going to be fantastic. One of the first things we're going to do is we're going to take Dragon, the industry standard of secondary analysis, bring it to the GPU. We're going to do that so we can expand the market so that Dragon can be deployed wherever there are NVIDIA GPUs in every public cloud, in every sovereign nation, in every data center and expand the market.์ด ํŒŒํŠธ๋„ˆ์‹ญ์€ ์ •๋ง ํ™˜์ƒ์ ์ผ ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์šฐ๋ฆฌ๊ฐ€ ๊ฐ€์žฅ ๋จผ์ € ํ•  ์ผ ์ค‘ ํ•˜๋‚˜๋Š” 2์ฐจ ๋ถ„์„์˜ ์—…๊ณ„ ํ‘œ์ค€์ธ Dragon์„ GPU๋กœ ๊ฐ€์ ธ์˜ค๋Š” ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์ด๋ฅผ ํ†ตํ•ด ์‹œ์žฅ์„ ํ™•์žฅํ•˜์—ฌ Dragon์ด ๋ชจ๋“  ํผ๋ธ”๋ฆญ ํด๋ผ์šฐ๋“œ, ๋ชจ๋“  ์ฃผ๊ถŒ ๊ตญ๊ฐ€, ๋ชจ๋“  ๋ฐ์ดํ„ฐ ์„ผํ„ฐ์—์„œ NVIDIA GPU๊ฐ€ ์žˆ๋Š” ๊ณณ์ด๋ผ๋ฉด ์–ด๋””๋“  ๋ฐฐํฌ๋  ์ˆ˜ ์žˆ๋„๋ก ํ•˜๊ณ  ์‹œ์žฅ์„ ํ™•์žฅํ•  ๊ฒƒ์ž…๋‹ˆ๋‹ค.
And then we're going to work together to build new applications in the Illumina connected analytics, bringing genomics foundation models, bringing single-cell applications that are essentially real time at very, very large scale and expand the genomics accessibility in utility into new markets like drug discovery. We still have so much to learn about the foundational building blocks of biology.๊ทธ๋ฆฌ๊ณ  ์šฐ๋ฆฌ๋Š” ํ•จ๊ป˜ ํ˜‘๋ ฅํ•˜์—ฌ Illumina ์ปค๋„ฅํ‹ฐ๋“œ ์• ๋„๋ฆฌํ‹ฑ์Šค์—์„œ ์ƒˆ๋กœ์šด ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์„ ๊ตฌ์ถ•ํ•  ์˜ˆ์ •์ž…๋‹ˆ๋‹ค. ์œ ์ „์ฒดํ•™ ํŒŒ์šด๋ฐ์ด์…˜ ๋ชจ๋ธ์„ ๋„์ž…ํ•˜๊ณ , ๋งค์šฐ ๋Œ€๊ทœ๋ชจ์—์„œ ๋ณธ์งˆ์ ์œผ๋กœ ์‹ค์‹œ๊ฐ„์œผ๋กœ ์ž‘๋™ํ•˜๋Š” ๋‹จ์ผ์„ธํฌ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์„ ๊ตฌ์ถ•ํ•˜๋ฉฐ, ์‹ ์•ฝ ๊ฐœ๋ฐœ๊ณผ ๊ฐ™์€ ์ƒˆ๋กœ์šด ์‹œ์žฅ์œผ๋กœ ์œ ์ „์ฒดํ•™์˜ ์ ‘๊ทผ์„ฑ๊ณผ ์œ ์šฉ์„ฑ์„ ํ™•์žฅํ•ด ๋‚˜๊ฐˆ ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์šฐ๋ฆฌ๋Š” ์—ฌ์ „ํžˆ ์ƒ๋ฌผํ•™์˜ ๊ธฐ์ดˆ ๊ตฌ์„ฑ ์š”์†Œ์— ๋Œ€ํ•ด ๋ฐฐ์›Œ์•ผ ํ•  ๊ฒƒ์ด ๋„ˆ๋ฌด๋‚˜ ๋งŽ์Šต๋‹ˆ๋‹ค.
It's in the amazing data that Illumina is generating, and now we're going to be working together to build those tools, build those capabilities and expand the genomics market into the next generation, into Generative AI genomics, and all the way to making it more accessible to the research community, to the biopharma and biotech community, and even expanding it into clinical and research clinical applications. Super excited about this partnership and we will be announcing more opportunities here as the time goes on. Okay. Last chapter. Here we go. Every sensor, every patient room, every hospital will integrate physical AI. It's a new concept.์ผ๋ฃจ๋ฏธ๋‚˜๊ฐ€ ์ƒ์„ฑํ•˜๊ณ  ์žˆ๋Š” ๋†€๋ผ์šด ๋ฐ์ดํ„ฐ์— ์žˆ์œผ๋ฉฐ, ์ด์ œ ์šฐ๋ฆฌ๋Š” ํ•จ๊ป˜ ํ˜‘๋ ฅํ•˜์—ฌ ์ด๋Ÿฌํ•œ ๋„๊ตฌ๋“ค์„ ๊ตฌ์ถ•ํ•˜๊ณ , ์ด๋Ÿฌํ•œ ์—ญ๋Ÿ‰๋“ค์„ ๊ตฌ์ถ•ํ•˜๋ฉฐ, ์œ ์ „์ฒดํ•™ ์‹œ์žฅ์„ ์ฐจ์„ธ๋Œ€๋กœ, ์ƒ์„ฑํ˜• AI ์œ ์ „์ฒดํ•™์œผ๋กœ ํ™•์žฅํ•˜๊ณ , ์—ฐ๊ตฌ ์ปค๋ฎค๋‹ˆํ‹ฐ, ๋ฐ”์ด์˜คํŒŒ๋งˆ ๋ฐ ๋ฐ”์ด์˜คํ…Œํฌ ์ปค๋ฎค๋‹ˆํ‹ฐ์— ๋”์šฑ ์ ‘๊ทผ ๊ฐ€๋Šฅํ•˜๊ฒŒ ๋งŒ๋“ค๊ณ , ์‹ฌ์ง€์–ด ์ž„์ƒ ๋ฐ ์—ฐ๊ตฌ ์ž„์ƒ ์‘์šฉ ๋ถ„์•ผ๋กœ๊นŒ์ง€ ํ™•์žฅํ•  ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์ด๋ฒˆ ํŒŒํŠธ๋„ˆ์‹ญ์— ๋Œ€ํ•ด ๋งค์šฐ ๊ธฐ๋Œ€ํ•˜๊ณ  ์žˆ์œผ๋ฉฐ, ์‹œ๊ฐ„์ด ์ง€๋‚˜๋ฉด์„œ ๋” ๋งŽ์€ ๊ธฐํšŒ๋“ค์„ ๋ฐœํ‘œํ•  ์˜ˆ์ •์ž…๋‹ˆ๋‹ค.

์ข‹์Šต๋‹ˆ๋‹ค. ๋งˆ์ง€๋ง‰ ์žฅ์ž…๋‹ˆ๋‹ค. ์‹œ์ž‘ํ•˜๊ฒ ์Šต๋‹ˆ๋‹ค. ๋ชจ๋“  ์„ผ์„œ, ๋ชจ๋“  ํ™˜์ž์‹ค, ๋ชจ๋“  ๋ณ‘์›์ด ๋ฌผ๋ฆฌ์  AI๋ฅผ ํ†ตํ•ฉํ•˜๊ฒŒ ๋  ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์ด๋Š” ์ƒˆ๋กœ์šด ๊ฐœ๋…์ž…๋‹ˆ๋‹ค.
But the simple way to think about physical AI is that it understands the physical world. It understands the environment. It understands that you're taking this picture right now, and it thinks, I made a good slide. But the real opportunity here is healthcare delivery is a very labor-intensive place. And we have thousands of hospitals around the world. We have hundreds, hundreds of thousands, excuse me, we have hundreds of thousands of operating rooms. We have millions of patient rooms and billions of devices. And I said, each one of them is going to be robotic. We have been building the three computers to realize physical AI.ํ•˜์ง€๋งŒ ๋ฌผ๋ฆฌ์  AI์— ๋Œ€ํ•ด ๊ฐ„๋‹จํžˆ ์ƒ๊ฐํ•ด๋ณด๋ฉด, ์ด๋Š” ๋ฌผ๋ฆฌ์  ์„ธ๊ณ„๋ฅผ ์ดํ•ดํ•œ๋‹ค๋Š” ๊ฒƒ์ž…๋‹ˆ๋‹ค. ํ™˜๊ฒฝ์„ ์ดํ•ดํ•ฉ๋‹ˆ๋‹ค. ์ง€๊ธˆ ์—ฌ๋Ÿฌ๋ถ„์ด ์ด ์‚ฌ์ง„์„ ์ฐ๊ณ  ์žˆ๋‹ค๋Š” ๊ฒƒ์„ ์ดํ•ดํ•˜๊ณ , '์ข‹์€ ์Šฌ๋ผ์ด๋“œ๋ฅผ ๋งŒ๋“ค์—ˆ๊ตฌ๋‚˜'๋ผ๊ณ  ์ƒ๊ฐํ•ฉ๋‹ˆ๋‹ค. ํ•˜์ง€๋งŒ ์—ฌ๊ธฐ์„œ ์ง„์ •ํ•œ ๊ธฐํšŒ๋Š” ์˜๋ฃŒ ์„œ๋น„์Šค ์ œ๊ณต์ด ๋งค์šฐ ๋…ธ๋™์ง‘์•ฝ์ ์ธ ๋ถ„์•ผ๋ผ๋Š” ์ ์ž…๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ์šฐ๋ฆฌ๋Š” ์ „ ์„ธ๊ณ„์— ์ˆ˜์ฒœ ๊ฐœ์˜ ๋ณ‘์›์„ ๋ณด์œ ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์ˆ˜์‹ญ๋งŒ ๊ฐœ์˜... ์ฃ„์†กํ•ฉ๋‹ˆ๋‹ค, ์ˆ˜์‹ญ๋งŒ ๊ฐœ์˜ ์ˆ˜์ˆ ์‹ค์„ ๋ณด์œ ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์ˆ˜๋ฐฑ๋งŒ ๊ฐœ์˜ ํ™˜์ž์‹ค๊ณผ ์ˆ˜์‹ญ์–ต ๊ฐœ์˜ ๊ธฐ๊ธฐ๋“ค์„ ๋ณด์œ ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ์ œ๊ฐ€ ๋ง์”€๋“œ๋ฆฐ ๋ฐ”์™€ ๊ฐ™์ด, ์ด๋“ค ๊ฐ๊ฐ์ด ๋กœ๋ด‡ํ™”๋  ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์šฐ๋ฆฌ๋Š” ๋ฌผ๋ฆฌ์  AI๋ฅผ ๊ตฌํ˜„ํ•˜๊ธฐ ์œ„ํ•ด ์„ธ ๊ฐ€์ง€ ์ปดํ“จํ„ฐ๋ฅผ ๊ตฌ์ถ•ํ•ด์™”์Šต๋‹ˆ๋‹ค.
The first computer that you well know us for is the computer that runs the real-time AI. Think about what we've done for self-driving cars, the car that lives in the computer, or think about the embedded products that lives inside the surgical robot. That's the robot brain. That's the brain in the robot. And the other computer, you also know us for, that's to develop the AI, NVIDIA DGX for NVIDIA AI. But we've been missing the third part of the trinity, and that is Omniverse for developing virtual worlds that are grounded in physics.์ €ํฌ๊ฐ€ ์ž˜ ์•Œ๋ ค์ง„ ์ฒซ ๋ฒˆ์งธ ์ปดํ“จํ„ฐ๋Š” ์‹ค์‹œ๊ฐ„ AI๋ฅผ ๊ตฌ๋™ํ•˜๋Š” ์ปดํ“จํ„ฐ์ž…๋‹ˆ๋‹ค. ์ž์œจ์ฃผํ–‰์ฐจ๋ฅผ ์œ„ํ•ด ์ €ํฌ๊ฐ€ ํ•ด์˜จ ์ผ์„ ์ƒ๊ฐํ•ด๋ณด์‹œ๋ฉด, ์ปดํ“จํ„ฐ ์•ˆ์— ์‚ด์•„์žˆ๋Š” ์ž๋™์ฐจ, ๋˜๋Š” ์ˆ˜์ˆ  ๋กœ๋ด‡ ๋‚ด๋ถ€์— ๋‚ด์žฅ๋œ ์ž„๋ฒ ๋””๋“œ ์ œํ’ˆ๋“ค์„ ๋– ์˜ฌ๋ ค๋ณด์‹ญ์‹œ์˜ค. ๊ทธ๊ฒƒ์ด ๋ฐ”๋กœ ๋กœ๋ด‡์˜ ๋‘๋‡Œ์ž…๋‹ˆ๋‹ค. ๋กœ๋ด‡ ์•ˆ์˜ ๋‡Œ์ธ ๊ฒƒ์ด์ฃ . ๊ทธ๋ฆฌ๊ณ  ์ €ํฌ๊ฐ€ ์•Œ๋ ค์ง„ ๋˜ ๋‹ค๋ฅธ ์ปดํ“จํ„ฐ๋Š” AI๋ฅผ ๊ฐœ๋ฐœํ•˜๋Š” ๊ฒƒ์œผ๋กœ, NVIDIA AI๋ฅผ ์œ„ํ•œ NVIDIA DGX์ž…๋‹ˆ๋‹ค. ํ•˜์ง€๋งŒ ์ €ํฌ๋Š” ์‚ผ์œ„์ผ์ฒด์˜ ์„ธ ๋ฒˆ์งธ ๋ถ€๋ถ„์„ ๋†“์น˜๊ณ  ์žˆ์—ˆ๋Š”๋ฐ, ๊ทธ๊ฒƒ์ด ๋ฐ”๋กœ ๋ฌผ๋ฆฌํ•™์— ๊ธฐ๋ฐ˜ํ•œ ๊ฐ€์ƒ ์„ธ๊ณ„๋ฅผ ๊ฐœ๋ฐœํ•˜๋Š” Omniverse์ž…๋‹ˆ๋‹ค.
They understand the laws of physics and they can synthetically generate virtual worlds so that the robot itself does not need to learn solely in a physical real-world environment. It is just too expensive, too costly, too dangerous. And so Omniverse, and I'm going to show you a little bit more about NVIDIA Cosmos, that was just announced at CES, is going to create that third trinity computer to allow for physical AI and the rapid capabilities of robotics in the future. We've been building NVIDIA Holoscan for many, many years now. This is our healthcare AI robot platform. In fact, Medtronic, Moon Surgical, our FDA-approved devices built on this and in market today.๊ทธ๋“ค์€ ๋ฌผ๋ฆฌ ๋ฒ•์น™์„ ์ดํ•ดํ•˜๊ณ  ๊ฐ€์ƒ ์„ธ๊ณ„๋ฅผ ํ•ฉ์„ฑ์ ์œผ๋กœ ์ƒ์„ฑํ•  ์ˆ˜ ์žˆ์–ด์„œ, ๋กœ๋ด‡ ์ž์ฒด๊ฐ€ ๋ฌผ๋ฆฌ์ ์ธ ์‹ค์ œ ํ™˜๊ฒฝ์—์„œ๋งŒ ํ•™์Šตํ•  ํ•„์š”๊ฐ€ ์—†์Šต๋‹ˆ๋‹ค. ๊ทธ๋Ÿฐ ๋ฐฉ์‹์€ ๋„ˆ๋ฌด ๋น„์‹ธ๊ณ , ๋น„์šฉ์ด ๋งŽ์ด ๋“ค๋ฉฐ, ์œ„ํ—˜ํ•˜๊ธฐ ๋•Œ๋ฌธ์ž…๋‹ˆ๋‹ค. ๋”ฐ๋ผ์„œ Omniverse์™€ - CES์—์„œ ๋ฐฉ๊ธˆ ๋ฐœํ‘œ๋œ NVIDIA Cosmos์— ๋Œ€ํ•ด ์กฐ๊ธˆ ๋” ๋ณด์—ฌ๋“œ๋ฆฌ๊ฒ ์Šต๋‹ˆ๋‹ค๋งŒ - ์ด๊ฒƒ์ด ๋ฌผ๋ฆฌ์  AI์™€ ๋ฏธ๋ž˜ ๋กœ๋ณดํ‹ฑ์Šค์˜ ์‹ ์†ํ•œ ์—ญ๋Ÿ‰์„ ๊ฐ€๋Šฅํ•˜๊ฒŒ ํ•˜๋Š” ์„ธ ๋ฒˆ์งธ ํŠธ๋ฆฌ๋‹ˆํ‹ฐ ์ปดํ“จํ„ฐ๋ฅผ ๋งŒ๋“ค์–ด๋‚ผ ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์ €ํฌ๋Š” ์ˆ˜๋…„๊ฐ„ NVIDIA Holoscan์„ ๊ตฌ์ถ•ํ•ด์™”์Šต๋‹ˆ๋‹ค. ์ด๊ฒƒ์€ ์ €ํฌ์˜ ํ—ฌ์Šค์ผ€์–ด AI ๋กœ๋ด‡ ํ”Œ๋žซํผ์ž…๋‹ˆ๋‹ค. ์‹ค์ œ๋กœ Medtronic, Moon Surgical ๋“ฑ ์ด ํ”Œ๋žซํผ์„ ๊ธฐ๋ฐ˜์œผ๋กœ ๊ตฌ์ถ•๋œ FDA ์Šน์ธ ๊ธฐ๊ธฐ๋“ค์ด ํ˜„์žฌ ์‹œ์žฅ์— ์ถœ์‹œ๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค.
It really applies -- it's a universal computing platform. It's sensor-agnostic. It's the real-time processing pipeline to go from sensor in to insights out. And we've built the full stack so that now the industry can just focus on the applications and all of that computing platform has already been built. We need these devices to understand the physical world. So you've heard about language models but now we have vision language models. NVIDIA Cosmos NeMoTron, the family of models that was just announced last week has that visual understanding where you can watch a video, ask it a question, it can tell you what's going on in this video. This gives AI both temporal and spatial awareness.์ด๊ฒƒ์€ ์ •๋ง๋กœ ์ ์šฉ๋ฉ๋‹ˆ๋‹ค -- ๋ฒ”์šฉ ์ปดํ“จํŒ… ํ”Œ๋žซํผ์ž…๋‹ˆ๋‹ค. ์„ผ์„œ์— ๊ตฌ์• ๋ฐ›์ง€ ์•Š์Šต๋‹ˆ๋‹ค. ์„ผ์„œ ์ž…๋ ฅ์—์„œ ์ธ์‚ฌ์ดํŠธ ์ถœ๋ ฅ๊นŒ์ง€์˜ ์‹ค์‹œ๊ฐ„ ์ฒ˜๋ฆฌ ํŒŒ์ดํ”„๋ผ์ธ์ž…๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ์šฐ๋ฆฌ๋Š” ํ’€์Šคํƒ์„ ๊ตฌ์ถ•ํ–ˆ๊ธฐ ๋•Œ๋ฌธ์— ์ด์ œ ์—…๊ณ„๋Š” ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์—๋งŒ ์ง‘์ค‘ํ•˜๋ฉด ๋˜๊ณ , ๋ชจ๋“  ์ปดํ“จํŒ… ํ”Œ๋žซํผ์€ ์ด๋ฏธ ๊ตฌ์ถ•๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค. ์šฐ๋ฆฌ๋Š” ์ด๋Ÿฌํ•œ ๋””๋ฐ”์ด์Šค๋“ค์ด ๋ฌผ๋ฆฌ์  ์„ธ๊ณ„๋ฅผ ์ดํ•ดํ•  ์ˆ˜ ์žˆ๋„๋ก ํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค. ์–ธ์–ด ๋ชจ๋ธ์— ๋Œ€ํ•ด ๋“ค์–ด๋ณด์…จ๊ฒ ์ง€๋งŒ, ์ด์ œ ์šฐ๋ฆฌ์—๊ฒŒ๋Š” ๋น„์ „ ์–ธ์–ด ๋ชจ๋ธ์ด ์žˆ์Šต๋‹ˆ๋‹ค. ์ง€๋‚œ์ฃผ์— ๋ฐœํ‘œ๋œ NVIDIA Cosmos NeMoTron ๋ชจ๋ธ ํŒจ๋ฐ€๋ฆฌ๋Š” ๋น„๋””์˜ค๋ฅผ ์‹œ์ฒญํ•˜๊ณ , ์งˆ๋ฌธ์„ ๋ฐ›์œผ๋ฉด ํ•ด๋‹น ๋น„๋””์˜ค์—์„œ ๋ฌด์Šจ ์ผ์ด ์ผ์–ด๋‚˜๊ณ  ์žˆ๋Š”์ง€ ์•Œ๋ ค์ค„ ์ˆ˜ ์žˆ๋Š” ์‹œ๊ฐ์  ์ดํ•ด ๋Šฅ๋ ฅ์„ ๊ฐ–์ถ”๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์ด๊ฒƒ์€ AI์—๊ฒŒ ์‹œ๊ฐ„์ , ๊ณต๊ฐ„์  ์ธ์‹ ๋Šฅ๋ ฅ์„ ๋ชจ๋‘ ์ œ๊ณตํ•ฉ๋‹ˆ๋‹ค.
This has just come about in the last six or nine months. This is a necessary component but not the only necessary component. The other part of achieving physical AI is how can you create physics-aware virtual worlds. And NVIDIA Cosmos is a family of what we call world foundation models. In fact, it won number one overall at CES just last week because it's that big of a breakthrough. Imagine you have the CAD design of a robot. You can now introduce it with real physics data. Here, we're introducing it with CT scans and other type of multimodal data.์ด๊ฒƒ์€ ์ง€๋‚œ 6๊ฐœ์›”์—์„œ 9๊ฐœ์›” ์‚ฌ์ด์— ๋‚˜ํƒ€๋‚œ ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์ด๋Š” ํ•„์š”ํ•œ ๊ตฌ์„ฑ ์š”์†Œ์ด์ง€๋งŒ ์œ ์ผํ•œ ํ•„์ˆ˜ ๊ตฌ์„ฑ ์š”์†Œ๋Š” ์•„๋‹™๋‹ˆ๋‹ค. ๋ฌผ๋ฆฌ์  AI๋ฅผ ๋‹ฌ์„ฑํ•˜๋Š” ๋˜ ๋‹ค๋ฅธ ๋ถ€๋ถ„์€ ๋ฌผ๋ฆฌํ•™์„ ์ธ์‹ํ•˜๋Š” ๊ฐ€์ƒ ์„ธ๊ณ„๋ฅผ ์–ด๋–ป๊ฒŒ ๋งŒ๋“ค ์ˆ˜ ์žˆ๋А๋ƒ๋Š” ๊ฒƒ์ž…๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  NVIDIA Cosmos๋Š” ์šฐ๋ฆฌ๊ฐ€ ์„ธ๊ณ„ ํŒŒ์šด๋ฐ์ด์…˜ ๋ชจ๋ธ์ด๋ผ๊ณ  ๋ถ€๋ฅด๋Š” ์ œํ’ˆ๊ตฐ์ž…๋‹ˆ๋‹ค. ์‹ค์ œ๋กœ ์ง€๋‚œ์ฃผ CES์—์„œ ์ข…ํ•ฉ 1์œ„๋ฅผ ์ฐจ์ง€ํ–ˆ๋Š”๋ฐ, ๊ทธ๋งŒํผ ํฐ ํ˜์‹ ์ด๊ธฐ ๋•Œ๋ฌธ์ž…๋‹ˆ๋‹ค. ๋กœ๋ด‡์˜ CAD ์„ค๊ณ„๋„๊ฐ€ ์žˆ๋‹ค๊ณ  ์ƒ์ƒํ•ด๋ณด์„ธ์š”. ์ด์ œ ์‹ค์ œ ๋ฌผ๋ฆฌํ•™ ๋ฐ์ดํ„ฐ์™€ ํ•จ๊ป˜ ๋„์ž…ํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ์—ฌ๊ธฐ์„œ๋Š” CT ์Šค์บ”๊ณผ ๊ธฐํƒ€ ์œ ํ˜•์˜ ๋ฉ€ํ‹ฐ๋ชจ๋‹ฌ ๋ฐ์ดํ„ฐ์™€ ํ•จ๊ป˜ ๋„์ž…ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.
These models can generate physically aware scenarios, for which we can change the camera view but it has object permanence so it knows where things are. And it can train robots in essentially infinite scenarios, which is the only way we're going to have safe and effective robots in the healthcare industry. And so we're building towards this, and it's coming along very, very quickly now. So here's a couple of industry firsts. You see amazing applications popping up out of everywhere. We're really leveraging and benefiting from this three computer trinity, Syncron, incredible.์ด๋Ÿฌํ•œ ๋ชจ๋ธ๋“ค์€ ๋ฌผ๋ฆฌ์ ์œผ๋กœ ์ธ์‹ ๊ฐ€๋Šฅํ•œ ์‹œ๋‚˜๋ฆฌ์˜ค๋ฅผ ์ƒ์„ฑํ•  ์ˆ˜ ์žˆ์œผ๋ฉฐ, ์นด๋ฉ”๋ผ ์‹œ์ ์„ ๋ณ€๊ฒฝํ•  ์ˆ˜ ์žˆ์ง€๋งŒ ๊ฐ์ฒด ์˜์†์„ฑ์„ ๊ฐ€์ง€๊ณ  ์žˆ์–ด ์‚ฌ๋ฌผ์ด ์–ด๋””์— ์žˆ๋Š”์ง€ ์•Œ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ๋ณธ์งˆ์ ์œผ๋กœ ๋ฌดํ•œํ•œ ์‹œ๋‚˜๋ฆฌ์˜ค์—์„œ ๋กœ๋ด‡์„ ํ›ˆ๋ จ์‹œํ‚ฌ ์ˆ˜ ์žˆ๋Š”๋ฐ, ์ด๋Š” ํ—ฌ์Šค์ผ€์–ด ์‚ฐ์—…์—์„œ ์•ˆ์ „ํ•˜๊ณ  ํšจ๊ณผ์ ์ธ ๋กœ๋ด‡์„ ํ™•๋ณดํ•  ์ˆ˜ ์žˆ๋Š” ์œ ์ผํ•œ ๋ฐฉ๋ฒ•์ž…๋‹ˆ๋‹ค. ์ €ํฌ๋Š” ์ด๋ฅผ ์œ„ํ•ด ๊ตฌ์ถ•ํ•˜๊ณ  ์žˆ์œผ๋ฉฐ, ํ˜„์žฌ ๋งค์šฐ ๋น ๋ฅด๊ฒŒ ์ง„์ „๋˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์—ฌ๊ธฐ ๋ช‡ ๊ฐ€์ง€ ์—…๊ณ„ ์ตœ์ดˆ ์‚ฌ๋ก€๋“ค์ด ์žˆ์Šต๋‹ˆ๋‹ค. ๋ชจ๋“  ๊ณณ์—์„œ ๋†€๋ผ์šด ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜๋“ค์ด ๋“ฑ์žฅํ•˜๊ณ  ์žˆ๋Š” ๊ฒƒ์„ ๋ณผ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ์ €ํฌ๋Š” ์ด ์„ธ ๊ฐ€์ง€ ์ปดํ“จํ„ฐ ํŠธ๋ฆฌ๋‹ˆํ‹ฐ์ธ Syncron์„ ์ •๋ง๋กœ ํ™œ์šฉํ•˜๊ณ  ๊ทธ๋กœ๋ถ€ํ„ฐ ํ˜œํƒ์„ ๋ฐ›๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์ •๋ง ๋†€๋ž์Šต๋‹ˆ๋‹ค.
They're [stent] (ph) for brain computer interfaces so that humans who can't move can interact with the physical world. Moon Surgical's Maestro is a surgical assistant. They're training task automation where the robot will be able to move on its own. Neptune Medical and their Pathfinder for robotic endoscopy and Virtual Incision, their MIRA and their next-generation miniature robotic surgery and even being used for tele-operation. So we're just starting to see some amazing breakthrough devices come to market. So we're announcing, this is our final announcement that we are partnering with the Mayo Clinic, the number one hospital in the world for many, many specialties.๋‡Œ-์ปดํ“จํ„ฐ ์ธํ„ฐํŽ˜์ด์Šค์šฉ [์Šคํ…ํŠธ](ph)๋กœ, ์›€์ง์ผ ์ˆ˜ ์—†๋Š” ํ™˜์ž๋“ค์ด ๋ฌผ๋ฆฌ์  ์„ธ๊ณ„์™€ ์ƒํ˜ธ์ž‘์šฉํ•  ์ˆ˜ ์žˆ๊ฒŒ ํ•ด์ค๋‹ˆ๋‹ค. Moon Surgical์˜ Maestro๋Š” ์ˆ˜์ˆ  ๋ณด์กฐ ์žฅ์น˜์ž…๋‹ˆ๋‹ค. ์ด๋“ค์€ ๋กœ๋ด‡์ด ์ž์œจ์ ์œผ๋กœ ์›€์ง์ผ ์ˆ˜ ์žˆ๋Š” ์ž‘์—… ์ž๋™ํ™”๋ฅผ ํ›ˆ๋ จ์‹œํ‚ค๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. Neptune Medical๊ณผ ๊ทธ๋“ค์˜ ๋กœ๋ด‡ ๋‚ด์‹œ๊ฒฝ์šฉ Pathfinder, ๊ทธ๋ฆฌ๊ณ  Virtual Incision์˜ MIRA์™€ ์ฐจ์„ธ๋Œ€ ์†Œํ˜• ๋กœ๋ด‡ ์ˆ˜์ˆ  ์žฅ๋น„๋Š” ์›๊ฒฉ ์กฐ์ž‘์—๋„ ์‚ฌ์šฉ๋˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์ด์ฒ˜๋Ÿผ ์šฐ๋ฆฌ๋Š” ์‹œ์žฅ์— ์ถœ์‹œ๋˜๋Š” ๋†€๋ผ์šด ํ˜์‹ ์  ์˜๋ฃŒ๊ธฐ๊ธฐ๋“ค์„ ์ด์ œ ๋ง‰ ๋ณด๊ธฐ ์‹œ์ž‘ํ–ˆ์Šต๋‹ˆ๋‹ค. ๋งˆ์ง€๋ง‰ ๋ฐœํ‘œ๋กœ, ์šฐ๋ฆฌ๋Š” ์—ฌ๋Ÿฌ ์ „๋ฌธ ๋ถ„์•ผ์—์„œ ์„ธ๊ณ„ 1์œ„ ๋ณ‘์›์ธ ๋ฉ”์ด์š” ํด๋ฆฌ๋‹‰๊ณผ ํŒŒํŠธ๋„ˆ์‹ญ์„ ๋งบ๋Š”๋‹ค๊ณ  ๋ฐœํ‘œํ•ฉ๋‹ˆ๋‹ค.
And we are accelerating the next generation of AI-driven digital pathology. Mayo today announced the Mayo Digital Pathology platform. This is built on an incredibly state-of-the-art robotic digital pathology labs that has amassed a huge and unique data set of over 20 million whole-slide images that has 10 million associated patient records. We're going to -- they're going to be building this on their new NVIDIA DGX Blackwell, which has 1.4 terabytes of GPU memory, absolutely necessary for these gigantic data sets. We're going to be applying the future of Cosmos and these multimodal vision to really understand much, much more deeply about biology. So this is an incredible partnership.๊ทธ๋ฆฌ๊ณ  ์šฐ๋ฆฌ๋Š” AI ๊ธฐ๋ฐ˜ ์ฐจ์„ธ๋Œ€ ๋””์ง€ํ„ธ ๋ณ‘๋ฆฌํ•™์„ ๊ฐ€์†ํ™”ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ๋ฉ”์ด์š” ํด๋ฆฌ๋‹‰์€ ์˜ค๋Š˜ ๋ฉ”์ด์š” ๋””์ง€ํ„ธ ๋ณ‘๋ฆฌํ•™ ํ”Œ๋žซํผ์„ ๋ฐœํ‘œํ–ˆ์Šต๋‹ˆ๋‹ค. ์ด๋Š” 2์ฒœ๋งŒ ๊ฐœ ์ด์ƒ์˜ ์ „์ฒด ์Šฌ๋ผ์ด๋“œ ์ด๋ฏธ์ง€์™€ 1์ฒœ๋งŒ ๊ฐœ์˜ ๊ด€๋ จ ํ™˜์ž ๊ธฐ๋ก์œผ๋กœ ๊ตฌ์„ฑ๋œ ๊ฑฐ๋Œ€ํ•˜๊ณ  ๋…ํŠนํ•œ ๋ฐ์ดํ„ฐ์…‹์„ ์ถ•์ ํ•œ ์ตœ์ฒจ๋‹จ ๋กœ๋ด‡ ๋””์ง€ํ„ธ ๋ณ‘๋ฆฌํ•™ ์—ฐ๊ตฌ์†Œ๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ๊ตฌ์ถ•๋˜์—ˆ์Šต๋‹ˆ๋‹ค. ์ด๋“ค์€ 1.4ํ…Œ๋ผ๋ฐ”์ดํŠธ์˜ GPU ๋ฉ”๋ชจ๋ฆฌ๋ฅผ ๊ฐ–์ถ˜ ์ƒˆ๋กœ์šด NVIDIA DGX Blackwell ์œ„์— ์ด๋ฅผ ๊ตฌ์ถ•ํ•  ์˜ˆ์ •์ธ๋ฐ, ์ด๋Š” ์ด๋Ÿฌํ•œ ๊ฑฐ๋Œ€ํ•œ ๋ฐ์ดํ„ฐ์…‹์— ์ ˆ๋Œ€์ ์œผ๋กœ ํ•„์š”ํ•œ ์‚ฌ์–‘์ž…๋‹ˆ๋‹ค. ์šฐ๋ฆฌ๋Š” ๋ฏธ๋ž˜์˜ Cosmos์™€ ์ด๋Ÿฌํ•œ ๋ฉ€ํ‹ฐ๋ชจ๋‹ฌ ๋น„์ „์„ ์ ์šฉํ•˜์—ฌ ์ƒ๋ฌผํ•™์— ๋Œ€ํ•ด ํ›จ์”ฌ ๋” ๊นŠ์ด ์ดํ•ดํ•  ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์ด๋Š” ์ •๋ง ๋†€๋ผ์šด ํŒŒํŠธ๋„ˆ์‹ญ์ž…๋‹ˆ๋‹ค.
And we're tackling biology at all scale from the genomics scale all the way through the tissue scale. And by doing this, you're going to see we're going to understand biology at multiscale in the not-too-distant future. So exciting times ahead. So I will -- I'm going to wrap up now. AI agents, develop and deploy domain-specific agents. It's so important. The LLMs themselves are not domain enough in order to have the impact that we need, which is why partnering with IQVIA, their domain expertise, stretches decades of deep, deep understanding what it takes to do clinical trials. And you need to put that knowledge and understanding and deeply embed it as the AI agent.๊ทธ๋ฆฌ๊ณ  ์šฐ๋ฆฌ๋Š” ์œ ์ „์ฒดํ•™ ๊ทœ๋ชจ๋ถ€ํ„ฐ ์กฐ์ง ๊ทœ๋ชจ์— ์ด๋ฅด๊ธฐ๊นŒ์ง€ ๋ชจ๋“  ๊ทœ๋ชจ์—์„œ ์ƒ๋ฌผํ•™์„ ๋‹ค๋ฃจ๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์ด๋ฅผ ํ†ตํ•ด ๋จธ์ง€์•Š์€ ๋ฏธ๋ž˜์— ์šฐ๋ฆฌ๊ฐ€ ๋‹ค์ค‘ ๊ทœ๋ชจ์—์„œ ์ƒ๋ฌผํ•™์„ ์ดํ•ดํ•˜๊ฒŒ ๋  ๊ฒƒ์„ ๋ณด์‹œ๊ฒŒ ๋  ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์ •๋ง ํฅ๋ฏธ์ง„์ง„ํ•œ ์‹œ๋Œ€๊ฐ€ ์•ž์— ์žˆ์Šต๋‹ˆ๋‹ค.

์ด์ œ ๋งˆ๋ฌด๋ฆฌํ•˜๊ฒ ์Šต๋‹ˆ๋‹ค. AI ์—์ด์ „ํŠธ, ๋„๋ฉ”์ธ๋ณ„ ํŠนํ™” ์—์ด์ „ํŠธ๋ฅผ ๊ฐœ๋ฐœํ•˜๊ณ  ๋ฐฐํฌํ•˜๋Š” ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์ด๊ฒƒ์€ ๋งค์šฐ ์ค‘์š”ํ•ฉ๋‹ˆ๋‹ค. LLM ์ž์ฒด๋งŒ์œผ๋กœ๋Š” ์šฐ๋ฆฌ๊ฐ€ ํ•„์š”๋กœ ํ•˜๋Š” ์ž„ํŒฉํŠธ๋ฅผ ๊ฐ€์ง€๊ธฐ์— ์ถฉ๋ถ„ํžˆ ๋„๋ฉ”์ธ๋ณ„๋กœ ํŠนํ™”๋˜์–ด ์žˆ์ง€ ์•Š์Šต๋‹ˆ๋‹ค. ์ด๊ฒƒ์ด ๋ฐ”๋กœ IQVIA์™€์˜ ํŒŒํŠธ๋„ˆ์‹ญ์ด ์ค‘์š”ํ•œ ์ด์œ ์ž…๋‹ˆ๋‹ค. ๊ทธ๋“ค์˜ ๋„๋ฉ”์ธ ์ „๋ฌธ์„ฑ์€ ์ž„์ƒ์‹œํ—˜ ์ˆ˜ํ–‰์— ํ•„์š”ํ•œ ๊ฒƒ์ด ๋ฌด์—‡์ธ์ง€์— ๋Œ€ํ•œ ์ˆ˜์‹ญ ๋…„๊ฐ„์˜ ๊นŠ๊ณ  ๊นŠ์€ ์ดํ•ด๋ฅผ ๋ฐ”ํƒ•์œผ๋กœ ํ•ฉ๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ๊ทธ๋Ÿฌํ•œ ์ง€์‹๊ณผ ์ดํ•ด๋ฅผ AI ์—์ด์ „ํŠธ์— ๊นŠ์ด ๋‚ด์žฌํ™”์‹œ์ผœ์•ผ ํ•ฉ๋‹ˆ๋‹ค.
AI drug discovery to generate, predict, and design new molecular ideas for different types of diseases, different new disease or therapeutic platforms that are continuing to evolve. We've got to add a lot more activity and discovery in the early part to see and meet the need for medicines, the tens of thousands of disease. And we know that so much of the healthcare spend is in the form of services. We need to augment them in all ways that we can. So we see this as a several hundred billion-dollar opportunity of AI factories where we're assisting patients, doctors, nurses, drug discoverers, both digitally and physically. So don't take my word for it.AI ์‹ ์•ฝ ๊ฐœ๋ฐœ์„ ํ†ตํ•ด ๋‹ค์–‘ํ•œ ์งˆ๋ณ‘ ์œ ํ˜•๊ณผ ์ง€์†์ ์œผ๋กœ ์ง„ํ™”ํ•˜๊ณ  ์žˆ๋Š” ์ƒˆ๋กœ์šด ์งˆ๋ณ‘ ๋˜๋Š” ์น˜๋ฃŒ ํ”Œ๋žซํผ์„ ์œ„ํ•œ ์ƒˆ๋กœ์šด ๋ถ„์ž ์•„์ด๋””์–ด๋ฅผ ์ƒ์„ฑ, ์˜ˆ์ธก, ์„ค๊ณ„ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์ˆ˜๋งŒ ๊ฐ€์ง€ ์งˆ๋ณ‘์— ๋Œ€ํ•œ ์˜์•ฝํ’ˆ ์ˆ˜์š”๋ฅผ ์ถฉ์กฑํ•˜๊ธฐ ์œ„ํ•ด์„œ๋Š” ์ดˆ๊ธฐ ๋‹จ๊ณ„์—์„œ ํ›จ์”ฌ ๋” ๋งŽ์€ ํ™œ๋™๊ณผ ๋ฐœ๊ฒฌ์„ ์ถ”๊ฐ€ํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ์˜๋ฃŒ๋น„ ์ง€์ถœ์˜ ์ƒ๋‹น ๋ถ€๋ถ„์ด ์„œ๋น„์Šค ํ˜•ํƒœ๋ผ๋Š” ๊ฒƒ์„ ์•Œ๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์šฐ๋ฆฌ๋Š” ๊ฐ€๋Šฅํ•œ ๋ชจ๋“  ๋ฐฉ๋ฒ•์œผ๋กœ ์ด๋ฅผ ๋ณด๊ฐ•ํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค. ๋”ฐ๋ผ์„œ ์šฐ๋ฆฌ๋Š” ์ด๋ฅผ ํ™˜์ž, ์˜์‚ฌ, ๊ฐ„ํ˜ธ์‚ฌ, ์‹ ์•ฝ ๊ฐœ๋ฐœ์ž๋“ค์„ ๋””์ง€ํ„ธ๊ณผ ๋ฌผ๋ฆฌ์  ๋ฐฉ์‹ ๋ชจ๋‘๋กœ ์ง€์›ํ•˜๋Š” AI ํŒฉํ† ๋ฆฌ์˜ ์ˆ˜์ฒœ์–ต ๋‹ฌ๋Ÿฌ ๊ทœ๋ชจ ๊ธฐํšŒ๋กœ ๋ณด๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์ œ ๋ง๋งŒ ๋ฏฟ์ง€ ๋งˆ์‹œ๊ณ ์š”.
I'll leave you with one last video, and then Harlan and I will chat. [Video Presentation]

Thank you, everybody. Have a great JPMorgan.
๋งˆ์ง€๋ง‰์œผ๋กœ ํ•œ ๊ฐ€์ง€ ์˜์ƒ์„ ๋” ๋ณด์—ฌ๋“œ๋ฆฌ๊ณ , ๊ทธ ๋‹ค์Œ์— ํ• ๋ž€๊ณผ ์ œ๊ฐ€ ๋Œ€ํ™”๋ฅผ ๋‚˜๋ˆ„๊ฒ ์Šต๋‹ˆ๋‹ค. [์˜์ƒ ํ”„๋ ˆ์  ํ…Œ์ด์…˜]

๊ฐ์‚ฌํ•ฉ๋‹ˆ๋‹ค, ์—ฌ๋Ÿฌ๋ถ„. ์ข‹์€ JP๋ชจ๊ฑด ์ปจํผ๋Ÿฐ์Šค ๋˜์‹œ๊ธฐ ๋ฐ”๋ž๋‹ˆ๋‹ค.

๐Ÿ“Œ ์š”์•ฝ

โ€ข ์ฃผ์š” ๋ฐœํ‘œ ๋‚ด์šฉ:
- IQVIA์™€์˜ ํŒŒํŠธ๋„ˆ์‹ญ: ์ž„์ƒ์‹œํ—˜ ๊ฐ€์†ํ™” ๋ฐ ์˜๋ฃŒ ๋ฐ์ดํ„ฐ ํ™œ์šฉ์„ ์œ„ํ•œ AI ๋ชจ๋ธ ๊ฐœ๋ฐœ ํ˜‘๋ ฅ
- Arc Institute์™€ ์ƒ๋ฌผํ•™ ๊ธฐ๋ฐ˜ ๋ชจ๋ธ ๊ฐœ๋ฐœ ํŒŒํŠธ๋„ˆ์‹ญ ์ฒด๊ฒฐ
- Illumina์™€ ์œ ์ „์ฒด ๋ถ„์„ ํ”Œ๋žซํผ ํ˜‘๋ ฅ ๋ฐœํ‘œ
- Mayo Clinic๊ณผ AI ๊ธฐ๋ฐ˜ ๋””์ง€ํ„ธ ๋ณ‘๋ฆฌํ•™ ํ”Œ๋žซํผ ๊ฐœ๋ฐœ ํ˜‘๋ ฅ

โ€ข ์‚ฌ์—… ์ „๋žต ๋ฐ ์‹œ์žฅ ๊ธฐํšŒ:
- ํ—ฌ์Šค์ผ€์–ด AI ์‹œ์žฅ์„ ์ˆ˜์ฒœ์–ต ๋‹ฌ๋Ÿฌ ๊ทœ๋ชจ๋กœ ์ „๋ง
- NVIDIA Clara ํ”Œ๋žซํผ์„ ํ†ตํ•œ ์˜๋ฃŒ ์ƒํƒœ๊ณ„ ์ง€์›
- 3,500๊ฐœ ์ด์ƒ์˜ ํ—ฌ์Šค์ผ€์–ด ์Šคํƒ€ํŠธ์—…์ด NVIDIA Inception ํ”„๋กœ๊ทธ๋žจ ์ฐธ์—ฌ

โ€ข ๊ธฐ์ˆ  ํ˜์‹ :
- GenMol ๋“ฑ ์‹ ๊ทœ AI ๊ธฐ๋ฐ˜ ๋ถ„์ž ์„ค๊ณ„ ๋ชจ๋ธ ์ถœ์‹œ
- ๋ฌผ๋ฆฌ์  AI์™€ ๋กœ๋ด‡๊ณตํ•™ ๊ธฐ์ˆ ์„ ํ†ตํ•œ ์˜๋ฃŒ ์„œ๋น„์Šค ํ˜์‹ 
- NVIDIA Cosmos๋ฅผ ํ†ตํ•œ ๊ฐ€์ƒ ํ™˜๊ฒฝ์—์„œ์˜ ๋กœ๋ด‡ ํ•™์Šต ์ง€์›


โ“ Q&A

Original Translation
Question-and-Answer Session

Harlan Sur

Okay, perfect. Yeah, that was a great overview and update. Thank you, Kimberly. Oftentimes for myself, the sell-side analysts that cover NVIDIA, especially when it comes to the vertical industry segments like healthcare, your customer base, your partner base is so diverse, large pharma companies, small tech bio companies, educational institutions, research organizations. Your platform approach has meant full stack solutions for most programs, big or small, on-prem, hybrid, cloud-based, on the hardware side, DGX, compute on-prem, DGX Cloud, et cetera. And then on top of that, you've got your software stacks, right?
**์งˆ์˜์‘๋‹ต ์„ธ์…˜**

**ํ• ๋ž€ ์„œ(Harlan Sur)**

์ข‹์Šต๋‹ˆ๋‹ค, ์™„๋ฒฝํ•˜๋„ค์š”. ์ •๋ง ํ›Œ๋ฅญํ•œ ๊ฐœ์š”์™€ ์—…๋ฐ์ดํŠธ์˜€์Šต๋‹ˆ๋‹ค. ๊ฐ์‚ฌํ•ฉ๋‹ˆ๋‹ค, ํ‚ด๋ฒŒ๋ฆฌ. ์ €๋ฅผ ํฌํ•จํ•ด์„œ NVIDIA๋ฅผ ๋‹ด๋‹นํ•˜๋Š” ์…€์‚ฌ์ด๋“œ ์• ๋„๋ฆฌ์ŠคํŠธ๋“ค์—๊ฒŒ๋Š”, ํŠนํžˆ ํ—ฌ์Šค์ผ€์–ด์™€ ๊ฐ™์€ ์ˆ˜์ง ์‚ฐ์—… ๋ถ€๋ฌธ์— ๊ด€ํ•ด์„œ๋Š”, ๊ณ ๊ฐ ๊ธฐ๋ฐ˜๊ณผ ํŒŒํŠธ๋„ˆ ๊ธฐ๋ฐ˜์ด ๋งค์šฐ ๋‹ค์–‘ํ•ฉ๋‹ˆ๋‹ค. ๋Œ€ํ˜• ์ œ์•ฝํšŒ์‚ฌ๋“ค, ์†Œ๊ทœ๋ชจ ํ…Œํฌ ๋ฐ”์ด์˜ค ๊ธฐ์—…๋“ค, ๊ต์œก ๊ธฐ๊ด€๋“ค, ์—ฐ๊ตฌ ์กฐ์ง๋“ค๊นŒ์ง€ ๋ง์ด์ฃ . ๊ท€ํ•˜์˜ ํ”Œ๋žซํผ ์ ‘๊ทผ ๋ฐฉ์‹์€ ๊ทœ๋ชจ์˜ ๋Œ€์†Œ๋ฅผ ๋ถˆ๋ฌธํ•˜๊ณ  ๋Œ€๋ถ€๋ถ„์˜ ํ”„๋กœ๊ทธ๋žจ์— ๋Œ€ํ•ด ํ’€์Šคํƒ ์†”๋ฃจ์…˜์„ ์˜๋ฏธํ•ด์™”์Šต๋‹ˆ๋‹ค. ์˜จํ”„๋ ˆ๋ฏธ์Šค, ํ•˜์ด๋ธŒ๋ฆฌ๋“œ, ํด๋ผ์šฐ๋“œ ๊ธฐ๋ฐ˜ ๋ชจ๋‘์—์„œ ๋ง์ด์ฃ . ํ•˜๋“œ์›จ์–ด ์ธก๋ฉด์—์„œ๋Š” DGX, ์˜จํ”„๋ ˆ๋ฏธ์Šค ์ปดํ“จํŒ…, DGX ํด๋ผ์šฐ๋“œ ๋“ฑ์ด ์žˆ๊ณ ์š”. ๊ทธ๋ฆฌ๊ณ  ๊ทธ ์œ„์— ์†Œํ”„ํŠธ์›จ์–ด ์Šคํƒ๋“ค๋„ ์žˆ์ฃ , ๋งž๋‚˜์š”?
You've got base command, enterprise AI, software stacks. You've got NIMs. And then you have your different vertical industry sort of frameworks like Clara, BioNeMo, MONAI, Holoscan. I think we often wonder, how does someone like yourself managing the vertical industries, like, how do you map all of these monetization opportunities to your hardware and software programs, right? So, can you give us a sense for the mix of your business? Is it mainly subscription cloud-based type of engagements on the hardware side? Or are most of your customer engagements on-prem hardware-based systems? And what is the subscription license attach of your software and off-the-shelf frameworks?๊ธฐ๋ณธ ๋ช…๋ น์–ด, ์—”ํ„ฐํ”„๋ผ์ด์ฆˆ AI, ์†Œํ”„ํŠธ์›จ์–ด ์Šคํƒ๋“ค์ด ์žˆ์Šต๋‹ˆ๋‹ค. NIM๋“ค๋„ ์žˆ๊ณ ์š”. ๊ทธ๋ฆฌ๊ณ  Clara, BioNeMo, MONAI, Holoscan๊ณผ ๊ฐ™์€ ๋‹ค์–‘ํ•œ ์ˆ˜์ง ์‚ฐ์—…๋ณ„ ํ”„๋ ˆ์ž„์›Œํฌ๋“ค๋„ ์žˆ์Šต๋‹ˆ๋‹ค. ์ €ํฌ๋Š” ์ข…์ข… ๊ถ๊ธˆํ•ดํ•˜๋Š”๋ฐ, ์ˆ˜์ง ์‚ฐ์—…์„ ๊ด€๋ฆฌํ•˜์‹œ๋Š” ๋ถ„๊ป˜์„œ๋Š” ์ด ๋ชจ๋“  ์ˆ˜์ตํ™” ๊ธฐํšŒ๋“ค์„ ํ•˜๋“œ์›จ์–ด ๋ฐ ์†Œํ”„ํŠธ์›จ์–ด ํ”„๋กœ๊ทธ๋žจ์— ์–ด๋–ป๊ฒŒ ๋งคํ•‘ํ•˜์‹œ๋Š”์ง€์š”? ๊ทธ๋ž˜์„œ ๊ท€ํ•˜์˜ ์‚ฌ์—… ๋ฏน์Šค์— ๋Œ€ํ•œ ๊ฐ์„ ์ฃผ์‹ค ์ˆ˜ ์žˆ๋‚˜์š”? ์ฃผ๋กœ ๊ตฌ๋… ๊ธฐ๋ฐ˜ ํด๋ผ์šฐ๋“œ ํ˜•ํƒœ์˜ ํ•˜๋“œ์›จ์–ด ์ธก๋ฉด ์ฐธ์—ฌ์ธ๊ฐ€์š”? ์•„๋‹ˆ๋ฉด ๋Œ€๋ถ€๋ถ„์˜ ๊ณ ๊ฐ ์ฐธ์—ฌ๊ฐ€ ์˜จํ”„๋ ˆ๋ฏธ์Šค ํ•˜๋“œ์›จ์–ด ๊ธฐ๋ฐ˜ ์‹œ์Šคํ…œ์ธ๊ฐ€์š”? ๊ทธ๋ฆฌ๊ณ  ๊ท€ํ•˜์˜ ์†Œํ”„ํŠธ์›จ์–ด ๋ฐ ๊ธฐ์„ฑ ํ”„๋ ˆ์ž„์›Œํฌ์˜ ๊ตฌ๋… ๋ผ์ด์„ ์Šค ๋ถ€์ฐฉ๋ฅ ์€ ์–ด๋–ป๊ฒŒ ๋˜๋‚˜์š”?
Kimberly Powell

Yeah. I think the way to think about this is everything that we build in the NVIDIA Clara platform and the ability to democratize it, we want it to -- we're driving consumption and it's going to terminate on a GPU somewhere, and we actually don't care. And so there is a huge shift in the world to cloud and we're delighted by it. And so we are working -- it used to be the days where you had to build the supercomputer. But now NVIDIA DGX Cloud is available in every public cloud, right? And so whether it'd be inferencing platforms, very high-performance computing platforms, NVIDIA Holoscan, which you, in your mind, probably think about it as running on an edge computer.
ํ‚ด๋ฒŒ๋ฆฌ ํŒŒ์›”

๋„ค. ์ด์— ๋Œ€ํ•ด ์ƒ๊ฐํ•ด๋ณด๋Š” ๋ฐฉ์‹์€ ์šฐ๋ฆฌ๊ฐ€ NVIDIA Clara ํ”Œ๋žซํผ์—์„œ ๊ตฌ์ถ•ํ•˜๋Š” ๋ชจ๋“  ๊ฒƒ๊ณผ ์ด๋ฅผ ๋ฏผ์ฃผํ™”ํ•˜๋Š” ๋Šฅ๋ ฅ์— ๊ด€ํ•œ ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์šฐ๋ฆฌ๋Š” ์ด๊ฒƒ์ด ์†Œ๋น„๋ฅผ ์ด‰์ง„ํ•˜๊ธฐ๋ฅผ ์›ํ•˜๋ฉฐ, ๊ฒฐ๊ตญ ์–ด๋”˜๊ฐ€์˜ GPU์—์„œ ์ข…๋ฃŒ๋  ๊ฒƒ์ž…๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ์‹ค์ œ๋กœ ์šฐ๋ฆฌ๋Š” ์–ด๋””์„œ๋“  ์ƒ๊ด€์—†์Šต๋‹ˆ๋‹ค. ์„ธ์ƒ์—๋Š” ํด๋ผ์šฐ๋“œ๋กœ์˜ ๊ฑฐ๋Œ€ํ•œ ์ „ํ™˜์ด ์ผ์–ด๋‚˜๊ณ  ์žˆ์œผ๋ฉฐ ์šฐ๋ฆฌ๋Š” ์ด๋ฅผ ๋งค์šฐ ๊ธฐ๋ปํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ๊ทธ๋ž˜์„œ ์šฐ๋ฆฌ๋Š” ์ž‘์—…ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค -- ์˜ˆ์ „์—๋Š” ์Šˆํผ์ปดํ“จํ„ฐ๋ฅผ ๊ตฌ์ถ•ํ•ด์•ผ ํ–ˆ๋˜ ์‹œ์ ˆ์ด ์žˆ์—ˆ์Šต๋‹ˆ๋‹ค. ํ•˜์ง€๋งŒ ์ด์ œ NVIDIA DGX Cloud๋Š” ๋ชจ๋“  ํผ๋ธ”๋ฆญ ํด๋ผ์šฐ๋“œ์—์„œ ์ด์šฉ ๊ฐ€๋Šฅํ•ฉ๋‹ˆ๋‹ค, ๋งž์ฃ ? ๊ทธ๋ž˜์„œ ์ถ”๋ก  ํ”Œ๋žซํผ์ด๋“ , ๋งค์šฐ ๊ณ ์„ฑ๋Šฅ ์ปดํ“จํŒ… ํ”Œ๋žซํผ์ด๋“ , NVIDIA Holoscan์ด๋“  -- ์—ฌ๋Ÿฌ๋ถ„์ด ๋งˆ์Œ์†์œผ๋กœ๋Š” ์•„๋งˆ ์—ฃ์ง€ ์ปดํ“จํ„ฐ์—์„œ ์‹คํ–‰๋˜๋Š” ๊ฒƒ์œผ๋กœ ์ƒ๊ฐํ•˜์‹ค ํ…๋ฐ์š”.
It runs on the cloud just the same. Every customer, no matter domain, needs to have that flexibility, especially in healthcare, right? You need to have the ability to sometimes run on-prem because that might be the company or institution's preference. It might need to run in-country because that is the sovereign AI strategy that is upon us. And so there's a huge shift into cloud. But it has a very hybrid capability because you're going to want to run the models where it makes the most sense in terms of data sovereignty, in terms of latency. But the platform is completely hybrid cloud. And so the customers get the choice. All of our software has a very simple go-to-market.ํด๋ผ์šฐ๋“œ์—์„œ๋„ ๋™์ผํ•˜๊ฒŒ ์ž‘๋™ํ•ฉ๋‹ˆ๋‹ค. ๋„๋ฉ”์ธ์— ๊ด€๊ณ„์—†์ด ๋ชจ๋“  ๊ณ ๊ฐ์€ ๊ทธ๋Ÿฌํ•œ ์œ ์—ฐ์„ฑ์„ ํ•„์š”๋กœ ํ•˜๋ฉฐ, ํŠนํžˆ ํ—ฌ์Šค์ผ€์–ด ๋ถ„์•ผ์—์„œ๋Š” ๋”์šฑ ๊ทธ๋ ‡์Šต๋‹ˆ๋‹ค. ๋•Œ๋กœ๋Š” ์˜จํ”„๋ ˆ๋ฏธ์Šค์—์„œ ์‹คํ–‰ํ•  ์ˆ˜ ์žˆ๋Š” ๋Šฅ๋ ฅ์ด ํ•„์š”ํ•œ๋ฐ, ์ด๋Š” ํ•ด๋‹น ํšŒ์‚ฌ๋‚˜ ๊ธฐ๊ด€์˜ ์„ ํ˜ธ๋„์ผ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ๋˜ํ•œ ๊ตญ๊ฐ€ ๋‚ด์—์„œ ์‹คํ–‰ํ•ด์•ผ ํ•  ํ•„์š”๊ฐ€ ์žˆ์„ ์ˆ˜ ์žˆ๋Š”๋ฐ, ์ด๋Š” ์šฐ๋ฆฌ ์•ž์— ๋†“์ธ ์ฃผ๊ถŒ AI ์ „๋žต ๋•Œ๋ฌธ์ž…๋‹ˆ๋‹ค. ๊ทธ๋ž˜์„œ ํด๋ผ์šฐ๋“œ๋กœ์˜ ๋Œ€๊ทœ๋ชจ ์ „ํ™˜์ด ์ผ์–ด๋‚˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ํ•˜์ง€๋งŒ ์ด๋Š” ๋งค์šฐ ํ•˜์ด๋ธŒ๋ฆฌ๋“œํ•œ ์—ญ๋Ÿ‰์„ ๊ฐ€์ง€๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ์™œ๋ƒํ•˜๋ฉด ๋ฐ์ดํ„ฐ ์ฃผ๊ถŒ ์ธก๋ฉด์—์„œ, ์ง€์—ฐ์‹œ๊ฐ„ ์ธก๋ฉด์—์„œ ๊ฐ€์žฅ ํ•ฉ๋ฆฌ์ ์ธ ๊ณณ์—์„œ ๋ชจ๋ธ์„ ์‹คํ–‰ํ•˜๊ณ  ์‹ถ์–ดํ•  ๊ฒƒ์ด๊ธฐ ๋•Œ๋ฌธ์ž…๋‹ˆ๋‹ค. ํ•˜์ง€๋งŒ ํ”Œ๋žซํผ์€ ์™„์ „ํžˆ ํ•˜์ด๋ธŒ๋ฆฌ๋“œ ํด๋ผ์šฐ๋“œ์ž…๋‹ˆ๋‹ค. ๋”ฐ๋ผ์„œ ๊ณ ๊ฐ๋“ค์ด ์„ ํƒ๊ถŒ์„ ๊ฐ–๊ฒŒ ๋ฉ๋‹ˆ๋‹ค. ์ €ํฌ์˜ ๋ชจ๋“  ์†Œํ”„ํŠธ์›จ์–ด๋Š” ๋งค์šฐ ๊ฐ„๋‹จํ•œ ์‹œ์žฅ์ง„์ถœ ์ „๋žต์„ ๊ฐ€์ง€๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.
It's through NVIDIA AI Enterprise subscriptions. It's a very publicly known price. It's $4,500 per GPU per year, which is essentially $1 per GPU-hour. And that is the subscription of all the NVIDIA Enterprise software. And that's how we go to market. Harlan Sur

There has been a lot of discussions back at the earnings call, back at, as you mentioned, CES about the opportunities in agentic AI and multi-agent systems that address the challenges of workforce shortages, rising healthcare costs. You announced here, right, your efforts with IQVIA to accelerate research using AI agents. Like what else is NVIDIA doing in this fast-growing subsegment of digital health agents? Kimberly Powell

Yeah.
NVIDIA AI Enterprise ๊ตฌ๋…์„ ํ†ตํ•ด์„œ์ž…๋‹ˆ๋‹ค. ์ด๋Š” ๋งค์šฐ ๊ณต๊ฐœ์ ์œผ๋กœ ์•Œ๋ ค์ง„ ๊ฐ€๊ฒฉ์œผ๋กœ, GPU๋‹น ์—ฐ๊ฐ„ 4,500๋‹ฌ๋Ÿฌ, ์ฆ‰ ๋ณธ์งˆ์ ์œผ๋กœ GPU ์‹œ๊ฐ„๋‹น 1๋‹ฌ๋Ÿฌ์ž…๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ์ด๊ฒƒ์ด ๋ชจ๋“  NVIDIA Enterprise ์†Œํ”„ํŠธ์›จ์–ด์˜ ๊ตฌ๋…๋ฃŒ์ž…๋‹ˆ๋‹ค. ์ด๊ฒƒ์ด ์šฐ๋ฆฌ๊ฐ€ ์‹œ์žฅ์— ์ง„์ถœํ•˜๋Š” ๋ฐฉ์‹์ž…๋‹ˆ๋‹ค.

**ํ• ๋ž€ ์„œ(Harlan Sur)**

์‹ค์  ๋ฐœํ‘œ์—์„œ, ๊ทธ๋ฆฌ๊ณ  ๋ง์”€ํ•˜์‹  ๋Œ€๋กœ CES์—์„œ ์ธ๋ ฅ ๋ถ€์กฑ, ์˜๋ฃŒ๋น„ ์ƒ์Šน ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ•˜๋Š” ์—์ด์ „ํ‹ฑ AI์™€ ๋ฉ€ํ‹ฐ ์—์ด์ „ํŠธ ์‹œ์Šคํ…œ์˜ ๊ธฐํšŒ์— ๋Œ€ํ•œ ๋งŽ์€ ๋…ผ์˜๊ฐ€ ์žˆ์—ˆ์Šต๋‹ˆ๋‹ค. ์—ฌ๊ธฐ์„œ AI ์—์ด์ „ํŠธ๋ฅผ ํ™œ์šฉํ•œ ์—ฐ๊ตฌ ๊ฐ€์†ํ™”๋ฅผ ์œ„ํ•œ IQVIA์™€์˜ ํ˜‘๋ ฅ ๋…ธ๋ ฅ์„ ๋ฐœํ‘œํ•˜์…จ๋Š”๋ฐ์š”. ๋น ๋ฅด๊ฒŒ ์„ฑ์žฅํ•˜๋Š” ๋””์ง€ํ„ธ ํ—ฌ์Šค ์—์ด์ „ํŠธ ํ•˜์œ„ ๋ถ€๋ฌธ์—์„œ NVIDIA๊ฐ€ ์ถ”๊ฐ€๋กœ ํ•˜๊ณ  ์žˆ๋Š” ์ผ์€ ๋ฌด์—‡์ธ๊ฐ€์š”?

**ํ‚ด๋ฒŒ๋ฆฌ ํŒŒ์›”(Kimberly Powell)**

๋„ค.
We're -- so NVIDIA AI Enterprise platform, as I said, one, we build the micro services. We turn these incredibly capable AI models, and we turn them into a super easy-to-use services micro service, which is an API. It's literally an API which means it can be integrated or embedded into any application out there, right? So it can just essentially become a new capability overnight by a couple of lines of code. So one is having a huge library of these models. The next thing that we're doing to democratize this is we're turning them into reference applications, these Blueprints. These are not end market applications.์ €ํฌ๋Š” -- NVIDIA AI Enterprise ํ”Œ๋žซํผ์— ๋Œ€ํ•ด ๋ง์”€๋“œ๋ฆฌ๋ฉด, ์ฒซ์งธ๋กœ ์ €ํฌ๊ฐ€ ๋งˆ์ดํฌ๋กœ์„œ๋น„์Šค๋ฅผ ๊ตฌ์ถ•ํ•ฉ๋‹ˆ๋‹ค. ์ €ํฌ๋Š” ์ด๋Ÿฌํ•œ ๋†€๋ผ์šธ ์ •๋„๋กœ ๋›ฐ์–ดํ•œ AI ๋ชจ๋ธ๋“ค์„ ๊ฐ€์ ธ์™€์„œ ๋งค์šฐ ์‚ฌ์šฉํ•˜๊ธฐ ์‰ฌ์šด ์„œ๋น„์Šค ๋งˆ์ดํฌ๋กœ์„œ๋น„์Šค, ์ฆ‰ API๋กœ ์ „ํ™˜ํ•ฉ๋‹ˆ๋‹ค. ๋ง ๊ทธ๋Œ€๋กœ API์ด๊ธฐ ๋•Œ๋ฌธ์— ๊ธฐ์กด์˜ ๋ชจ๋“  ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์— ํ†ตํ•ฉ๋˜๊ฑฐ๋‚˜ ์ž„๋ฒ ๋“œ๋  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ๋”ฐ๋ผ์„œ ๋ช‡ ์ค„์˜ ์ฝ”๋“œ๋งŒ์œผ๋กœ ํ•˜๋ฃจ์•„์นจ์— ์ƒˆ๋กœ์šด ๊ธฐ๋Šฅ์ด ๋  ์ˆ˜ ์žˆ๋Š” ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์ฒซ ๋ฒˆ์งธ๋Š” ์ด๋Ÿฌํ•œ ๋ชจ๋ธ๋“ค์˜ ๋ฐฉ๋Œ€ํ•œ ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ๋ฅผ ๋ณด์œ ํ•˜๋Š” ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์ด๋ฅผ ๋ฏผ์ฃผํ™”ํ•˜๊ธฐ ์œ„ํ•ด ์ €ํฌ๊ฐ€ ํ•˜๊ณ  ์žˆ๋Š” ๋‹ค์Œ ์ž‘์—…์€ ์ด๋“ค์„ ์ฐธ์กฐ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์ธ ๋ธ”๋ฃจํ”„๋ฆฐํŠธ๋กœ ์ „ํ™˜ํ•˜๋Š” ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์ด๋“ค์€ ์ตœ์ข… ์‹œ์žฅ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์ด ์•„๋‹™๋‹ˆ๋‹ค.
They're not a SaaS you subscribe to but they're blueprints for our partners to take advantage of. And in fact, you've seen a lot of announcements more recently with NVIDIA. We're really, really getting involved with the global system integrators. Accenture and Deloitte, Quantify, Capgemini are a necessary part of the strategy so that we can take these blueprints but do that last-mile customization integration into the domain expert agents they need to become. And so that is how we really approach this. And then we have NeMo, which is the flywheel, the data flywheel. Every enterprise should be capturing all customer interactions.์ด๋“ค์€ ๊ตฌ๋…ํ•˜๋Š” SaaS๊ฐ€ ์•„๋‹ˆ๋ผ ํŒŒํŠธ๋„ˆ๋“ค์ด ํ™œ์šฉํ•  ์ˆ˜ ์žˆ๋Š” ์ฒญ์‚ฌ์ง„์ž…๋‹ˆ๋‹ค. ์‹ค์ œ๋กœ ์ตœ๊ทผ NVIDIA์™€์˜ ๋งŽ์€ ๋ฐœํ‘œ๋ฅผ ๋ณด์…จ์„ ํ…๋ฐ์š”. ์ €ํฌ๋Š” ๊ธ€๋กœ๋ฒŒ ์‹œ์Šคํ…œ ํ†ตํ•ฉ์—…์ฒด๋“ค๊ณผ ์ •๋ง๋กœ ๊นŠ์ด ๊ด€์—ฌํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. Accenture, Deloitte, Quantify, Capgemini๋Š” ์ €ํฌ ์ „๋žต์˜ ํ•„์ˆ˜์ ์ธ ๋ถ€๋ถ„์ž…๋‹ˆ๋‹ค. ์ด๋ฅผ ํ†ตํ•ด ์ด๋Ÿฌํ•œ ์ฒญ์‚ฌ์ง„๋“ค์„ ๊ฐ€์ ธ์™€์„œ ๊ทธ๋“ค์ด ํ•„์š”๋กœ ํ•˜๋Š” ๋„๋ฉ”์ธ ์ „๋ฌธ๊ฐ€ ์—์ด์ „ํŠธ๋กœ ๋ณ€๋ชจํ•˜๊ธฐ ์œ„ํ•œ ๋ผ์ŠคํŠธ๋งˆ์ผ ๋งž์ถคํ™” ํ†ตํ•ฉ์„ ์ˆ˜ํ–‰ํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ์ด๊ฒƒ์ด ๋ฐ”๋กœ ์ €ํฌ๊ฐ€ ์ ‘๊ทผํ•˜๋Š” ๋ฐฉ์‹์ž…๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ์ €ํฌ์—๊ฒŒ๋Š” NeMo๊ฐ€ ์žˆ๋Š”๋ฐ, ์ด๋Š” ํ”Œ๋ผ์ดํœ , ์ฆ‰ ๋ฐ์ดํ„ฐ ํ”Œ๋ผ์ดํœ ์ž…๋‹ˆ๋‹ค. ๋ชจ๋“  ๊ธฐ์—…์€ ๋ชจ๋“  ๊ณ ๊ฐ ์ƒํ˜ธ์ž‘์šฉ์„ ํฌ์ฐฉํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค.
They should be capturing everything about how their products are being used, and taking that as new data and intelligence to refine these agents to perform better and better and better over time. And so we've really taken that full stack approach from helping amazing companies who have unique data sets build their own proprietary foundation models, turning them into micro services, blueprint as agents and then, of course, running them on these large-scale AI factories that they can scale across the world. Harlan Sur

One of the big unveils at your developers conference last year called GTC was physical AI. You touched upon it in your presentation.
๊ทธ๋“ค์€ ์ž์‹ ๋“ค์˜ ์ œํ’ˆ์ด ์–ด๋–ป๊ฒŒ ์‚ฌ์šฉ๋˜๊ณ  ์žˆ๋Š”์ง€์— ๋Œ€ํ•œ ๋ชจ๋“  ๊ฒƒ์„ ํฌ์ฐฉํ•˜๊ณ , ์ด๋ฅผ ์ƒˆ๋กœ์šด ๋ฐ์ดํ„ฐ์™€ ์ธํ…”๋ฆฌ์ „์Šค๋กœ ํ™œ์šฉํ•˜์—ฌ ์ด๋Ÿฌํ•œ ์—์ด์ „ํŠธ๋“ค์ด ์‹œ๊ฐ„์ด ์ง€๋‚จ์— ๋”ฐ๋ผ ์ ์  ๋” ๋‚˜์€ ์„ฑ๋Šฅ์„ ๋ฐœํœ˜ํ•˜๋„๋ก ๊ฐœ์„ ํ•ด์•ผ ํ•ฉ๋‹ˆ๋‹ค. ๊ทธ๋ž˜์„œ ์ €ํฌ๋Š” ๊ณ ์œ ํ•œ ๋ฐ์ดํ„ฐ์…‹์„ ๋ณด์œ ํ•œ ๋†€๋ผ์šด ๊ธฐ์—…๋“ค์ด ์ž์ฒด์ ์ธ ๋…์  ํŒŒ์šด๋ฐ์ด์…˜ ๋ชจ๋ธ์„ ๊ตฌ์ถ•ํ•˜๊ณ , ์ด๋ฅผ ๋งˆ์ดํฌ๋กœ์„œ๋น„์Šค๋กœ ์ „ํ™˜ํ•˜๋ฉฐ, ์—์ด์ „ํŠธ๋กœ ์ฒญ์‚ฌ์ง„์„ ๋งŒ๋“ค๊ณ , ๋ฌผ๋ก  ์ „ ์„ธ๊ณ„์ ์œผ๋กœ ํ™•์žฅํ•  ์ˆ˜ ์žˆ๋Š” ์ด๋Ÿฌํ•œ ๋Œ€๊ทœ๋ชจ AI ํŒฉํ† ๋ฆฌ์—์„œ ์šด์˜ํ•  ์ˆ˜ ์žˆ๋„๋ก ๋•๋Š” ํ’€์Šคํƒ ์ ‘๊ทผ ๋ฐฉ์‹์„ ์ •๋ง๋กœ ์ทจํ•ด์™”์Šต๋‹ˆ๋‹ค.

ํ• ๋ž€ ์„œ(Harlan Sur)

์ž‘๋…„ GTC๋ผ๊ณ  ๋ถˆ๋ฆฌ๋Š” ๊ฐœ๋ฐœ์ž ์ปจํผ๋Ÿฐ์Šค์—์„œ ๊ณต๊ฐœ๋œ ํฐ ๋ฐœํ‘œ ์ค‘ ํ•˜๋‚˜๊ฐ€ ํ”ผ์ง€์ปฌ AI์˜€์Šต๋‹ˆ๋‹ค. ํ”„๋ ˆ์  ํ…Œ์ด์…˜์—์„œ ์ด์— ๋Œ€ํ•ด ์–ธ๊ธ‰ํ•˜์…จ๋Š”๋ฐ์š”.
Jensen talked about it at the recent Consumer Electronics Show. Models that perceive, understand, interact with the physical world with one of the use cases being to train AI robots, right? Omniverse is the team's platform to simulate virtual physical roles that adhere to the laws of physics and is used to train robots. We saw the add-on capabilities with the announcement of Cosmos. Again, you talked about it, right, again, focused on physical AI. We can extend this to healthcare where robots can be used to assist in complex surgery, medical procedures. Training these robots in a virtual environment reduces risk, as you mentioned, associated with real-world trials.์  ์Šจ์ด ์ตœ๊ทผ ์†Œ๋น„์ž ์ „์ž ์ „์‹œํšŒ์—์„œ ์ด์— ๋Œ€ํ•ด ์–ธ๊ธ‰ํ–ˆ์Šต๋‹ˆ๋‹ค. ๋ฌผ๋ฆฌ์  ์„ธ๊ณ„๋ฅผ ์ธ์‹ํ•˜๊ณ , ์ดํ•ดํ•˜๋ฉฐ, ์ƒํ˜ธ์ž‘์šฉํ•˜๋Š” ๋ชจ๋ธ๋“ค๋กœ, ๊ทธ ํ™œ์šฉ ์‚ฌ๋ก€ ์ค‘ ํ•˜๋‚˜๊ฐ€ AI ๋กœ๋ด‡์„ ํ›ˆ๋ จ์‹œํ‚ค๋Š” ๊ฒƒ์ด์ฃ . ์˜ด๋‹ˆ๋ฒ„์Šค๋Š” ๋ฌผ๋ฆฌ ๋ฒ•์น™์„ ์ค€์ˆ˜ํ•˜๋Š” ๊ฐ€์ƒ ๋ฌผ๋ฆฌ์  ์—ญํ• ์„ ์‹œ๋ฎฌ๋ ˆ์ด์…˜ํ•˜๋Š” ํŒ€์˜ ํ”Œ๋žซํผ์œผ๋กœ, ๋กœ๋ด‡ ํ›ˆ๋ จ์— ์‚ฌ์šฉ๋ฉ๋‹ˆ๋‹ค. ์ฝ”์Šค๋ชจ์Šค ๋ฐœํ‘œ์™€ ํ•จ๊ป˜ ์• ๋“œ์˜จ ๊ธฐ๋Šฅ๋“ค์„ ํ™•์ธํ–ˆ์Šต๋‹ˆ๋‹ค. ๋‹ค์‹œ ๋ง์”€ํ•˜์‹  ๋ฐ”์™€ ๊ฐ™์ด, ๋ฌผ๋ฆฌ์  AI์— ์ดˆ์ ์„ ๋งž์ถ˜ ๊ฒƒ์ด์ฃ . ์ด๋ฅผ ํ—ฌ์Šค์ผ€์–ด ๋ถ„์•ผ๋กœ ํ™•์žฅํ•˜๋ฉด ๋กœ๋ด‡์ด ๋ณต์žกํ•œ ์ˆ˜์ˆ ์ด๋‚˜ ์˜๋ฃŒ ์‹œ์ˆ ์„ ๋ณด์กฐํ•˜๋Š” ๋ฐ ํ™œ์šฉ๋  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ๊ฐ€์ƒ ํ™˜๊ฒฝ์—์„œ ์ด๋Ÿฌํ•œ ๋กœ๋ด‡๋“ค์„ ํ›ˆ๋ จ์‹œํ‚ค๋Š” ๊ฒƒ์€ ๋ง์”€ํ•˜์‹  ๋Œ€๋กœ ์‹ค์ œ ํ™˜๊ฒฝ์—์„œ์˜ ์‹œํ—˜๊ณผ ๊ด€๋ จ๋œ ์œ„ํ—˜์„ ์ค„์—ฌ์ค๋‹ˆ๋‹ค.
It seems like you guys have actually adopted Omniverse Cosmos like framework, but whereas Jensen will talk about Omniverse and Cosmos, building simulation, automotive simulation and so on, right? What's the framework for building Omniverse and Cosmos from a healthcare perspective? Kimberly Powell

Yeah. As I said, our mission in healthcare business unit is how do you take these incredible technologies but make them more domain-specific or teach the world how to use them in a domain context. And so that example I gave you where you saw the robot picking up the needle, that's doing research with a Da Vinci Research Kit, an actual research kit, right?
๋„ค, ๋ง์”€ํ•˜์‹  ๋Œ€๋กœ ์—ฌ๋Ÿฌ๋ถ„์ด ์‹ค์ œ๋กœ Omniverse Cosmos์™€ ๊ฐ™์€ ํ”„๋ ˆ์ž„์›Œํฌ๋ฅผ ์ฑ„ํƒํ•˜์‹  ๊ฒƒ ๊ฐ™์Šต๋‹ˆ๋‹ค. ์  ์Šจ์ด Omniverse์™€ Cosmos, ์‹œ๋ฎฌ๋ ˆ์ด์…˜ ๊ตฌ์ถ•, ์ž๋™์ฐจ ์‹œ๋ฎฌ๋ ˆ์ด์…˜ ๋“ฑ์— ๋Œ€ํ•ด ์ด์•ผ๊ธฐํ•˜๋Š” ๋ฐ˜๋ฉด, ํ—ฌ์Šค์ผ€์–ด ๊ด€์ ์—์„œ Omniverse์™€ Cosmos๋ฅผ ๊ตฌ์ถ•ํ•˜๋Š” ํ”„๋ ˆ์ž„์›Œํฌ๋Š” ๋ฌด์—‡์ธ๊ฐ€์š”?

ํ‚ด๋ฒŒ๋ฆฌ ํŒŒ์›”:

๋„ค, ๋ง์”€๋“œ๋ฆฐ ๋ฐ”์™€ ๊ฐ™์ด ํ—ฌ์Šค์ผ€์–ด ์‚ฌ์—…๋ถ€์˜ ๋ฏธ์…˜์€ ์ด๋Ÿฌํ•œ ๋†€๋ผ์šด ๊ธฐ์ˆ ๋“ค์„ ์–ด๋–ป๊ฒŒ ๋”์šฑ ๋„๋ฉ”์ธ๋ณ„๋กœ ํŠนํ™”์‹œํ‚ค๊ฑฐ๋‚˜, ๋„๋ฉ”์ธ ๋งฅ๋ฝ์—์„œ ์ด๋ฅผ ์‚ฌ์šฉํ•˜๋Š” ๋ฐฉ๋ฒ•์„ ์„ธ์ƒ์— ๊ฐ€๋ฅด์น˜๋Š” ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์ œ๊ฐ€ ๋ณด์—ฌ๋“œ๋ฆฐ ๋กœ๋ด‡์ด ๋ฐ”๋Š˜์„ ์ง‘๋Š” ์˜ˆ์‹œ๋Š” ์‹ค์ œ ์—ฐ๊ตฌ ํ‚คํŠธ์ธ ๋‹ค๋นˆ์น˜ ์—ฐ๊ตฌ ํ‚คํŠธ(Da Vinci Research Kit)๋ฅผ ์‚ฌ์šฉํ•œ ์—ฐ๊ตฌ์ž…๋‹ˆ๋‹ค.
And then using our NVIDIA AI imaging research team to understand how to take medical imaging data and turn that into physically-aware synthetic data, and then start to train this robot different policies by doing that in a simulated environment. And then we publish all of that understanding, what we've learned either in open source but also in our Holoscan reference applications. It's all on GitHub with the recipes, with the data sources, with the models that we've used. So that, again, we're trying to democratize this and show everybody how to use these amazing platforms in a domain-specific way. Harlan Sur

Well, we're just about out of time. Kimberly, thank you for your participation.
๊ทธ๋ฆฌ๊ณ  ์ €ํฌ NVIDIA AI ์ด๋ฏธ์ง• ์—ฐ๊ตฌํŒ€์„ ํ™œ์šฉํ•˜์—ฌ ์˜๋ฃŒ ์ด๋ฏธ์ง• ๋ฐ์ดํ„ฐ๋ฅผ ๋ฌผ๋ฆฌ์  ์ธ์‹์ด ๊ฐ€๋Šฅํ•œ ํ•ฉ์„ฑ ๋ฐ์ดํ„ฐ๋กœ ๋ณ€ํ™˜ํ•˜๋Š” ๋ฐฉ๋ฒ•์„ ์ดํ•ดํ•˜๊ณ , ์‹œ๋ฎฌ๋ ˆ์ด์…˜ ํ™˜๊ฒฝ์—์„œ ์ด๋ฅผ ํ†ตํ•ด ๋กœ๋ด‡์—๊ฒŒ ๋‹ค์–‘ํ•œ ์ •์ฑ…๋“ค์„ ํ›ˆ๋ จ์‹œํ‚ค๊ธฐ ์‹œ์ž‘ํ•ฉ๋‹ˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  ์ €ํฌ๊ฐ€ ํ•™์Šตํ•œ ๋ชจ๋“  ์ดํ•ด์™€ ์ง€์‹์„ ์˜คํ”ˆ์†Œ์Šค๋กœ ๋ฐœํ‘œํ•˜๊ฑฐ๋‚˜ Holoscan ์ฐธ์กฐ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์—์„œ๋„ ๊ณต๊ฐœํ•ฉ๋‹ˆ๋‹ค. ๋ ˆ์‹œํ”ผ, ๋ฐ์ดํ„ฐ ์†Œ์Šค, ์ €ํฌ๊ฐ€ ์‚ฌ์šฉํ•œ ๋ชจ๋ธ๋“ค๊ณผ ํ•จ๊ป˜ ๋ชจ๋“  ๊ฒƒ์ด GitHub์— ์˜ฌ๋ผ๊ฐ€ ์žˆ์Šต๋‹ˆ๋‹ค. ๋‹ค์‹œ ๋งํ•ด, ์ €ํฌ๋Š” ์ด๋ฅผ ๋ฏผ์ฃผํ™”ํ•˜๊ณ  ์ด๋Ÿฌํ•œ ๋†€๋ผ์šด ํ”Œ๋žซํผ๋“ค์„ ๋„๋ฉ”์ธ๋ณ„๋กœ ํŠนํ™”๋œ ๋ฐฉ์‹์œผ๋กœ ์‚ฌ์šฉํ•˜๋Š” ๋ฐฉ๋ฒ•์„ ๋ชจ๋“  ์‚ฌ๋žŒ์—๊ฒŒ ๋ณด์—ฌ์ฃผ๋ ค๊ณ  ๋…ธ๋ ฅํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.

ํ• ๋ž€ ์„œ(Harlan Sur)

์‹œ๊ฐ„์ด ๊ฑฐ์˜ ๋‹ค ๋˜์—ˆ๋„ค์š”. ํ‚ด๋ฒŒ๋ฆฌ, ์ฐธ์—ฌํ•ด ์ฃผ์…”์„œ ๊ฐ์‚ฌํ•ฉ๋‹ˆ๋‹ค.
We look forward to monitoring the execution of the team. Kimberly Powell

Thank you so much. Harlan Sur

Thank you very much. Kimberly Powell

Thank you.
ํŒ€์˜ ์‹คํ–‰ ๊ณผ์ •์„ ์ง€์ผœ๋ณด๊ธฐ๋ฅผ ๊ธฐ๋Œ€ํ•ฉ๋‹ˆ๋‹ค. ํ‚ด๋ฒŒ๋ฆฌ ํŒŒ์›”

์ •๋ง ๊ฐ์‚ฌํ•ฉ๋‹ˆ๋‹ค. ํ• ๋ž€ ์„œ

๋Œ€๋‹จํžˆ ๊ฐ์‚ฌํ•ฉ๋‹ˆ๋‹ค. ํ‚ด๋ฒŒ๋ฆฌ ํŒŒ์›”

๊ฐ์‚ฌํ•ฉ๋‹ˆ๋‹ค.

๐Ÿ“Œ ์š”์•ฝ

โ€ข NVIDIA ์˜๋ฃŒ ์‚ฌ์—…์˜ ์ฃผ์š” ํฌ์ธํŠธ๋ฅผ ๋‹ค์Œ๊ณผ ๊ฐ™์ด ์š”์•ฝํ•ฉ๋‹ˆ๋‹ค:

โ€ข ๋น„์ฆˆ๋‹ˆ์Šค ๋ชจ๋ธ:
- GPU ๊ธฐ๋ฐ˜ ํด๋ผ์šฐ๋“œ ๋ฐ ์˜จํ”„๋ ˜ ํ•˜์ด๋ธŒ๋ฆฌ๋“œ ๋ฐฉ์‹ ์ œ๊ณต
- NVIDIA AI Enterprise ๊ตฌ๋… ๋ชจ๋ธ: GPU๋‹น ์—ฐ๊ฐ„ $4,500 (์‹œ๊ฐ„๋‹น $1)
- ํ•˜๋“œ์›จ์–ด, ์†Œํ”„ํŠธ์›จ์–ด, ํ”„๋ ˆ์ž„์›Œํฌ๋ฅผ ํฌํ•จํ•œ ํ’€์Šคํƒ ์†”๋ฃจ์…˜ ์ œ๊ณต

โ€ข ์ฃผ์š” ์ „๋žต์  ๋ฐฉํ–ฅ:
- Accenture, Deloitte ๋“ฑ ๊ธ€๋กœ๋ฒŒ SI ์—…์ฒด๋“ค๊ณผ์˜ ํŒŒํŠธ๋„ˆ์‹ญ ๊ฐ•ํ™”
- AI ์—์ด์ „ํŠธ ๊ฐœ๋ฐœ์„ ์œ„ํ•œ ๋ธ”๋ฃจํ”„๋ฆฐํŠธ ์ œ๊ณต
- ์˜๋ฃŒ ๋ถ„์•ผ ํŠนํ™” ๋ฌผ๋ฆฌ์  AI ๋ฐ ๋กœ๋ด‡ ์‹œ๋ฎฌ๋ ˆ์ด์…˜ ๊ธฐ์ˆ  ๊ฐœ๋ฐœ

โ€ข ๋ฆฌ์Šคํฌ/๊ธฐํšŒ ์š”์ธ:
- ๋ฐ์ดํ„ฐ ์ฃผ๊ถŒ ๋ฐ ๊ตญ๊ฐ€๋ณ„ ๊ทœ์ œ์— ๋”ฐ๋ฅธ ํด๋ผ์šฐ๋“œ/์˜จํ”„๋ ˜ ์œ ์—ฐ์„ฑ ํ•„์š”
- ์˜๋ฃŒ ์ธ๋ ฅ ๋ถ€์กฑ ๋ฐ ๋น„์šฉ ์ƒ์Šน ๋ฌธ์ œ ํ•ด๊ฒฐ์„ ์œ„ํ•œ AI ์†”๋ฃจ์…˜ ์ˆ˜์š” ์ฆ๊ฐ€