輝達 (NVDA) 2020 Q2 法說會逐字稿

Good afternoon. My name is Christina, and I will be your conference operator today. Welcome to NVIDIA's financial results conference call. (Operator Instructions)

午安.我叫克里斯蒂娜，今天我將擔任您的會議主持人。歡迎參加 NVIDIA 財務績效電話會議。（操作員指示）

I'll now turn the call over to Simona Jankowski from Investor Relations to begin your conference.

現在我將把電話轉給投資者關係部的 Simona Jankowski，開始您的會議。

Thank you. Good afternoon, everyone, and welcome to NVIDIA's Conference Call for the Second Quarter of Fiscal 2020. With me on the call today from NVIDIA are Jen-Hsun Huang, President and Chief Executive Officer; and Colette Kress, Executive Vice President and Chief Financial Officer.

謝謝。大家下午好，歡迎參加NVIDIA 2020財年第二季電話會議。今天與我一起參加電話會議的還有 NVIDIA 總裁兼執行長黃仁勳；以及執行副總裁兼財務長 Colette Kress。

I'd like to remind you that our call is being webcast live on NVIDIA's Investor Relations website. The webcast will be available for replay until the conference call to discuss our financial results for the third quarter of fiscal 2020. The content of today's call is NVIDIA's property. It can't be reproduced or transcribed without our prior written consent.

我想提醒您，我們的電話會議正在 NVIDIA 的投資者關係網站上進行網路直播。網路直播將可重播，直至電話會議討論我們 2020 財年第三季的財務業績。今天電話會議的內容屬於 NVIDIA 的財產。未經我們事先書面同意，不得複製或轉錄。

During this call, we may make forward-looking statements based on current expectations. These are subject to a number of significant risks and uncertainties, and our actual results may differ materially. For a discussion of factors that could affect our future financial results and business, please refer to the disclosure in today's earnings release, our most recent Form 10-K and 10-Q and the reports that we may file on Form 8-K with the Securities and Exchange Commission. All our statements are made as of today, August 15, 2019, based on information currently available to us. Except as required by law, we assume no obligation to update any such statements.

在本次電話會議中，我們可能會根據目前預期做出前瞻性陳述。這些都受到許多重大風險和不確定性的影響，我們的實際結果可能會有重大差異。有關可能影響我們未來財務表現和業務的因素的討論，請參閱今天的收益報告中的披露內容、我們最新的 10-K 和 10-Q 表格以及我們可能向美國證券交易委員會提交的 8-K 表格報告。我們所有的聲明都是根據我們目前掌握的資訊截至 2019 年 8 月 15 日做出的。除法律要求外，我們不承擔更新任何此類聲明的義務。

During this call, we will discuss non-GAAP financial measures. You can find a reconciliation of these non-GAAP financial measures to GAAP financial measures in our CFO commentary, which is posted on our website.

在本次電話會議中，我們將討論非公認會計準則財務指標。您可以在我們網站上發布的 CFO 評論中找到這些非 GAAP 財務指標與 GAAP 財務指標的對帳表。

With that, let me turn the call over to Colette.

說完這些，讓我把電話轉給科萊特。

Thanks, Simona. Q2 revenue was $2.58 billion, in line with our outlook, down 17% year-on-year and up 16% sequentially.

謝謝，西蒙娜。第二季營收為 25.8 億美元，符合我們的預期，年減 17%，季增 16%。

Starting with our gaming business. Revenue of $1.31 billion was down 27% year-on-year and up 24% sequentially. We are pleased with the strong sequential growth in the quarter when we launched our RTX SUPER lineup for desktop gamers, wrapped up our greatest ever number of gaming laptops and launched our new RTX studio laptops for creators. In July, we unveiled 3 GeForce RTX SUPER GPUs, delivering the best-in-class gaming performance and power efficiency and real-time ray tracing for both current and next-generation games. These GPUs delivered a performance boost of up to 24% from our initial Turing GPUs launched a year earlier. The SUPER lineup strengthens our leadership in the high end of the market, and the response has been great. We look forward to delighting gamers with the best performance in ray tracing as we get into the back to school and holiday shopping seasons.

從我們的遊戲業務開始。營收為 13.1 億美元，年減 27%，季增 24%。我們對本季強勁的連續成長感到滿意，我們為桌上遊戲玩家推出了 RTX SUPER 系列，推出了有史以來數量最多的遊戲筆記型電腦，並為創作者推出了新的 RTX 工作室筆記型電腦。7 月，我們推出了 3 款 GeForce RTX SUPER GPU，為當前和下一代遊戲提供一流的遊戲性能和能源效率以及即時光線追蹤。與我們一年前推出的初始 Turing GPU 相比，這些 GPU 的效能提升了高達 24%。SUPER系列產品鞏固了我們在高端市場的領導地位，並且反應非常好。隨著開學季和假期購物季的到來，我們期待以光線追蹤的最佳性能取悅遊戲玩家。

Ray tracing is taking the gaming industry by storm and have quickly come to define the modern era of computer graphics. A growing number of blockbuster AAA titles have announced support for NVIDIA RTX ray tracing, including Call of Duty: Modern Warfare, Super Punk 2077 (sic) [Cyberpunk 2077], Watch Dogs: Legion and Wolfenstein: Youngblood. Excitement around these titles is tremendous. GameSpot called Cyberpunk one of the most anticipated games of the decade. NVIDIA GeForce RTX are the only graphic cards in the market with hardware support for ray tracing. They deliver a 2 to 3x performance speed up over GPUs without a dedicated ray tracing core.

光線追蹤正在席捲遊戲產業，並迅速定義了現代電腦圖形時代。越來越多的 AAA 大作宣布支持 NVIDIA RTX 光線追踪，其中包括《決勝時刻：現代戰爭》、《超級朋克 2077》（原文如此）[賽博朋克 2077]、《看門狗：軍團》和《德軍總部：新血統》。這些頭銜引發了巨大的興奮。GameSpot 稱 Cyber​​punk 是十年來最受期待的遊戲之一。NVIDIA GeForce RTX 是市場上唯一支援光線追蹤硬體的顯示卡。與沒有專用光線追蹤核心的 GPU 相比，它們的效能提高了 2 到 3 倍。

The laptop business continues to be a standout growth driver as OEMs are ramping a record 100-plus gaming laptop models ahead of the back to school and holiday season. The combination of our energy-efficient Turing architecture and Max-Q technology enables beautifully crafted thin and light form factors that can deliver the performance of high-end gaming desktop or our next-generation console.

筆記型電腦業務繼續成為突出的成長動力，因為原始設備製造商在返校和假期來臨之前創紀錄地推出了 100 多款遊戲筆記型電腦。我們的節能圖靈架構與 Max-Q 技術的結合，打造出精美的輕薄外形，可提供高階遊戲桌上型電腦或下一代遊戲機的效能。

At Computex in May, we unveiled NVIDIA RTX Studio laptops, a new design artist platform that extends our reach to the large, underserved market of creators. In the age of YouTube, creators and freelancers are rapidly growing population, but they have traditionally not had access to professional-grade workstations through online and retail channels. RTX Studio laptops are designed to meet their increasing complex workflows such as photorealistic ray tracing, AI image enhancement and ultra high-resolution video. Powered by our RTX GPUs and optimized software, RTX Studio laptops deliver performance that's up to 7x faster than that of the MacBook Pro. A total of 27 RTX Studio models have been announced by major OEMs.

在 5 月的台北國際電腦展上，我們推出了 NVIDIA RTX Studio 筆記型電腦，這是一個全新的設計藝術家平台，將我們的業務範圍擴展到龐大且服務不足的創作者市場。在 YouTube 時代，創作者和自由工作者的數量正在迅速增長，但他們傳統上無法透過線上和零售管道獲得專業級工作站。RTX Studio 筆記型電腦旨在滿足日益複雜的工作流程，例如逼真的光線追蹤、AI 影像增強和超高解析度視訊。RTX Studio 筆記型電腦採用我們的 RTX GPU 和最佳化軟體，效能比 MacBook Pro 快 7 倍。各大 OEM 廠商共發表了 27 款 RTX Studio 車型。

Sequential growth also benefited from the production ramp of the 2 new models of Nintendo Switch gaming console. We are expecting our console business to remain strong in Q3 before the seasonal production slowdown in Q4 when console-related revenue is expected to be fairly minimal, similar to last year.

連續成長也得益於兩款新型 Nintendo Switch 遊戲機的產量提升。我們預計，我們的遊戲機業務將在第三季保持強勁，但在第四季季節性生產放緩之前，預計遊戲機相關收入將相當少，與去年類似。

Moving to data center. Revenue was $655 million, down 14% year-on-year and up 3% sequentially. In the vertical industries portion of the business, expanding AI workload drove sequential and year-over-year growth. In hyperscale portion, we continue to be impacted by relatively weak overall spending at a handful of CPU -- CSPs. Sales of NVIDIA GPUs for use in the cloud were solid. While sales of internal hyperscale use were muted, the engineering focus on AI is growing.

移至資料中心。營收為 6.55 億美元，年減 14%，季增 3%。在垂直產業業務部分，不斷擴大的人工智慧工作量推動了連續和同比成長。在超大規模部分，我們持續受到少數 CPU-CSP 整體支出相對較弱的影響。用於雲端的 NVIDIA GPU 的銷售表現穩健。雖然內部超大規模使用的銷售不景氣，但工程部門對人工智慧的關注度卻日益增長。

Let me give some color on each of these areas. We are building a broad base of customers across multiple industries as they adopt NVIDIA's platforms to harness the power of AI. Public sectors, higher education and financial services were among the key verticals driving growth this quarter. In addition, we won Lighthouse account deals in important industries that are on the cusp of being transformed by AI. For example, in retail, Walmart is using NVIDIA GPUs to run some of its product demand forecasting models, slashing the time to do so in just 4 hours from several weeks on CPUs. By accelerating its data science workflow, Walmart can improve its algorithms, reduce development cycles and test new features.

讓我對每個領域進行一些說明。隨著各行各業採用 NVIDIA 平台來發揮 AI 的力量，我們正在建立廣泛的客戶基礎。公共部門、高等教育和金融服務是本季推動成長的主要垂直產業。此外，我們也贏得了即將被人工智慧轉型的重要產業的燈塔帳戶交易。例如，在零售業，沃爾瑪正在使用 NVIDIA GPU 來運行其部分產品需求預測模型，將運行時間從 CPU 上的數週縮短至僅需 4 小時。透過加速其資料科學工作流程，沃爾瑪可以改進其演算法、縮短開發週期並測試新功能。

Earlier this week, we announced breakthroughs for the fastest training and inference of the state-of-the-art model for natural language process understanding called BERT, or Bidirectional Encoder Representations of -- from Transformers, a breakthrough AI language model that achieves a deeper sense of language, context and meaning. This can enable mere human comprehension in real-time by chat box, intelligent personal assistants and search engines. We are working with Microsoft as an early adopter of these advances.

本週早些時候，我們宣布了用於自然語言過程理解的最快訓練和推理的突破，該模型稱為 BERT，即 Transformers 的雙向編碼器表示，這是一種突破性的人工智慧語言模型，可以更深入地理解語言、上下文和含義。這可以使人類透過聊天框、智慧個人助理和搜尋引擎即時理解。我們正在與微軟合作，作為這些進步的早期採用者。

AI computing leadership is a high priority for NVIDIA. Last month, we set records for training deep learning neural network models on the latest MLPerf benchmarks, particularly in the most demanding areas. In just 7 months, we have achieved up to 80% speed-ups enabled by new algorithms and software optimizations across the full stack while using the same hardware. This is a direct result of the productive programming environment and flexibility of CUDA.

人工智慧運算領導地位是 NVIDIA 的首要任務。上個月，我們在最新的 MLPerf 基準上創下了訓練深度學習神經網路模型的記錄，尤其是在要求最嚴格的領域。在短短 7 個月內，我們在使用相同硬體的情況下，透過新演算法和軟體優化，實現了整個堆疊高達 80% 的加速。這是 CUDA 高效程式設計環境和靈活性的直接結果。

Delivering AI at scale isn't just about silicon. It's about optimizing across the entire high-performance computing system. In fact, the NVIDIA AI platform is getting progressively faster. Every month, we publish new optimization and performance improvements to CUDA-X AI libraries, supporting every AI framework and development environment. All in, our ecosystem of developers is now 1.4 million strong.

大規模交付人工智慧不僅涉及矽。它是關於整個高效能運算系統的最佳化。事實上，NVIDIA AI 平台的速度正在不斷加快。我們每月都會發布針對 CUDA-X AI 庫的新優化和效能改進，支援每個 AI 框架和開發環境。總而言之，我們的開發者生態系統現已擁有 140 萬名開發者。

In setting these MLPerf records, we leveraged our new DGX SuperPOD AI supercomputer, demonstrating that leadership in AI research demands leadership in computing infrastructure. This system debuted in June at #22 on the TOP500 list of the world's fastest supercomputers at the annual International Supercomputing Conference. Used to meet the massive demand for autonomous vehicle development program, it is powered by more than 1,500 NVIDIA V100 Tensor Core GPUs linked with Mellanox interconnects. We've made DGX SuperPOD available commercially to customers, essentially providing them with the turnkey supercomputer that they can assemble in weeks rather than months. It is roughly 400x smaller in size than other similarly performing TOP500 systems, which are built from thousands of servers.

在創造這些 MLPerf 記錄時，我們利用了新的 DGX SuperPOD AI 超級計算機，證明了 AI 研究的領導地位需要計算基礎設施的領導地位。該系統於 6 月在年度國際超級運算大會上首次亮相，位列全球最快超級電腦 TOP500 榜單第 22 名。它用於滿足自動駕駛汽車開發計劃的巨大需求，由 1,500 多個透過 Mellanox 互連的 NVIDIA V100 Tensor Core GPU 提供支援。我們已將 DGX SuperPOD 商用提供給客戶，本質上是為他們提供可以在幾週而不是幾個月內組裝的交鑰匙超級電腦。與其他由數千台伺服器建置的具有類似效能的 TOP500 系統相比，它的規模大約小 400 倍。

Also at the conference, we announced that by next year's end, we will make available to the ARM ecosystem NVIDIA's full stack of AI and HPC software, which accelerates more than 600 HPC applications and all AI frameworks. With this announcement, NVIDIA will accelerate all major CPU architectures, including x86, POWER and ARM.

在此次大會上，我們也宣布，到明年年底，我們將向 ARM 生態系統提供 NVIDIA 的全套 AI 和 HPC 軟體，以加速 600 多個 HPC 應用程式和所有 AI 框架。透過此次發布，NVIDIA 將加速所有主流 CPU 架構，包括 x86、POWER 和 ARM。

Lastly, regarding our pending acquisition of Mellanox, we have received regulatory approval in the U.S. and are engaged with regulators in Europe and China. The approval process is progressing as expected, and we continue to work toward closing the deal by the end of this calendar year.

最後，關於我們即將收購 Mellanox 的交易，我們已獲得美國監管部門的批准，並正在與歐洲和中國的監管機構接洽。審批流程正在按預期進行，我們將繼續努力在今年年底前完成交易。

Moving to pro visualization. Revenue reached $291 million, up 4% from our prior year and up 9% sequentially. Year-on-year and sequential growth was led by record revenue for mobile workstations with strong demand for new thin and light form factors. We had a great showing at SIGGRAPH, the computer graphics industry's biggest annual conference held in Los Angeles. Our researchers won several Best in Show awards. In just a year since the launch of RTX ray tracing, over 40 design and creative applications with RTX technology had been announced by leading software vendors, including Adobe, Autodesk and Dassault systems and many others. NVIDIA RTX technology has reinvigorated the computer graphics industry by enabling researchers and developers to take a leap in photorealistic rendering, augmented reality and virtual reality.

轉向專業視覺化。營收達 2.91 億美元，比上年成長 4%，比上一季成長 9%。年比和環比成長主要得益於行動工作站創紀錄的收入以及市場對新型輕薄外形的強勁需求。我們在洛杉磯舉辦的電腦圖形產業最大的年度會議 SIGGRAPH 上有著出色的表現。我們的研究人員贏得了多項最佳展示獎。RTX 光線追蹤推出僅一年時間，Adobe、Autodesk、達梭系統等多家領先軟體供應商就已宣布推出 40 多個採用 RTX 技術的設計和創意應用程式。NVIDIA RTX 技術使研究人員和開發人員能夠在照片層級渲染、擴增實境和虛擬實境領域取得飛躍，從而為電腦圖形產業注入了新的活力。

Finally, turning to automotive. Q2 revenue was $209 million, up 30% from a year ago and up 26% sequentially. This reflects growing adoption of next-generation AI cockpit solutions and autonomous vehicle development projects, including 1 particularly sizable development services transaction that was recognized in the quarter. In addition, in June, we announced a new partnership with the Volvo Group to develop AI and autonomous trucks utilizing NVIDIA's end-to-end AI platform for training, simulation and in-vehicle computing. The strategic partnership will enable Volvo Group to develop a wide range of autonomous driving solutions for freight transport, recycling collection, public transport, construction, mining, forestry and more. This collaboration is a great validation of our long-held position that every vehicle, not just cars but also trucks, shuttles, business, taxis and many others, will have autonomous capability 1 day.

最後，轉向汽車。第二季營收為 2.09 億美元，較去年同期成長 30%，較上一季成長 26%。這反映了下一代人工智慧駕駛艙解決方案和自動駕駛汽車開發項目的日益普及，其中包括本季確認的一項特別大規模的開發服務交易。此外，今年 6 月，我們宣布與沃爾沃集團建立新的合作夥伴關係，利用 NVIDIA 的端到端 AI 平台進行訓練、模擬和車載運算，開發 AI 和自動駕駛卡車。此策略合作夥伴關係將使沃爾沃集團能夠為貨運、回收收集、公共交通、建築、採礦、林業等領域開發廣泛的自動駕駛解決方案。這項合作大大驗證了我們長期以來的立場：未來某一天，每輛車，不僅是汽車，還有卡車、接駁車、商務車、計程車等等，都將具備自動駕駛能力。

Autonomous features can bring enormous value to the trucking industry, in particular as the demand of online shopping put ever greater stress on the world's transport systems. Expectations for overnight or same-day deliveries create challenges that can only be met by autonomous trucks, which can operate 24 hours a day. To help address these needs, NVIDIA has created an end-to-end platform for autonomous vehicles from AI computing infrastructure to large-scale simulation to in-car computing. Multiple customers from OEMs like Mercedes-Benz, Toyota and Volvo to Tier 1s like Bosch, Continental and ZF are already onboard. We see this as a $30 billion addressable market by 2025.

自動駕駛功能可以為卡車運輸業帶來巨大的價值，特別是當網上購物的需求給世界運輸系統帶來越來越大的壓力時。隔夜或當日送達的期望帶來了挑戰，只有能夠全天 24 小時運行的自動駕駛卡車才能應對。為了幫助滿足這些需求，NVIDIA 為自動駕駛汽車創建了一個端到端平台，從 AI 運算基礎設施到大規模模擬再到車載運算。眾多客戶，從梅賽德斯-奔馳、豐田和沃爾沃等原始設備製造商到博世、大陸和 ZF 等一級供應商，都已加入。我們認為到 2025 年，這個潛在市場規模將達到 300 億美元。

Moving to the rest of the P&L. Q2 GAAP gross margins was 59.8% and non-GAAP was 60.1%, up sequentially, reflecting higher automotive development services, a favorable mix in gaming and lower component cost. GAAP operating expenses were $970 million, and non-GAAP operating expenses were $749 million, up 19% and 8% year-on-year, respectively. We remain on track for high single-digit OpEx growth in fiscal 2020 while continuing to invest in the key platforms driving our long-term growth, namely graphics, AI and self-driving cars. GAAP EPS was $0.90, down 49% from a year earlier. Non-GAAP EPS was $1.24, down 36% from a year ago.

轉到損益表的其餘部分。第二季 GAAP 毛利率為 59.8%，非 GAAP 毛利率為 60.1%，較上季成長，反映了汽車開發服務的增加、遊戲業務的良好組合以及零件成本的降低。GAAP營運費用為9.7億美元，非GAAP營運費用為7.49億美元，分別較去年同期成長19%和8%。我們仍有望在 2020 財年實現高個位數營運支出成長，同時繼續投資於推動我們長期成長的關鍵平台，即圖形、人工智慧和自動駕駛汽車。GAAP EPS 為 0.90 美元，較去年同期下降 49%。非公認會計準則每股收益為 1.24 美元，較去年同期下降 36%。

With that, let me turn to the outlook for the third quarter of fiscal 2020. We expect revenue to be $2.9 billion, plus or minus 2%. GAAP and non-GAAP gross margins are expected to be 62% and 62.5%, respectively, plus or minus 50 basis points. GAAP and non-GAAP operating expenses are expected to be approximately $980 million and $765 million, respectively. GAAP and non-GAAP OI&E are both expected to be income of approximately $25 million. GAAP and non-GAAP tax rates are both expected to be 10%, plus or minus 1%, excluding discrete items. Capital expenditures are expected to be approximately $100 million to $120 million. Further financial details are included in the CFO commentary and other information available on our IR website.

接下來，讓我來談談 2020 財年第三季的展望。我們預計營收為 29 億美元，上下浮動 2%。預計 GAAP 和非 GAAP 毛利率分別為 62% 和 62.5%，上下浮動 50 個基點。預計 GAAP 和非 GAAP 營運費用分別約為 9.8 億美元和 7.65 億美元。預計 GAAP 和非 GAAP OI&E 收入均為約 2,500 萬美元。預計 GAAP 和非 GAAP 稅率均為 10%，上下浮動 1%，不包括單一項目。預計資本支出約1億至1.2億美元。進一步的財務細節包含在財務長評論和我們 IR 網站上提供的其他資訊中。

In closing, let me highlight upcoming events for the financial community. We will be at the Jefferies conference, hardware and communications infrastructure summit, on August 27 and at the Citi Global Technology Conference on September 25.

最後，讓我重點介紹一下金融界即將發生的事件。我們將於 8 月 27 日參加 Jefferies 會議、硬體和通訊基礎設施高峰會，並於 9 月 25 日參加花旗全球技術會議。

With that, we will now open the call for questions. Operator, would you please poll for the questions?

現在，我們將開始提問。接線生，請您調查問題好嗎？

(Operator Instructions) And your first question comes from the line of C.J. Muse with Evercore.

（操作員指示）您的第一個問題來自 Evercore 的 C.J. Muse。

I guess first question on gaming, how should we think about your outlook into the October quarter vis-à-vis kind of normal seasonality? How are you thinking about Switch within that? And considering now that you have full Turing lineup as well as content truly coming to the forefront here, how do you think about trends beyond the October quarter?

我想第一個問題是關於遊戲的，相對於正常的季節性，我們應該如何看待您對 10 月季度的展望？如何看待 Switch 呢？考慮到現在您已經擁有完整的圖靈產品線以及真正佔據領先地位的內容，您如何看待十月季度之後的趨勢？

Sure. Colette, why don't you take the Switch question? And then I'll take the rest of the RTX questions.

當然。科萊特，為什麼不回答 Switch 的問題呢？然後我會回答剩下的 RTX 問題。

Sure. From a gaming perspective, the overall Switch or the overall console business definitely is a seasonal business. We usually expect to see production ramping in Q2 and in Q3, with it coming down likely in Q4. So you should see Switch to be a portion definitely of our gaming business in Q3.

當然。從遊戲角度來看，整體 Switch 或整個遊戲機業務絕對是季節性業務。我們通常預期第二季和第三季產量會上升，而第四季產量可能會下降。因此，您應該會看到 Switch 肯定會成為我們第三季遊戲業務的一部分。

Yes. C.J., thanks for the question. RTX, as you know, is -- first of all, RTX is doing great. I think we've put all the pieces in place to bring ray tracing into the future of games. The number of games, the blockbuster games that adopted RTX is really snowballing. We announced several -- 6 games in the last couple of months. There's going to be some exciting announcements next week at gamescom. It's pretty clear now the future of gaming will include ray tracing. The number of software developers that create -- with creative tools that adopted RTX is really quite spectacular. We now have 40 -- over 40 ISV tools that was announced at SIGGRAPH that have accelerated ray tracing and video editing. And some of the applications' amazing AI capabilities for image optimization enhancement support RTX. And so looking forward, this is what I expect. I expect that ray tracing is going to drive a reinvigoration of gaming graphics. I expect that the over 100 laptops that we have RTX designed -- RTX GPUs designed into is going to contribute our growth. Notebook gaming is one of the fastest-growing segments of the gaming platform world. The number of notebooks that are able to game is only a few percent, so it's extremely underexposed. And yet, we know that gamers are -- like the rest of us, they like thin and light notebooks, but they like it to be able to run powerful games. And so this is an area that has grown significantly for us year-over-year, and we're expecting it to grow through the end of the -- through the second half and through next year.

是的。C.J.，謝謝你的提問。如您所知，RTX——首先，RTX 表現出色。我認為我們已經做好了一切準備，將光線追蹤帶入遊戲的未來。採用 RTX 的遊戲數量和大片遊戲數量確實在滾雪球般增長。過去幾個月我們宣布了幾款遊戲——6 款。下週的科隆遊戲展將會發布一些令人興奮的公告。現在很明顯，遊戲的未來將包括光線追蹤。採用 RTX 進行創作的軟體開發人員的數量確實相當驚人。我們現在有 40 多個 ISV 工具在 SIGGRAPH 上發布，它們可以加速光線追蹤和影片編輯。而一些應用程式令人驚嘆的影像優化增強AI功能也支援RTX。展望未來，這就是我所期望的。我預計光線追蹤將會推動遊戲圖形的復興。我預計，我們設計的 100 多台採用 RTX 的筆記型電腦（RTX GPU）將促進我們的成長。筆記型遊戲是遊戲平台領域成長最快的領域之一。能夠玩遊戲的筆記型電腦數量只有百分之幾，因此曝光度極低。然而，我們知道遊戲玩家和我們其他人一樣，他們喜歡輕薄的筆記型電腦，但他們喜歡它能夠運行強大的遊戲。因此，這是我們同比成長顯著的領域，我們預計它將在今年下半年和明年繼續成長。

And one of the things that's really exciting is our RTX Studio line that we introduced recently. We observed, and through our discussions with the PC industry, that the creatives are really underexposed and underserved by the latest technologies. And they want notebooks and they want PCs that have powerful graphics. They use it for 3D content creation and high definition video editing and image optimization and things like that. And we introduced a brand-new line of computers that we call RTX Studio. Now the OEMs were so excited about it. And at SIGGRAPH, we now have 27 different laptops shipping and more coming. And so I think RTX is really geared for growth. We have great games coming. We got the SUPER line of GPUs. We have all of our notebooks that were designed into that we're ramping and, of course, the new RTX Studio line. And so I expect this to be a growth market for us.

其中真正令人興奮的是我們最近推出的 RTX Studio 系列。透過與 PC 產業的討論，我們觀察到創意人員確實沒有得到充分的關注，也沒有得到最新技術的充分服務。他們想要筆記型電腦和具有強大圖形功能的個人電腦。他們使用它來創建 3D 內容、編輯高清影片、優化圖像等等。我們推出了名為 RTX Studio 的全新電腦系列。現在 OEM 對此感到非常興奮。在 SIGGRAPH 上，我們現在已經推出了 27 款不同的筆記型電腦，而且還會有更多。因此我認為 RTX 確實適合成長。我們即將迎來精彩的比賽。我們擁有 SUPER 系列 GPU。我們所有的筆記型電腦都是為此而設計的，當然還有新的 RTX Studio 系列。因此我預計這對我們來說將是一個成長市場。

Very helpful. If I could follow up on the data center side, perhaps you can speak directly just to the hyperscale side, both internal and cloud, and whether you're seeing any green shoots, any signs of life there and how you're thinking about what that rate of recovery could look like over time.

非常有幫助。如果我可以跟進資料中心方面的問題，也許您可以直接談論超大規模方面，包括內部和雲端，以及您是否看到任何綠芽、任何生命跡象，以及您如何看待這種復甦速度隨著時間的推移會是什麼樣子。

With the exception of a couple of hyperscalers, C.J., I would -- we're seeing broad-based growth in data centers. In the area of training, the thing that's really exciting everybody, and everybody is racing towards, is training these large gigantic natural language understanding models, language models. The transformer model that was introduced by Google, called BERT, has since been enhanced into XLned and RoBERTa and, gosh, so many different, GP2, and Microsoft's MASS. And there's so many different versions of these language models. And in the AI, NLU, natural language understanding, is one of the most important areas that everybody's racing to go to. And so these models are really, really large. It's over 1,000x larger than image models that we're training just a few years ago, and they're just gigantic models. It's one of the reasons why we built the DGX SuperPOD so that we could train these gigantic models in a reasonable amount of time. The second area -- so that's training in the hyperscalers.

除了幾個超大規模企業外，C.J.，我們看到資料中心正在廣泛成長。在訓練領域，真正讓每個人興奮、每個人都在競相實現的事情是訓練這些大型的自然語言理解模型、語言模型。谷歌推出的轉換器模型稱為 BERT，後來被增強為 XLned 和 RoBERTa，天哪，還有這麼多不同的東西，還有 GP2 和微軟的 MASS。這些語言模型有很多不同的版本。在人工智慧領域，NLU（自然語言理解）是每個人都競相進入的最重要的領域之一。所以這些模型真的非常大。它比我們幾年前訓練的圖像模型大 1,000 多倍，而且它們只是巨大的模型。這就是我們建立 DGX SuperPOD 的原因之一，以便我們能夠在合理的時間內訓練這些龐大的模型。第二個領域－即超大規模資料中心的培訓。

The second area where we're seeing enormous amounts of activity has to do with trying to put these conversational AI models into services so that they could be interactive and in real time. Whereas photo tagging and photo enhancement is something that you could put off-line and you could do that while you have excess capacity when it's off of the most busy time of the day. You can't do that with language and conversational AI. You better to respond to the person in real time. And so the performance that's required is significant. But more importantly, the number of models necessary for conversational AI from speech recognition to language understanding to recommendation systems to text-to-speech to wave synthesis, these 5, 6, 7 models have to be processed in real time -- in series and in real time so that you can have a reasonable conversation with the AI agent.

我們看到大量活動的第二個領域是嘗試將這些對話式人工智慧模型放入服務中，以便它們能夠進行互動和即時。而照片標記和照片增強功能則可以離線進行，並且您可以在一天中最繁忙的時間之外，利用剩餘容量進行這些操作。你無法透過語言和對話式人工智慧做到這一點。你最好即時回覆對方。因此，所需的性能非常重要。但更重要的是，對話式人工智慧所需的模型數量從語音識別到語言理解到推薦系統到文字轉語音到波形合成，這 5、6、7 個模型必須即時處理——串行、即時地處理，這樣你才能與人工智慧代理進行合理的對話。

And so these type of activities is really driving interest and activity at all of the hyperscalers. My expectation is that this is going to continue to be a big growth opportunity for us. But more importantly, in addition to that, we're seeing that AI is -- the wave of AI is going from the cloud to the enterprise to the edge and all the way out to the autonomous systems. The place where we're seeing a lot of excitement, and we talked about that in the past and we're seeing growth there, has to do with the vertical industry enterprises that are starting to adopt AI to create new products, whether it's a delivery robot or some kind of a chat bot or the ability to detect fraud in financial services, these applications in vertical industries are really spreading all over the place. There's some over 4,000 AI start-ups around the world. And the way that we engage them is they use our platform to start developing AI in the cloud. And as you know, we're the only AI platform that's available on-prem and in every single cloud. And so they can use our AI platforms for -- in all the clouds, which is driving our cloud computing, external cloud computing growth. And then they can also use it on-prem if their usage really grows significantly. And that's one of the reasons why our Tesla for OEMs and DGX is growing. And so we're seeing broad-based excitement around AI as they use it for their products and new services. And these 4,000, 4,500 start-ups around the world is really driving consumption of that.

因此，這些類型的活動確實激發了所有超大規模企業的興趣和活動。我期望這將繼續成為我們的一個巨大的成長機會。但更重要的是，除此之外，我們看到人工智慧的浪潮正在從雲端蔓延到企業、邊緣，一直到自主系統。我們看到很多令人興奮的事情，我們過去也討論過這個問題，並且我們看到了那裡的增長，這與垂直行業企業開始採用人工智能來創造新產品有關，無論是送貨機器人還是某種聊天機器人，還是檢測金融服務欺詐的能力，這些垂直行業的應用確實遍布各地。全球有超過 4,000 家人工智慧新創公司。我們與他們互動的方式是，他們使用我們的平台開始在雲端開發人工智慧。如您所知，我們是唯一可在本地和所有雲端使用的 AI 平台。因此他們可以在所有雲端中使用我們的 AI 平台，從而推動我們的雲端運算和外部雲端運算的成長。如果他們的使用量確實大幅增長，他們也可以在本地使用它。這就是我們的特斯拉 OEM 和 DGX 不斷增長的原因之一。因此，我們看到大家對人工智慧的廣泛興趣，並將其應用於他們的產品和新服務。全球各地的這 4,000 到 4,500 家新創公司確實推動了這一領域的消費。

And your next question comes from the line of Vivek Arya with Bank of America Merrill Lynch.

您的下一個問題來自美銀美林的 Vivek Arya。

I actually had 2 as well, one quick one for Colette and one for Jensen. Colette, good to see the gross margin recovery getting into October. Is this 62% to 63% range a more sustainable level and perhaps a level you could grow off of as sales get more normalized levels? And then a bigger question is for Jensen. Again, on the data center side, Jensen, when I look back between -- 2015 to 2018, your data center business essentially grew 10x. And then the last year has been a tough one with the slowdown in cloud CapEx and so forth. When do you think your data center starts to grow back on a year-to-year -- on a year-on-year basis? Can that happen sometime -- later this year? And then just longer term, what is the right way to think about this business? Does it go back to prior levels? Does it go at a different phase? This is the one part of the business that I think is toughest for us to model, so any color would be very helpful.

我實際上也吃了 2 顆，一顆給 Colette，另一顆給 Jensen。科萊特，很高興看到十月毛利率回升。62% 到 63% 的範圍是否是一個更可持續的水平，或者隨著銷售額達到更正常的水平，您是否可以從這個水平開始增長？接下來，詹森將面臨一個更大的問題。再說一次，在資料中心方面，Jensen，當我回顧 2015 年至 2018 年期間時，您的資料中心業務基本上增長了 10 倍。去年是艱難的一年，因為雲端運算資本支出等放緩。您認為您的資料中心何時會開始逐年成長？這會在今年晚些時候發生嗎？那麼從長遠來看，如何正確看待這項業務？它會回到之前的水平嗎？它會在不同的階段進行嗎？我認為這是我們業務中最難建模的部分，因此任何顏色都會非常有幫助。

Great. So let me start first with your question, Vivek, regarding gross margins. Yes, thanks for recognizing that we are moving towards our expectations that, over time, we'll continue to see our overall volumes improve. Essentially, our business is normalized. We've reached normalized levels through the last couple of quarters. And this quarter, just very similar to what we will see going forward, is mix is the largest driver, what drives our overall gross margins and our gross margin improvements.

偉大的。那麼，Vivek，讓我先回答你關於毛利率的問題。是的，感謝您認識到我們正在朝著我們的期望邁進，隨著時間的推移，我們將繼續看到我們的整體銷售量不斷提高。本質上，我們的業務已經正常化了。經過過去幾個季度，我們已經達到了正常水平。本季度，與我們未來將看到的非常相似，產品組合是最大的驅動力，推動著我們的整體毛利率和毛利率的提高。

Yes, Vivek, if you look at the last several years, there's no question our data center business has grown a lot. And my expectation is that it's going to grow a lot more, and let me explain to you why. Aside from a couple -- a few of uncontrollable circumstances and the exception of a couple of large customers, the overall trend, the broad-based trend, of our data center business is upward, to the right. And it is growing very nicely. There's a couple of different dynamics that's causing that on first principles to grow. And of course, one of them is as AI is well known now to require accelerated computing, our computing architecture is really ideal for it. AI is not just one network. It's thousands of different types of networks, and these networks are getting more and more complex over time, the amount of data you have to process is enormous. And so like all software programs, you can't predict exactly how the software is going to get programmed. And having a programmable architecture like CUDA and yet optimized for AI like Tensor Cores that we've created is really the ideal architecture.

是的，Vivek，如果你回顧過去幾年，毫無疑問我們的資料中心業務已經成長了很多。我預計它會進一步增長，讓我來向你們解釋一下原因。除了少數不可控的情況和幾個大客戶外，我們的資料中心業務的整體趨勢、廣泛的趨勢是向上、向右的。而且它生長得非常好。有幾種不同的動力導致其在第一原則上發展。當然，其中之一就是眾所周知人工智慧需要加速運算，我們的運算架構非常適合它。人工智慧不僅僅是一個網路。網路類型有數千種，而且這些網路隨著時間的推移變得越來越複雜，需要處理的資料量非常龐大。因此，與所有軟體程式一樣，您無法準確預測軟體將如何進行程式設計。擁有像 CUDA 這樣的可程式架構，同時又針對 AI 進行了最佳化，就像我們創建的 Tensor Cores 一樣，這才是真正理想的架構。

We know also that AI is the most powerful technology force of our time. The ability for machines to learn and write software by itself and write software that no humans can write is pretty extraordinary. And the applications of AI, as you guys are watching yourself, are just spreading in every single industry. And so the way we think about AI is in waves, if you will. The first wave of AI is developing the computer architecture, and that was the first part where -- that's when a lot of people discovered who we are, and we emerged into the world of high-performance computing in AI. The second wave is applying the AI for cloud service providers or hyperscalers. They have a large amount of data. They have a lot of consumer applications. Many of them are not life-critical and so, therefore, the application of an early technology -- early-adoption technology was really viable. And so you saw hyperscalers adopt AI. And the thing that's really exciting for us is beyond recommendations, beyond image enhancement, the area where we believe the most important application for AI is likely conversational AI. Most people talking and asking questions and talking to their mobile devices and looking for something or asking for directions. Instead of having a page of -- a list of options, it responds with an answer that is very likely a good one.

我們也知道，人工智慧是我們這個時代最強大的科技力量。機器學習和自我編寫軟體以及編寫人類無法編寫的軟體的能力非常非凡。如你們所見，人工智慧的應用正在各個產業中蔓延。因此，如果你願意的話，我們對人工智慧的思考方式是波浪式的。人工智慧的第一波浪潮是開發電腦架構，這是第一部分——那時很多人發現了我們是誰，我們進入了人工智慧高效能運算的世界。第二波浪潮是將人工智慧應用於雲端服務供應商或超大規模企業。他們擁有大量數據。他們有很多消費者應用程式。其中許多並不是與生命息息相關的，因此，應用早期技術－早期採用的技術確實是可行的。所以你看到超大規模企業採用了人工智慧。對我們來說真正令人興奮的是，除了推薦和影像增強之外，我們認為人工智慧最重要的應用領域可能是對話式人工智慧。大多數人都在說話、問問題、對著他們的行動裝置說話、尋找某些東西或詢問方向。它不會給出一頁選項列表，而是給出一個很可能是好的答案。

The next phase of AI is what we call vertical industry enterprise AI. And this is where companies are using it not just to accelerate the business process internally, but they're using AI to create new products and services. They could be new medical instruments to IoT-based medical instruments to monitor your health. It could be something related to an application that -- used for financial services for forecasting or for fraud detection. It could be some kind of device that delivers pizza to you, delivery bots. And the combination of IoT and artificial intelligence, for the very first time, you actually have the software capabilities to make use of all of these sensors that you're putting all over the world. And that's the next phase of growth. And it affects companies from large industrials, transportation companies, retailers, you name it. Health care companies, you name it. And so that phase of growth of AI is the phase that we're about to enter into.

人工智慧的下一個階段是我們所謂的垂直產業企業人工智慧。公司不僅使用它來加速內部業務流程，而且還使用人工智慧來創造新產品和服務。它們可能是新的醫療儀器，也是基於物聯網的醫療儀器，用於監測您的健康狀況。它可能與用於金融服務預測或詐欺檢測的應用程式有關。它可能是某種送披薩的設備，送貨機器人。物聯網和人工智慧的結合，首次讓我們擁有了利用遍佈全球的所有感測器的軟體能力。這就是下一階段的成長。它影響到大型工業公司、運輸公司、零售商等等。醫療保健公司，隨便你怎麼說。因此，我們即將進入人工智慧的發展階段。

And then the longer term is an industry that we all know to be extremely large, but it takes time because it's life-critical, and it has to do with transportation. It's a $100 trillion industry. We know it's going to be automated. We know that everything that moves in the future will be autonomous or have autonomous capabilities. And that's just a matter of time before we realize its full potential.

而長期來看，我們都知道這個產業非常龐大，但它需要時間，因為它與生命至關重要，並且與交通運輸有關。這是一個價值100兆美元的產業。我們知道它將實現自動化。我們知道，未來所有運動的物體都將是自主的或具有自主能力。我們充分發揮其潛力只是時間問題。

And so the net of it all is that I believe that AI is the single most powerful technology force of our time, and that's why we're all in on it. And we know that acceleration and accelerated computing is the perfect model for that. And it started in the cloud, but it's going to keep moving out into the edge and through data centers and enterprises and hopefully -- well, eventually, all the way out into autonomous devices and machines in the real world. And so this is a big market, and I'm super enthusiastic about it.

總而言之，我相信人工智慧是我們這個時代最強大的技術力量，這就是我們全力投入其中的原因。我們知道加速和加速運算是實現這一目標的完美模型。它始於雲端，但會不斷向邊緣移動，穿過資料中心和企業，並希望最終能夠進入現實世界中的自主設備和機器。這是一個巨大的市場，我對此非常熱衷。

And your next question comes from the line of Toshiya Hari with Goldman Sachs.

您的下一個問題來自高盛的 Toshiya Hari。

I had 2 as well, one for Jensen and the other for Colette. Jensen, you guys called out inference as a significant contributor to growth in data center last quarter. I think you guys talked about it being a double-digit percentage contributor, curious what you saw from inference in the quarter. And more importantly, if you can talk about the outlook, both near term and long term, as it relates to inference, that'll be helpful. And then secondly, for Colette, just want to double click on the gross margin question. The sequential improvement that you're guiding to is a pretty significant number. So I was just hoping if you can kind of break it down for us in terms of overall volume growth mix dynamics, both between segments and within segments and also to the extent DRAM pricing is impacting that, any color on that will be helpful as well.

我也有 2 個，一個給 Jensen，另一個給 Colette。詹森，你們指出推理是上個季度資料中心成長的重要貢獻者。我認為你們談到了它是一個兩位數百分比的貢獻者，好奇你們從本季的推論中看到了什麼。更重要的是，如果您可以談論與推理相關的近期和長期前景，那將會很有幫助。其次，對於 Colette，我只想雙擊毛利率問題。您所指導的連續改進是一個相當顯著的數字。所以我只是希望您能為我們詳細分析整體銷售成長組合動態，包括各個細分市場之間和細分市場內部的動態，以及 DRAM 定價對此的影響程度，任何關於這方面的詳細資訊都將有所幫助。

Yes, Toshiya, I got to tell you, I'm less good at normal pre -- near-term productions than I am good at thinking about long-term dynamics. But let me talk to you about inference. Our inference business is -- remains robust. It's double digits. It's a large part of our business. And -- but more importantly, the 2 dynamics that I think are near term and that's going to drive growth, number one is interactive conversational AI, interactive conversational AI inference.

是的，Toshiya，我得告訴你，我不太擅長正常的近期生產，但我更擅長思考長期動態。但讓我和你談談推理。我們的推理業務－依然強勁。是兩位數。這是我們業務的很大一部分。而且 — — 但更重要的是，我認為近期將出現兩種推動成長的動力，第一是互動式對話式人工智慧、互動式對話式人工智慧推理。

If you simply ask a chat bot a simple question, where is the closest pizza and you would -- pizza shop, and you would like to have a conversation with this bot, it would have to do speech recognition, it has to understand what it is that you asked about, it has to look it up in a recommender based on the locations you're at, maybe your preferences of styles of pizza and the price ranges that you're interested and how far you're willing to go, to go get it. It has to recommend a pizza shop for you to go to. It has to then translate that from text-to-speech and then into human -- a human understand a voice.

如果你只是問聊天機器人一個簡單的問題，最近的披薩店在哪裡，你想和這個機器人對話，它必須進行語音識別，它必須理解你問的是什麼，它必鬚根據你所在的位置在推薦器中查找，也許是你對披薩風格的偏好和你感興趣的價格範圍以及你願意走多遠去買它。它必須推薦一家披薩店給你去。然後，它必須將文字轉換為語音，然後再轉換為人類能夠理解的聲音。

And those models have to happen in just a few -- ideally, a few hundred milliseconds. Currently, it's not that. And it makes it really hard for these services to be deployed quite broadly and used for all kinds of different applications. And so that's the near-term opportunity, it's interactive conversational AI inference. And you could just imagine every single hyperscaler racing to go make this possible because recently, we had some important breakthroughs in machine learning language models. The BERT model that I mentioned earlier is really, really an important development, and it's caused a large number of derivatives that has improved upon it. And so near-term conversational AI inference.

這些模型必須在短短幾毫秒內完成——理想情況下是幾百毫秒。目前還不是這樣。這使得這些服務很難廣泛部署並用於各種不同的應用。這就是近期的機會，即互動式對話式人工智慧推理。你可以想像每個超大規模企業都在競相實現這一目標，因為最近我們在機器學習語言模型方面取得了一些重要的突破。我之前提到的 BERT 模型確實是一個非常重要的進展，並且它引發了大量基於其改進的衍生模型。因此近期的對話式人工智慧推理。

We're also seeing near term the inference at the edge. There are many types of applications where because of the laws of physics reasons, the speed of light reasons or the economics reasons or data sovereignty reasons, it's not possible to stream the data to the cloud and have the inference done at the cloud. You have to do that at the edge. You need the latency to be low, the amount of data that you're streaming is continuous. And so you don't want to be paying for that line rate the whole time, and maybe the data is of great confidentiality or privacy. And so we're seeing a lot of excitement and a lot of development for edge AI. Smart retail, smart warehouses, smart factories, smart cities, smart airports, you just make a list of those kind of things, basically locations where there is a lot of activity, where safety or cost or large amount of materials is passing through, you could just imagine the applications. All of those really want to be edge computing systems and edge inference systems. And so those are near term -- 2 near-term drivers, and I think it's fair to say that both of them are quite large opportunities.

我們也看到了近期邊緣的推論。有許多類型的應用程式由於物理定律、光速、經濟或資料主權等原因，無法將資料傳輸到雲端並在雲端進行推理。你必須在邊緣做到這一點。您需要較低的延遲，並且傳輸的資料量是連續的。因此，您不想一直支付線路費用，而且資料可能具有很高的機密性或隱私性。因此，我們看到了邊緣人工智慧的諸多興奮和大量發展。智慧零售、智慧倉庫、智慧工廠、智慧城市、智慧機場，你只需列出這些事物的清單，基本上就是活動頻繁、安全或成本或大量材料經過的地方，你可以想像這些應用。所有這些都確實希望成為邊緣運算系統和邊緣推理系統。所以這些是近期的——兩個近期的驅動因素，我認為可以公平地說，它們都是相當大的機會。

So to answer your question regarding gross margin in a little bit more detail, probably our largest area that we expect improvement in terms of our mix is our mix return regarding our overall gaming business. We expect to have a full quarter of our SUPER lineup within the next quarter, including our RTX as well as our notebook becoming a bigger mix as well as it grows. These drivers are one of the largest reasons why we see that growth in our gross margin. We always think about our component cost, our overall cost of manufacturing, so this is always baked in over time, but we'll continue to see improvements on that as well.

因此，為了更詳細地回答您關於毛利率的問題，我們預計在產品組合方面改進的最大領域可能是我們整體遊戲業務的產品組合回報。我們預計下個季度我們的 SUPER 產品線將有完整季度的推出，包括我們的 RTX 以及我們的筆記本，隨著產品線的成長，其組合也會越來越大。這些驅動因素是我們毛利率成長的最大原因之一。我們始終考慮我們的零件成本和整體製造成本，因此這一點始終會隨著時間的推移而不斷完善，但我們也會繼續看到這方面的改進。

And your next question comes from the line of Harlan Sur with JPMorgan.

您的下一個問題來自摩根大通的 Harlan Sur。

Again, your data center business, many of your peers on the compute and storage side are seeing spending recovery by cloud and hyperscalers in the second half of this year after a similar weak first half of the year. You guys saw some growth in Q2 driven primarily by enterprise. It seems like you had some broadening out of the customer spending this quarter. Inferencing continues to see strong momentum. Would you guys expect that this translates into a double-digit percentage sequential growth in data center in Q3 off of the low base in Q2?

再次，您的資料中心業務，計算和儲存方面的許多同行在經歷了同樣疲軟的上半年之後，在今年下半年看到了雲端運算和超大規模的支出復甦。你們看到了第二季的一些成長，主要由企業推動。看起來本季您的客戶支出有所擴大。推理持續保持強勁勢頭。你們是否預計這會轉化為資料中心在第三季在第二季較低基數的基礎上實現兩位數的連續成長？

Our hyperscale data center with a few customers don't give us very much -- we don't get very much visibility from a handful of customers in hyperscale. However, we're seeing broad-based growth and excitement in data centers. And the way to think about data center, our data center business consists of hyperscale training, internal training, hyperscale inference, cloud computing -- and that's hyperscale, and that cloud is a public cloud. And then we have vertical industry enterprise, what sometimes we call enterprise, vertical industry enterprise, it could be transportation companies, retailers, telcos, vertical industry adoption of AI either to accelerate their business or to develop new products and services.

我們的超大規模資料中心只有少數客戶，但這並沒有為我們帶來太多好處——我們無法從少數超大規模客戶那裡獲得太多的可見性。然而，我們看到數據中心的廣泛增長和興奮。思考資料中心的方式是，我們的資料中心業務包括超大規模訓練、內部訓練、超大規模推理、雲端運算——這是超大規模的，而且雲端是公有雲。然後我們有垂直行業企業，有時我們稱之為企業，垂直行業企業，它可能是運輸公司、零售商、電信公司，垂直行業採用人工智慧來加速他們的業務或開發新產品和服務。

And then the -- so when you look at our data center from that perspective and these pieces, although we don't see as much -- we don't get as much visibility as we like in a couple of the large customers, the rest of the hyperscalers, we're seeing broad-based growth. And so we're experiencing the enthusiasm and the energy that maybe the others are. And so we'll keep reporting -- updating you guys as we go. We'll see how it goes.

然後 — — 因此，當您從這個角度和這些部分來看我們的資料中心時，儘管我們看不到那麼多 — — 我們對幾個大客戶、其餘超大規模客戶沒有獲得盡可能多的可見性，但我們看到廣泛的增長。因此，我們感受到了其他人可能也感受到的熱情和活力。因此，我們會持續向大家報道最新進展。我們將拭目以待。

And your next question comes from the line of Timothy Arcuri with UBS.

您的下一個問題來自瑞銀的 Timothy Arcuri。

I had 2. I guess first for Jensen, Volta's been around now for about 2 years. Do you see signs of demand maybe building up ahead of the new set of nanometer products, whenever that comes out? I guess I'm just wondering whether there's some element of this is more around product cadence that gets resolved as you do roll out the product. That's the first question.

我有 2 個。我想首先對於 Jensen 來說，Volta 已經存在大約 2 年了。無論新奈米產品何時推出，您是否看到需求在它們之前逐漸增強的跡象？我想我只是想知道這其中是否存在一些與產品節奏有關的因素，這些問題會在您推出產品時得到解決。這是第一個問題。

And then I guess, the second question, Colette, is of the $300 million growth into October, it sounds like Switch is pretty flat, but I'm wondering if you can give us maybe some qualitative sense of where the growth is coming from, is it maybe like 2/3 gaming and 1/3 data centers, something like that?

然後我想，第二個問題，科萊特，10 月份的 3 億美元增長中，聽起來 Switch 相當平穩，但我想知道您是否可以給我們一些定性的信息，說明增長來自哪裡，是不是 2/3 來自遊戲，1/3 來自數據中心，諸如此類？

Well, Volta -- data center products can churn that fast. We -- gamers could churn products quickly because they're bought and sold one at a time. But data centers -- data center infrastructure really has to be planned properly, and the build-out takes time. And we expect Volta to be successful all the way through next year. And software still continues to be improved on it. We're still improving systems on it. And in fact, just 1 year -- in just 1 year, we improved our AI performance on Volta by almost 2x, 80%. And so you could just imagine the amount of software that's built on top of Volta and all the Tensor Cores and all the GPUs connected with NVLink and the large number of nodes that are connected to build supercomputers.

嗯，Volta－資料中心產品可以快速更新換代。我們－遊戲玩家可以快速地更換產品，因為他們一次只能買賣一個產品。但是資料中心－資料中心基礎設施確實需要妥善規劃，而且建設需要時間。我們預計 Volta 明年將取得成功。且軟體仍在不斷改進。我們仍在改進該系統。事實上，僅僅一年時間——僅僅一年時間，我們就將 Volta 上的 AI 性能提高了近 2 倍，即 80%。因此，你可以想像一下在 Volta 之上構建的軟體數量、所有 Tensor Core 和所有透過 NVLink 連接的 GPU 以及建立超級電腦時連接的大量節點。

The software of building these systems, large-scale systems, is really, really hard. And that's one of the reasons why you hear people talk about chips, but they never show up because building the software is just an enormous undertaking. The number of software engineers we have in the company is in the thousands, and we have the benefit of having built on top of this architecture for over 1.5 decades. And so when we're able to deploy into data centers as quickly as we do, I think we kind of lose sight of how hard it is to do that in the first place. The last time a new processor entered into a data center was an x86 Xeon, and you just don't bring processors in the data centers that frequently or that easily. And so I think the way to think about Volta is that it's surely in its prime, and it's going to keep -- continue to do well all the way through next year.

建構這些系統、大型系統的軟體真的非常非常困難。這就是為什麼你聽到人們談論晶片，但他們從未出現的原因之一，因為建立軟體是一項艱鉅的任務。我們公司有數千名軟體工程師，並且我們已經在這個架構上建置了超過 15 年，這為我們帶來了好處。因此，當我們能夠如此快速地部署到資料中心時，我認為我們有點忽略了這樣做有多困難。上一次進入資料中心的新處理器是 x86 Xeon，而且處理器進入資料中心的頻率和難度並不大。因此，我認為 Volta 肯定正處於巔峰時期，並且明年它將繼續保持良好的表現。

In regard to our guidance on revenue, and we do guide in terms of the total. You have seen, in this last quarter, we executed a sequential increase really focusing on moving to a normalization of our gaming business. And we're now approaching the second half of the year getting ready for the back to school and the holidays. So you should expect also our gaming business to continue to grow to reach that full normalization by the end of Q3. We do expect the rest of our platforms to likely also grow. We have a couple different models on how that will come out. But yes, we do expect our data center business to grow, and then we'll see on the rest of our businesses as well.

關於我們對收入的指導，我們確實對總額進行了指導。你已經看到，在上個季度，我們實現了連續成長，真正專注於實現遊戲業務的正常化。現在我們正接近下半年，為開學和假期做準備。因此，您也應該預期我們的遊戲業務將繼續成長，並在第三季末完全正常化。我們確實預計我們其餘的平台也可能會成長。對於這一結果，我們有幾個不同的模型。但是的，我們確實預計我們的資料中心業務將會成長，然後我們也會看到我們其他業務的成長。

Your next question comes from the line of Matt Ramsay with Cowen.

您的下一個問題來自 Cowen 的 Matt Ramsay。

A couple of questions. I guess the first one is, Jensen, if you have any, I guess, high-level qualitative commentary on how the new SUPER upgrades of your Turing platform have been received in the market and how you might think about them progressing through the year. And then, I guess, the second question is a bigger one. Intel's talked quite openly about One API. The software stack at Xilinx is progressing with Versal ACAP. I mean you guys get a lot of credit for the decade of work that you've done on CUDA. But I wonder if you might comment on if you've seen any movement in the competitive landscape on the software side for the data center space.

有幾個問題。我想第一個問題是，Jensen，如果您有任何關於圖靈平台的新 SUPER 升級在市場上的反響，以及您對它們在今年的進展有何看法，我想您可以給出一些高水平的定性評論。然後，我想，第二個問題更大。英特爾非常公開地談論 One API。Xilinx 的軟體堆疊正在透過 Versal ACAP 取得進展。我的意思是，你們在 CUDA 上所做的十年工作獲得了很多讚譽。但我想知道您是否可以評論一下您是否看到資料中心領域軟體方面的競爭格局有任何變化。

SUPER is off to a great start. Goodness, SUPER is off to a super start. And if you look at -- if you do channel checks all over, even though we've got a lot of products in the channel and we -- last quarter was a transitional quarter for us actually. And we didn't -- we shipped SUPER later in the quarter. But because the entire ecosystem and all of our execution engines are so primed, we were able to ship a fair number through the channel. And so -- and yet, if you do spot checks all around the world, they're sold out almost everywhere. And the pricing in the spot market is drifting higher than MSRP. That just tells you something about demand. And so that's really exciting. SUPER is off to a super start for -- and at this point, it's a foregone conclusion that we're going to buy a new graphics card, and it's going to the last 2, 3, 4 years to not have ray tracing is just crazy. Ray tracing content just keeps coming out. And between the performance of SUPER and the fact that it has ray tracing hardware, it's going to be super well positioned for -- throughout all of next year.

SUPER 開局良好。天哪，SUPER 的開局非常好。如果你看一下——如果你對整個管道進行檢查，儘管我們在通路中有很多產品，但上個季度對我們來說實際上是一個過渡季度。但我們沒有——我們在本季度晚些時候發貨了 SUPER。但由於整個生態系統和我們所有的執行引擎都已準備就緒，我們能夠透過該管道運送相當數量的貨物。然而，如果你在世界各地進行抽查，你會發現它們幾乎在所有地方都已售罄。現貨市場的定價正高於製造商建議零售價。這只是告訴你一些有關需求的情況。這真是令人興奮。SUPER 已經有了一個非常好的開始——從這一點來看，我們肯定會購買一張新的顯示卡，而且在接下來的 2、3、4 年內沒有光線追蹤技術簡直是瘋了。光線追蹤內容不斷湧現。憑藉 SUPER 的出色性能以及其配備的光線追蹤硬件，它將在整個明年佔據非常有利的地位。

Your question about APIs and software programmability. APIs is just one of the issues. The large issue about processors is how do you program it. The reason why x86s and CPUs are so popular is because they solve the great challenge of software developers: how to program a computer. And how to program a computer and how to compile for that computer is a paramount concern to computer science, and it's an area of tremendous research. Going from single CPU to multi-core CPUs was a great challenge. Going from multi-core CPUs to multi-node multi-core CPUs is an enormous challenge. And yet, when we created CUDA in our GPUs, we went from 1 CPU core or one processor core to a few to now, in the case of large-scale systems, millions of processor cores. And how do you program such a computer across multi-GPU, multi-node? It's a concept that's not easy to grasp. And so I don't really know how one programming approach or a simple API is going to make 7 different type of weird things work together. And I can't make it fit in my head. But programming isn't as simple as a PowerPoint slide, I guess. And I think it's just -- time will tell whether one programming approach could fit 7 different types of processors when no time in history has it ever happened.

您關於 API 和軟體可編程性的問題。API 只是其中一個問題。關於處理器的最大問題是如何對其進行編程。x86 和 CPU 之所以如此受歡迎，是因為它們解決了軟體開發人員面臨的巨大挑戰：如何對電腦進行程式設計。如何編寫電腦程式以及如何為電腦進行編譯是電腦科學最關注的問題，也是一個需要大量研究的領域。從單 CPU 到多核心 CPU 是一個巨大的挑戰。從多核心 CPU 到多節點多核心 CPU 是一個巨大的挑戰。然而，當我們在 GPU 中建立 CUDA 時，我們從 1 個 CPU 核心或一個處理器核心發展到幾個，現在，在大型系統中，有數百萬個處理器核心。那麼如何跨多 GPU、多節點對這樣的電腦進行程式設計呢？這是一個不容易理解的概念。所以我真的不知道一種程式方法或一個簡單的 API 如何讓 7 種不同類型的奇怪東西協同工作。我無法將它融入我的腦海裡。但我想，程式設計並不像 PowerPoint 投影片那麼簡單。我認為時間會證明一種程式設計方法是否能夠適應 7 種不同類型的處理器，而歷史上從未發生過這種情況。

Your next question comes from the line of Joe Moore with Morgan Stanley.

您的下一個問題來自摩根士丹利的喬·摩爾。

I wonder if you could talk about the strength in the automotive business. Looks like the services piece of that is getting to be bigger. What's the outlook for that part of the business? And can you give us a sense of the mix between services and components at this point?

我想知道您是否可以談談汽車業務的優勢。看起來服務部分正在變得越來越大。這部分業務的前景如何？您能否讓我們了解一下目前服務和組件之間的混合情況？

Sure. Thanks, Joe. Our approach to autonomous vehicles comes in basically 2 parts. The first part is a full stack, which is building the architected processor, the system, the system software and all of the driving applications on top, including the deep neural nets. The second part of it, we call that a full stack self-driving car computer. The second part of DRIVE includes an end-to-end AV development system. For those who would like to use our processors, use our system software but create their own applications, we created a system that allows -- basically shares with them our computing infrastructure that we built for ourselves that allows them to do end-to-end development from deep learning development to the application of AV to simulating that application to doing regression testing of that application before they deploy it into a car. And the 2 systems that we use there is called DGX for training and Constellation for simulation and what is called Replay.

當然。謝謝，喬。我們自動駕駛汽車的方法基本上分為兩個部分。第一部分是全端，包括建立架構處理器、系統、系統軟體以及所有驅動應用程序，包括深度神經網路。第二部分，我們稱之為全端自動駕駛汽車電腦。DRIVE 的第二部分包括端對端 AV 開發系統。對於那些想要使用我們的處理器、使用我們的系統軟體但創建自己的應用程式的人，我們創建了一個系統，基本上允許與他們共享我們為自己構建的計算基礎設施，使他們能夠進行端到端開發，從深度學習開發到 AV 應用程序，再到模擬該應用程序，再到在將該應用程序部署到汽車之前對該應用程序進行測試。我們在那裡使用的兩個系統分別是用於訓練的 DGX、用於模擬的 Constellation 和所謂的 Replay。

And then the third part of our business model is development agreements, otherwise known as NRE. These 3 elements, full stack computer, end-to-end development flow and NRE project development -- product development consists of the overall DRIVE business. And so although the cars will take several years to go into production, we're seeing a lot of interest in working with us to develop self-driving cars using our development systems and entering into development projects. And so we're -- the number of autonomous vehicle projects is quite large around the world as you can imagine. And so my sense is that we're going to continue to do well here. The additional part of autonomous vehicles and where the capability has been derived and is going to seal up more near-term opportunities has to do with things like delivery shuttles, self-driving shuttles and maybe cargo movers inside walled warehouses. Those kind of autonomous machines require basically the same technology, but it's sooner and easier to deploy. And so we are seeing a lot of excitement around that area.

我們的商業模式的第三部分是開發協議，也稱為 NRE。全端電腦、端對端開發流程和 NRE 專案開發－產品開發這三個要素構成了整體 DRIVE 業務。因此，儘管汽車需要幾年時間才能投入生產，但我們看到許多人有興趣與我們合作，使用我們的開發系統開發自動駕駛汽車並參與開發專案。因此，正如您所想像的，世界各地的自動駕駛汽車項目數量相當龐大。所以我的感覺是，我們將繼續在這裡做得很好。自動駕駛汽車的附加部分以及其能力的來源以及將要鎖定的更多近期機會與送貨班車、自動駕駛班車以及圍牆倉庫內的貨物搬運車有關。這些類型的自主機器需要基本上相同的技術，但部署起來更快、更容易。因此我們看到該地區有很多令人興奮的事情。

Your next question comes from the line of Aaron Rakers with Wells Fargo.

您的下一個問題來自富國銀行的 Aaron Rakers。

Congratulations on the improved performance. At your Analyst Day back a couple of months ago, you had highlighted the installed base opportunity for RTX. And I think at that point in time, you talked about 50% being Pascal base, 48% being pre-Pascal. You also alluded to the fact that you were seeing a positive mix shift higher in terms of the price points of this RTX cycle. So I'm curious, where do we stand on the current product cycle? And what are you seeing currently as we go through this product cycle on the Turing platforms?

恭喜您業績提升。在幾個月前的分析師日上，您強調了 RTX 的安裝基礎機會。我認為在那個時候，您談到了 50% 是基於 Pascal 的，48% 是 pre-Pascal 的。您也提到，您看到 RTX 週期的價格點出現了積極的組合轉變。所以我很好奇，我們在目前的產品週期中處於什麼位置？當我們經歷圖靈平台的產品週期時，您目前看到了什麼？

We launched -- well, first of all, the answer is that RTX adoption is faster than Pascal's adoption if you normalize to time 0 of launch. The reason for that is Pascal launched top to bottom on the same day. And as you guys know, we weren't able to do that for Turing. But if we did that for Turing, the adoption rate is actually faster. And to me, it's a rather sensible. And the reason for that is because Pascal was basically DX12. And Maxwell was DX12. And Turing is the world's first DXR, the first ray tracing GPU, brand-new functionality, brand-new API and a lot more performance. And so I think it's sensible that Turing's adoption is going to be rapid.

我們推出了——嗯，首先，如果以發佈時間為 0 來計算，那麼 RTX 的採用速度比 Pascal 的採用速度更快。原因在於 Pascal 是在同一天從上到下發射的。正如你們所知，我們無法為圖靈做到這一點。但如果我們對圖靈也這樣做，採用速度實際上會更快。對我來說，這是相當明智的。原因在於 Pascal 基本上是 DX12。而Maxwell是DX12。Turing 是世界上第一個 DXR、第一個光線追蹤 GPU，具有全新的功能、全新的 API 和更強大的效能。因此我認為圖靈的普及將會非常迅速，這是理所當然的。

The second element of Turing is something that we've never talked about before. We're mentioning it more and more because it's such an exciting book market for us is notebooks. The install base of Pascal has a very, very little notebook in it. And the reason for that is because, in the past, we were never able to put a high performance gaming GPU into a thin and light notebook until we invented Max-Q. And in combination with our energy efficiency, we were able to -- we're now able to put a 2080 into a laptop, and it's still beautiful. And so this is effectively a brand-new growth market for us. And with so few people and so few gamers in the world that are able to game on a laptop, I think this is going to be a nice growth market for us.

圖靈的第二個要素是我們以前從未談論過的。我們越來越多地提及它，因為對我們來說，筆記本是一個非常令人興奮的圖書市場。Pascal 的安裝基礎中有一個非常小的筆記本。原因在於，過去，直到我們發明了 Max-Q，我們才能夠將高效能遊戲 GPU 放入輕薄筆記型電腦中。結合我們的能源效率，我們現在能夠將 2080 放入筆記型電腦中，而且它仍然很漂亮。因此，這對我們來說實際上是一個全新的成長市場。由於世界上能夠在筆記型電腦上玩遊戲的人和遊戲玩家如此之少，我認為這對我們來說將是一個很好的成長市場。

And then the new market that we introduced and launched this last quarter is called RTX Studio. And this is an underserved segment of the market where consumers, enthusiasts, they could be artists that are working on small firms, they need powerful computers to do their work. They need powerful computers to do rendering and high-definition video editing. And yet it's underserved by workstations because workstations are really sold on a B2B basis into large enterprises. And so we aligned all of the OEMs and created a whole new line of notebooks called RTX Studio. And the enthusiasm has been great. We've launched 27 different laptops, and I'm looking forward to seeing the results of that. This is tens of millions of people who are creators. Some of them professionals, some of them hobbyists. And they use Adobe suites, they use Autodesk in their suites and some of them use SolidWorks and some of them use all kinds of renders, like blender. And these are 3D artists and video artists, and this digital content creation is the modern way of creativity. And so this is an underserved market that we're excited to go serve with RTX Studio.

我們在上個季度推出的新市場叫做 RTX Studio。這是一個服務不足的市場，消費者、愛好者，他們可能是在小公司工作的藝術家，他們需要功能強大的電腦來完成他們的工作。他們需要功能強大的電腦來進行渲染和高清影片編輯。然而工作站的服務卻不足，因為工作站實際上是以 B2B 的方式銷售給大型企業的。因此，我們聯合所有 OEM 廠商，打造了一個名為 RTX Studio 的全新筆記型電腦系列。人們的熱情也非常高漲。我們已經推出了 27 款不同的筆記型電腦，我期待看到其成果。這是數千萬的創造者。他們有些是專業人士，有些是業餘愛好者。他們使用 Adob​​e 套件，在他們的套件中使用 Autodesk，其中一些使用 SolidWorks，還有一些使用各種渲染器，例如 blender。這些都是 3D 藝術家和錄像藝術家，這種數位內容創作是現代的創造力方式。因此，這是一個服務不足的市場，我們很高興能夠透過 RTX Studio 為其提供服務。

And your last question comes from the line of Stacy Rasgon with Bernstein Research.

您的最後一個問題來自伯恩斯坦研究公司的 Stacy Rasgon。

I have 2 for Colette. My first question is on data center. So I know you say that you have a broad-based growth except for a few hyperscalers. But you only grew at 3% sequentially, about $20 million. That doesn't sound like broad-based growth to me unless like -- did the hyperscalers get worse? Or are they just still so much bigger than like the rest of it? I guess, what's going on in data center? How do I wrap my head around like broad-based growth with relatively minimal growth observed?

我有 2 個給 Colette。我的第一個問題是關於資料中心的。所以我知道你們說除了少數超大規模企業外，你們都實現了廣泛的成長。但你的收入僅環比成長了 3%，約 2000 萬美元。在我看來，這聽起來不像是廣泛的成長，除非——超大規模企業的情況變得更糟了？或者它們仍然比其他部分大得多？我猜，資料中心發生了什麼事？我如何理解在觀察到相對較小的增長的情況下出現廣泛的增長？

So to answer your question here, Stacy, on what we refer to when we're discussing the broad-based growth is the substantial expansion that we have on the types of customers and the industries that we are now approaching. As you know, even a year ago, we had a very, very small base in terms of industry-based hyper -- excuse me, industry-based AI workloads that they were using. Over this last quarter, we're continuing to see strong growth as we roll out all different types of AI solutions, both across the U.S. and worldwide, to these overall customers. Our hyperscalers, again, a couple of them, not necessarily growing. Some of them are flat and some of them are growing depending on whether or not that's for cloud instances or whether or not they're using it for internal use. So we believe that our continued growth with the industries is important for us for the long term to expand the use of AI, and we're just really pleased with what we're seeing in that growth this quarter.

因此，史黛西，要回答你的問題，當我們討論廣泛的成長時，我們指的是我們現在所針對的客戶類型和行業的大幅擴張。如你所知，即使在一年前，我們在基於行業的超級——對不起，是基於行業的人工智慧工作負載——方面的基礎也非常小。在上個季度，隨著我們向美國和全球的整體客戶推出各種不同類型的人工智慧解決方案，我們繼續看到強勁的成長。我們的超大規模企業，其中有幾個，不一定會成長。其中一些是持平的，而另一些則在增長，這取決於是否用於雲端實例或是否將其用於內部用途。因此，我們相信，與各行業一起持續成長對於我們長期擴大人工智慧的應用至關重要，我們對本季的成長感到非常滿意。

I'll now turn the call back over to Jensen for any closing remarks.

現在我將把電話轉回給詹森，請他做最後發言。

Thanks, everyone. We're happy with our results this quarter and our return to growth across our platforms. Gaming is doing great. It's great to see NVIDIA RTX reinvigorating the industry. GeForce has several growth drivers. Ray traced games continue to gain momentum. A large number of gaming laptops are rolling out, and our new Studio platform is reaching the large underserved community of creators. Outside a few hyperscalers, we're seeing broad-based growth in data centers. AI is the most powerful technology force of our time and a once-in-a-lifetime opportunity. More and more enterprises are using AI to create new products and services while leveraging AI to drive ultra-efficiency and speed in their business. And with hyperscalers racing to harness recent breakthroughs in conversational AI, we see growing engagements in training as well as interactive conversational inference. RTX, CUDA accelerated computing, AI, autonomous vehicles, the work we're doing is important, impactful and incredibly fun. We're just grateful there is so much of it. We look forward to updating you on our progress next quarter.

謝謝大家。我們對本季的業績以及我們各個平台的恢復成長感到滿意。遊戲表現很棒。很高興看到 NVIDIA RTX 為這個行業注入新的活力。GeForce 有多個成長動力。光線追蹤遊戲繼續獲得發展勢頭。大量遊戲筆記型電腦正在推出，我們的新 Studio 平台正在覆蓋大量服務不足的創作者群體。除了少數超大規模企業外，我們還看到資料中心的廣泛成長。人工智慧是我們這個時代最強大的科技力量，也是千載難逢的機會。越來越多的企業正在利用人工智慧創造新產品和服務，同時利用人工智慧來提高業務的效率和速度。隨著超大規模企業競相利用對話式人工智慧的最新突破，我們看到訓練和互動式對話推理的參與度不斷提高。RTX、CUDA 加速運算、人工智慧、自動駕駛汽車，我們正在做的工作很重要、很有影響力，而且非常有趣。我們很感激有這麼多這樣的人。我們期待下個季度向您通報我們的進展。

This concludes today's conference call. You may now disconnect.

今天的電話會議到此結束。您現在可以斷開連線。