Now, the battle of the semiconductor space heating up with Intel, NVIDIA, and AMD all announcing new next-generation AI chips.

現在，隨著英特爾、英偉達（NVIDIA）和 AMD 都宣佈推出新的下一代人工智能芯片，半導體領域的爭奪戰進入白熱化。

But we're going to zoom in on NVIDIA.

不過，我們還是要放大英偉達公司。

It's already been a huge year for the stock, obviously.

很明顯，今年對這隻股票來說已經是意義非凡的一年。

Shares, though, up over 130% so far just in 2024, surpassing that $1,000 milestone for the first time a few weeks ago, and just this morning hitting another record high.

不過，就在 2024 年，公司股價迄今已上漲超過 130%，幾周前首次突破 1000 美元的里程碑，今天上午又創下新高。

But it hasn't just been this year, right?

但不只是今年，對吧？

NVIDIA's dominance, sending shares skyrocketing over 500% in the last two years.

英偉達™（NVIDIA®）的主導地位，使其股價在過去兩年中飆升了 500% 以上。

Take a look at that extreme climb there and that coming as investors double down on AI hype.

請看這裡的極速攀升，投資者對人工智能的炒作加倍。

All of NVIDIA's momentum leading to a lot of bullishness across Wall Street, Bank of America becoming the latest, hiking their price target on NVIDIA this week from $1,320 to $1,500.

英偉達的發展勢頭引發了華爾街的廣泛看好，美國銀行（Bank of America）就是最新的一家，本週他們將英偉達的目標價從 1320 美元上調至 1500 美元。

This is now the highest call on Wall Street for the chip giant.

現在，華爾街對這家芯片巨頭的呼聲最高。

The analyst behind the call, Vivek Arya, is Bank of America's senior semiconductor analyst, and he joins us now.

電話背後的分析師 Vivek Arya 是美國銀行的資深半導體分析師，他現在加入我們的談話。

Vivek, thank you so much for being here with us to talk about your bullish call on NVIDIA.

維韋克，非常感謝您能與我們一起討論您看漲英偉達的觀點。

I'm curious, why $1,500?

我很好奇，為什麼是 1500 美元？

Why not doubling your price target on this name considering the amount of bullishness and the fundamentals that back it up?

考慮到看漲的程度和支持它的基本面，為什麼不將目標價格翻倍呢？

Sure.

當然。

Good morning.

早上好。

Thank you for having me.

謝謝你們邀請我。

So I think there are two aspects to it.

是以，我認為這有兩個方面。

The first is where we are in the cycle of converting data centers to accelerated computing.

首先，我們正處於數據中心向加速計算轉換的週期中。

The concept of the data center is not new.

數據中心的概念並不新鮮。

It's been around since the 1940s.

自 20 世紀 40 年代以來，它就一直存在。

But then we have these multi-decade infrastructure upgrade cycles from the mainframe to the microcomputer to cloud computing.

但是，從大型機到微型計算機，再到雲計算，基礎設施的升級週期長達數十年。

And now we are at the start of what I think would be a decade-long conversion over to accelerated computing.

而現在，我們正處於我認為長達十年的向加速計算轉換的起點。

What is accelerated computing?

什麼是加速計算？

It's a way of taking a lot of very intensive data processing tasks that have to do with images, with voice, with video, with the processing of these large language models, and converting the data center with new kinds of semiconductors to make sure that processing can be done very efficiently.

這是一種處理大量密集型數據處理任務的方法，這些任務涉及影像、語音、視頻以及大型語言模型的處理，並通過新型半導體對數據中心進行轉換，以確保能夠非常高效地進行處理。

We are just at the start of this cycle.

我們才剛剛開始這個週期。

We think that the spending could be anywhere between $250 billion to $500 billion a year, and NVIDIA is And I know there is often this comparison to, let's compare it to Intel of the past, or let's compare it to Cisco of the past, or other parts of the technology upgrade cycle.

我們認為，每年的支出可能在 2500 億到 5000 億美元之間，而英偉達也是其中之一。

But what is different is that NVIDIA is bringing a combination of silicon, of systems, of networking, of software, of developers, and that is unprecedented.

但不同的是，英偉達帶來了芯片、系統、網絡、軟件和開發人員的組合，這是前所未有的。

So to come to your question, technology stocks, the price targets really move because of the size of the market, because of their execution, and even at this $1,500, the stock would essentially be trading at one times their earnings growth rate.

所以，對於你的問題，科技股的目標價格確實會因為市場規模和執行力而變動，即使是在 1,500 美元的價位上，股票的交易價格基本上也是其盈利增長率的 1 倍。

If you look at the S&P 500, it's trading at two times their earnings growth rate.

如果你看一下標準普爾 500 指數，它的交易價格是其盈利增長率的兩倍。

So I would claim that even at this kind of price level, it's actually trading below where the broader market is in terms of the price-to-earnings-to-growth ratio.

是以我認為，即使是在這樣的價格水平上，它的市盈率與增長比率實際上也低於大盤的水準。

High target on the street for you, Vivek, for NVIDIA.

你的目標很高，Vivek，英偉達。

What is the significance that you and deals like the one that is being reported this morning, XAI potentially going to be taking the chips that Tesla was supposed to be getting from NVIDIA here?

你們和今天早上報道的 XAI 這樣的交易，有可能從英偉達那裡獲得特斯拉的芯片，這有什麼意義？

Yeah, I think there is a race going on between technology companies of various types.

是的，我認為各類技術公司之間正在進行一場競賽。

We have seen that across the cloud of players who have raised their capital spending forecast to over 40% to 45% growth.

我們已經看到，許多公司都將資本支出預測提高到了 40% 至 45% 的增長。

Last year, we were thinking that growth would be 20%.

去年，我們認為增長率將達到 20%。

It's already 40% to 45%.

現在已經是 40% 到 45%。

There is a race between these different technology companies to be the first to support the next large language model, to make sure that they are the destination of choice for a lot of startups who are trying to really kickstart their own generative AI applications.

這些不同的技術公司之間都在爭先恐後地支持下一個大型語言模型，以確保它們成為許多試圖真正啟動自己的生成式人工智能應用的初創公司的首選目標。

And then this market is broadening from these cloud service providers to a lot of enterprise verticals.

然後，這個市場正在從這些雲服務提供商擴大到許多垂直企業。

And in this most recent quarter, NVIDIA did call out that automotive and automotive customers, which I assume includes Tesla, are part of that really large vertical.

在最近的這個季度，英偉達確實提到了汽車和汽車客戶，我想其中也包括特斯拉，它們都是這一大型垂直行業的一部分。

So we see the spreading from cloud customers towards automotive, towards healthcare, a lot of drug discoveries being done with this infrastructure.

是以，我們看到雲客戶正在向汽車和醫療保健領域擴展，很多藥物的發現都是通過這種基礎設施完成的。

Financial services are benefiting from it.

金融服務正從中受益。

And then the last but not least part that we will start to see more is the rise of this infrastructure amongst sovereign.

最後但並非最不重要的一點是，我們將開始看到主權國家中基礎設施的崛起。

So we have seen India, Europe, Japan, every country and region, the Middle East, they are trying to make sure that they're able to use their culture, their language to really go and train these large language models.

是以，我們看到印度、歐洲、日本、每個國家和地區、中東地區，都在努力確保他們能夠利用自己的文化和語言，真正去訓練這些大型語言模型。

And we think we are just at the start of what that, and that could be a really large infrastructure upgrade cycle as well.

我們認為，我們只是剛剛開始，這也可能是一個真正的大型基礎設施升級週期。

Vivek, you mentioned customers and it makes me think about Intel's latest chip announcement.

Vivek，你提到了客戶，這讓我想到了英特爾最新發布的芯片。

And the executives really speaking about this idea that their chips are not necessarily for the hyperscalers.

高管們確實在談論這樣一個觀點，即他們的芯片並不一定是為超級計算機準備的。

They're for more of the enterprise customers.

它們面向更多的企業客戶。

They're going to be a more affordable.

他們會更實惠。

To what extent do you see that as a winning strategy for Intel moving forward, particularly given the amount of headwinds facing that stock over the course of the past year here?

特別是考慮到英特爾股票在過去一年中所面臨的不利因素，您認為這在多大程度上是英特爾未來的制勝戰略？

Sure, absolutely.

當然，當然。

I think we have to give Intel the credit that they have very strong enterprise incumbency.

我認為，我們必須對英特爾表示讚賞，因為他們擁有非常強大的企業實力。

They are the brand name, one of the most trusted names in enterprise infrastructure.

他們是品牌，是企業基礎設施領域最值得信賴的品牌之一。

But broadly speaking, Intel is facing this challenge where they have been behind in AI.

但總的來說，英特爾正面臨著人工智能領域落後的挑戰。

They have the incumbency, but they don't quite have the deployed base, the range of software developers.

他們擁有現任職位，但卻沒有足夠的部署基礎和軟件開發人員。

For example, NVIDIA has 5 million developers on their CUDA platform.

例如，英偉達公司的 CUDA 平臺擁有 500 萬名開發人員。

So a lot of the enterprises, as they start to look towards their AI infrastructure, the first name they turn to is NVIDIA for the reason that there's a lot of developers who are already familiar with NVIDIA's cloud instances.

是以，很多企業在開始考慮其人工智能基礎架構時，首先會選擇英偉達，因為有很多開發人員已經熟悉了英偉達的雲實例。

And that too is where they find a lot of NVIDIA-based instances.

這也是他們發現大量基於英偉達™（NVIDIA®）處理器的實例的地方。

So broadly speaking, if I look at the overall AI chip market today, we think it's about $100 billion or so this year.

是以，從廣義上講，如果我看一下目前的整個人工智能芯片市場，我們認為今年的市場規模大約在 1000 億美元左右。

NVIDIA is about 80% of that.

英偉達約佔其中的 80%。

Custom chips from Broadcom, Marvell, and others are another 10 or 15%.

來自 Broadcom、Marvell 和其他公司的定製芯片則佔 10%或 15%。

And then you have this long tail, which has AMD, which has Intel, which has a number of startups.

然後是長尾，有 AMD，有英特爾，也有許多初創公司。

And even as I look forward over the next three to five years, as this market doubles or triples from here, we think that ratio will stay relatively in that range, that NVIDIA will continue to have 80%.

展望未來三到五年，隨著市場規模的翻番或翻三番，我們認為這一比例將相對保持在這一範圍內，英偉達將繼續保持 80% 的市場份額。

We think custom chips will have another 10 or 15%.

我們認為，定製芯片將再增加 10%或 15%。

And then you will continue to have this long tail of companies that have a presence, they benefit from the rising tide in this market, but they're never quite able to achieve the escape velocity that the leader in the market has.

然後，你將繼續看到長尾公司的存在，它們受益於市場的上升趨勢，但它們永遠無法達到市場領導者所擁有的逃逸速度。

And by the way, that is not too uncommon.

順便說一句，這種情況並不少見。

If you look at other parts of technology, search, social, e-commerce, operating systems, you see the leader that has the early start, that has the scale, the incumbency, the relationship with developers claiming 80 plus percent of the market.

縱觀其他技術領域，如搜索、社交、電子商務、作業系統等，你會發現起步早、規模大、地位穩固、與開發者關係融洽的領先企業佔據了 80% 以上的市場份額。

And we see exactly the same thing in the AI chip market as well.

我們在人工智能芯片市場也看到了同樣的情況。

So what would change that dynamic moving forward, Vivek?

那麼，維韋克，是什麼改變了這一態勢？

And I would love to get your answer on that in the context of some of the cloud providers that are trying to compete with NVIDIA.

我很想知道，在一些雲計算提供商試圖與英偉達競爭的背景下，你對此有何回答。

I was just in Amazon's chips lab, AWS's chips lab last week, taking a look at some chips that they have coming out here.

上週，我剛去過亞馬遜的芯片實驗室，也就是 AWS 的芯片實驗室，看了看他們即將推出的一些芯片。

And I'm curious what you think would make some of the cloud players like an Amazon a formidable opponent to NVIDIA?

我很好奇，您認為亞馬遜（Amazon）等雲計算公司如何才能成為英偉達（NVIDIA）的強勁對手？

No, absolutely.

不，絕對不。

I think we have to look at the cloud service providers as in some way competing with NVIDIA, but even more so being very strong customers for the company.

我認為，我們必須把雲服務提供商看作是英偉達的競爭對手，但更是公司的強大客戶。

Developing chips and developing cloud infrastructure are two different skill sets.

開發芯片和開發雲基礎設施是兩種不同的技能組合。

So if I'm developing a custom chip in a cloud, only that cloud can use it.

是以，如果我在雲中開發定製芯片，只有該雲才能使用它。

It's not going to scale any chip that Amazon develops.

亞馬遜開發的任何芯片都無法擴展。

It's not going to be used by Google.

谷歌不會使用它。

It's not going to be used by Microsoft.

微軟不會使用它。

It's not going to be used by Oracle or Meta.

甲骨文或 Meta 不會使用它。

So in some ways, it is serving a very specific need that Amazon might have.

是以，在某種程度上，它滿足了亞馬遜可能有的非常特殊的需求。

Similarly, the chips that Google develops, the TPUs, they serve very specific needs.

同樣，谷歌開發的芯片 TPU 也有非常特殊的需求。

What NVIDIA has developed is a so-called merchant reconfigurable programmable infrastructure that can be deployed across a very wide range of workloads, both for when these cloud companies are handling internal workloads, such as search, such as e-commerce, such as YouTube or TikTok-like video recommender engines, or when it is serving a lot of enterprise customers through their public cloud, which is why I mentioned that if you take a step back and look at the broader accelerated computing market, I do think that 10 to 15 percent will be these custom chips, but they will have several different custom chips across each of these specific cloud infrastructures.

英偉達™（NVIDIA®）所開發的是一種所謂的商家可重新配置的可編程基礎架構，可以部署在非常廣泛的工作負載中，既適用於這些雲公司處理內部工作負載，如搜索、電子商務、YouTube 或類似 TikTok 的視頻推薦引擎、這就是為什麼我提到，如果你退一步看更廣闊的加速計算市場，我確實認為10%到15%會是這些定製芯片，但他們會在每個特定的雲基礎設施中使用幾種不同的定製芯片。

But because they don't quite have a way to go from one cloud to another cloud, their range of workloads, their applicability, their scale tends to be limited to just that cloud infrastructure.

但由於它們沒有辦法從一個雲到另一個雲，其工作負載範圍、適用性和規模往往僅限於該雲基礎設施。