저자: 미상 · 외국인
요약 보기
AI 인프라 수요와 엔비디아 실적 기대를 다룬다. 주요 AI 연구소 경쟁과 평판 이슈가 언급되며, 데이터센터 임대 사업과 전력 확보가 핵심 병목으로 제시된다. 1.5GW 활성 전력, 이번 분기 약 500MW 추가, 51개 데이터센터, 1,040억 달러 수주잔고가 근거로 나온다. 뉴욕주 대규모 데이터센터 건설 모라토리엄 등 지역 반대가 리스크로 언급된다. 엔비디아 주가는 기술적 저항선을 회복했으나 실적 발표 전 기대가 높아 실망 가능성도 제기된다. 금융기관과 연계한 AI 인프라 자금조달 구조도 소개된다.
다듬은 글
model, it is 80-95% as good. This is a very complex dynamic for leading laboratories. By the way, do you see any changes? You know, for a long time, OpenAI and Sam have been under a barrage of criticism. There were big questions about trust and things like that. By the way, there may be even more questions because of the situation with Apple on Friday.
But it seems that Anthropic also has some problems. And this may be true for Alex Karp and others, but people think that either the costs don't make mathematical sense, or, if you look at the story with Figma, they're going into other businesses. anxiously. I mean, listen, let's be honest. There are three leaders of laboratory companies who have been in the spotlight for 20 years.
Elon, you have seen every facet of his personality and can judge 20 years of his behavior. Sundar and Demis, 20 years old. Zach, 20 years old. And I think these people have shown that they are essentially running a relatively simple and straightforward business, right? There are no tricky moments there.
Maybe they were on Facebook in the past. I think they fixed it, okay? But Google has definitely never done that, and neither has Elon. You may not like them for other reasons. And then suddenly these two guys appear who are incredible entrepreneurs, for sure. But the amount of data to draw on, the behavioral examples of when they had to make tough decisions, is still relatively limited because they have only been at such a public level and under such scrutiny for a few years.
So, if you look at the first few years of Facebook, it was a mess. I don't know about Google. Maybe the first few years to talk about you and your 20s. You'll still have time to talk about I mean, you know , I was portrayed as that guy on Facebook that made everyone cry. But when you look back, this is a $2 trillion company.
So, you know, these are crocodile tears. Can I just...IBM says that, you know, we just closed a few big deals. They always say something like that. But it's interesting. The customers they were trying to attract were distracted by industry-wide cybersecurity concerns. And what they were going to do If you have someone, if you have someone And now they'll say the weather was bad .
No, I have a real problem with this, and I've said it a few times already, so I'll just say it again . You can look at the graph of the financial needs of these companies. Yes. And you can tell exactly when they are hesitating between two different messages. Message. Hmm, geographically the difference isn't that big.
In terms of latency, the difference is small. People want to settle in North America, that is, in the USA and Canada. We see great demand in Europe. We recently announced data center expansion, expanding into the APAC region for the first time. But in general, enterprises are looking for much of the same infrastructure that AI labs use .
It's just a matter of the length of the contract. And let's talk about power. You have 1.5 gigawatts of active power. You added almost 500 megawatts this quarter alone. As Chief Development Officer, how do you overcome grid connection delays to deliver that much capacity in one quarter? Is this the biggest operational hurdle, or is there something else to achieve this result?
I think it's less about power delivery and more about how to physically deliver a data center or connected building. We do two things. We lease data center capacity from our trusted partners, and it's their responsibility to build and power the building , and then we go inside and deploy our GPU infrastructure and cloud infrastructure.
In addition, we ourselves participate in the construction of these facilities. This gives us a little more operating leverage for the entire portfolio. But it's less a question of access to electricity and more a question of whether there are enough electricians. Are there enough components to build these data centers, etc.?
This is not news to us, right? We have been overcoming these supply chain challenges for years . As I said, we have 51 data centers. It is our rich experience and development of partnerships in the supply chain that allows us to consistently meet deadlines. And how far in advance do you have to enter into power purchase agreements so that CoreWeave can meet this $104 billion order backlog?
I was looking at some data before our conversation, and the IEA predicts that electricity consumption by data centers worldwide could double by 2030 . So, my next question is: are you considering alternative energy sources, such as local nuclear or geothermal microgrids, to bypass traditional utilities?
Of course. We do not offer capacity to customers until we actually have a data center and the corresponding energy in our portfolio. We don't sign contracts for GPUs hoping to find free capacity in this limited and complex market. We changed this approach to ensure we don't disappoint our customers.
Regarding the types of power plants, we focus on grid electricity. Isn't that right ? Currently, we are almost not involved in autonomous generation or our own energy production projects. We consider grid energy to be the optimal resource and have found plenty of spare capacity across the US. As you mentioned about possible obstacles, one of them is local opposition to data centers.
I mean the NIMBY (not in my backyard) movement. The governor of New York has signed an executive order imposing a moratorium on the construction of large-scale data centers. Your CEO points out that such regulatory pressures are making your negotiations difficult. How are you adapting? How is your site selection strategy changing to circumvent this state and local resistance, and even bans?
Nvidia has become much more than indicators for the entire investment cycle in AI. As the earnings report approaches, investors are no longer just asking whether Nvidia will beat expectations. At this point, almost everyone is expecting another strong quarter. The real question is , will demand remain as strong in the years to come?
This is what will ultimately determine whether today's massive spending on AI infrastructure is justified. The shares have already made a confident recovery . They have returned above an important technical level and are now approaching the previous record high. This indicates a return of momentum, but there is a nuance.
The higher the stock rises before the report, the higher the expectations become. Therefore, Nvidia may show outstanding results, but the stock could still fall if investors were expecting something better. That's why the earnings conference call itself may be more important than the headline numbers.
Look at Amazon and Meta. Both companies reported strong results, but what caught investors' attention was their willingness to continue significant investments in AI infrastructure. Amazon detailed where its capital is going and provided strong forecasts. This is important because investors are looking for more than just a higher capex figure.
They need evidence that these costs are economically justified. And this is where Nvidia comes to the rescue. Nvidia actually supplies the hardware that underpins the infrastructure expansion. So when Nvidia reports, investors get another look at the state of the entire AI spending cycle. The market already knows that Nvidia makes money.
He already knows that demand is high. He already knows that hyperscalers spend billions. What investors need to know is how long can this last? Our customers are still actively placing orders for several years ahead. Our data centers are still struggling to provide enough computing power. Is supply still limited and are Nvidia customers seeing enough return to justify another wave of spending?
These answers may carry more weight than another earnings report. Now let's look at a large-scale project in Ohio related to OpenAI. This is not just about purchasing GPUs. The project included land, electricity, construction, and a huge data center campus. The developer is responsible for creating the physical infrastructure.
OpenAI becomes a customer through a long-term lease , and NVIDIA has secured exclusivity for computing hardware. This actually illustrates something important. AI infrastructure is becoming an extremely capital-intensive industry. Chips are just one component. Electricity, transmission infrastructure, buildings, cooling, networks and financing are needed.
And this is where NVIDIA's recent partnership with a major financial institution becomes particularly interesting. The company helps attract external idea is that institutional investors can provide financing for productive computing assets. This would allow customers to implement
Comments 0