저자: DanIves · 외국인
요약 보기
한줄결론: 애플은 메타와의 AI 협력으로 AI 시대에 본격 진입하며 장기적으로 서비스 매출 성장 수혜가 기대되고, AI 인프라 수요는 여전히 강하다는 낙관론이다. AI 파트너십: 애플과 메타는 서로 필요로 하는 관계로, 뮤즈(Muse) 관련 하드웨어 통합이 향후 12~18개월 동안 가속화될 것이라고 전망했다. 이번 협력이 애플의 다른 AI 제휴 가능성을 완전히 막는 것은 아니지만 비공식적으로는 제약이 될 것이라고 봤다. 애플 주가 및 가치: 애플은 올해 나스닥을 18%p 상회했다. 애플이 AI 경쟁에서 늦었다는 우려가 있지만, 아이폰 15억 대와 iOS 기기 25억 대의 설치 기반을 바탕으로 소비자 AI 접근의 통행료 징수자가 될 것이라고 주장했다. 전 세계 소비자의 20%가 결국 애플 기기를 통해 AI에 접근할 것으로 추정했다. AI와 수익화를 통해 장기적으로 애플 주가에 75달러 가치가 더해질 것이라는 계산을 제시했으며, 서비스 매출이 향후 수년간 월가 예상보다 약 15% 더 성장할 것으로 봤다. 반도체·메모리: 최근 한국 방문 결과 메모리 칩 수급비율이 13대 1로 수요가 공급을 크게 초과한다고 밝혔다. 한국 시장과 미국 반도체주가 지난주 큰 압박을 받았으나, 하이퍼스케일러 실적에서 수요 가속이 확인될 것이라고 봤다. 이번 주와 다음 주 실적발표가 중요하다고 강조했다. 코어위브·네오클라우드: 자금조달 우려에도 수요가 매우 강해 백로그가 크다고 평가했다. 시장이 우려를 과장했으며 코어위브, 네비우스 등이 매수 기회라고 봤다. 다만 마이크로소프트, 아마존, 구글 등 하이퍼스케일러가 네오클라우드보다 한 단계 위에 있으며, 부채 부담이 있는 오라클은 도전 과제가 있다고 지적했다. 리스크: 자금조달 비용 상승과 부채 부담이 리스크 요인이다. 메모리는 역사적으로 경기순환적이며 2029년경 공급 과잉 가능성이 제기되고 있다. 미드텀 선거를 앞두고 변동성이 높을 수 있다. 출연자 전망: AI 인프라 개발은 10~12년 주기의 3년차이며, 팔란티어·스노우플레이크·서비스나우 등 사용 사례 확산이 주가 상승의 관건이라고 봤다. 앤듀릴과 앤트로픽의 450억 달러 규모 클라우드 컴퓨팅 거래 보도도 언급했다.
다듬은 글
Yes, I think this is just the first of many potential partnerships between them, because they both need each other. When you think about what happened with Anthropic and OpenAI, it actually brings these two giants closer together. And I think the hardware integration that we're seeing with Muse is something that I think will accelerate over the next 12 to 18 months because these companies need each other.
I think Ternus understands, and it's very smart, that we need to put aside our doubts and move forward, and I think this is just the beginning of a larger collaboration between Meta and Apple. Does this rule out Apple collaborating with other companies, or is the field of possibilities now open? Yes, Guy, I would say this: maybe not officially, but unofficially, I think it blocks them.
I think this really shows that Apple is going to bet quite heavily on Meta. But it all comes down to Muse. That's why it's so important, what kind of technology is this, I think it's a real breakthrough in the field of AI. They understand that now is the time, especially given their conflicts with OpenAI and others, that this is just the beginning of the AI era for Apple .
Dan, when I think about AI and what's been happening in technology over the last year, I don't think of Apple, but Apple has outperformed the Nasdaq by 18 points for the year. My question to you is: what does the market see that most people, except you, are not so optimistic about? And do you think the discussion about the new product that has emerged in recent days, Apple Home, changes anything?
Again, home products with built-in AI. Yes, that's a great point. I know we've been talking about this on air for the past 6 months: is Apple going to be late to the AI party ? Without a doubt. They've stumbled in recent years, but with 1.5 billion iPhones and 2.5 billion iOS devices, that's a goldmine of installed base .
That is, they will essentially become toll collectors on the consumer AI highway. We estimate that 20% of consumers worldwide will eventually access AI through an Apple device. This is what the market sees. Together with Turners, we believe that AI and monetization will add $75 to Apple's stock value in the long term.
I mean, Apple has benefited from not being part of the so- called AI complex in recent days. And I just wanted to get your thoughts on what's happening with Coreweave and the investors' doubts about the ability to fund this expansion. We heard comments from Broadcom CEO Hock Tan, who believes that one of the limiting factors for development could be access to capital.
David Solomon said that not everyone with a chip in a data center will have access to cheap capital. Are we at a point now where we should start to doubt the ability to finance this? Yes, financing, especially considering profitability, will be a risk factor, but I still go back to demand. I literally just got back from Korea, and on the memory side, we're seeing a 13-to-1 supply-demand ratio for chips.
And as that grows, combined with what we're seeing in the earnings reports from the hyperscalers, the Coreweave backlog is just huge. So I'm not saying there won't be pressure , but I think the market has greatly exaggerated here. I find this very similar to the fears of the “ SaaS apocalypse” we are seeing.
I just think these are opportunities to buy shares in some of these companies, if you look at Nebulous, Coreweave and other technology firms. Dan and Karen, thanks for joining. Let me go back to what you said, which is the $75 figure that you mentioned. How did you come to this? Tell me how you calculated it.
What do you think about this? Yes, my view is that 15% of service revenue growth will be driven by AI in the coming years. When you think about developers, the heart and lungs of Apple, the monetization story is just beginning, given their ability to monetize their installed base and developers using that as a platform.
This also shows what they do with the news. I think Wall Street is largely underestimating services revenue, probably by about 15% in the coming years. I think this will happen through monetization. But the last few weeks have been turbulent. We were in Korea last week trying to better understand the memory situation and demand there.
But listen, it all comes down to this week and next week in terms of earnings reports. I mean, this is ultimately what Wall Street needs to see regarding hyperscalers and big tech companies. What did you take away from your stay in Korea? Obviously, this is an important point: we have seen their markets under tremendous pressure, and there are many companies in the semiconductor sector that have been quite vulnerable last week, as have American chipmakers.
But what did you learn from what you saw and heard? Yes, there's definitely nervousness, but demand exceeds supply for memory chips by a ratio of 13 to 1. So the way I look at it is this: you and I have been talking about this on the show for years, that we're going to go through these kinds of narratives.
Remember the “ SaaS apocalypse” earlier this year. Look what's happening here. I'm not saying it doesn't make sense, of course there is some validity to this nervousness given what we've seen with the fundamentals, even bonds and CDS spreads. But there is actually an acceleration in demand for chips, from hyperscalers, use cases, and that's what I think the reporting season will show.
So, how important is this week for restoring confidence in AI investment? I think this is very important because in a vacuum, when there are no numbers, different narratives, different reports can emerge, and I think this can sometimes dictate conditions, and it throws investors off track because there is a sense of nervousness.
We are approaching the midterm elections. During this period, there is always high volatility . But in the end, it all comes down to fundamentals. What is the demand for chips? What do enterprises do in interaction with hyperscalers? And what do the use cases look like? Palantir, Snowflake, ServiceNow and many others.
And that's what will determine whether stocks go up. And I still believe that we are in the third year of a 10-12 year development cycle. So what does it mean that we're in the third year of a 10- year development cycle, Dan? Here, for example, is Anduril. A headline appeared. Anduril and Anthropic are in talks for a $45 billion cloud computing deal .
So, essentially , Anduril will sell computing power to Anthropic. And this deal could be worth $45 billion. Interestingly, it could be finalized in the coming days, as reported by our Billy Lipschutz, Dina Bass, and Brody Ford. But this comes after we just published a story about Anduril's $30 billion valuation being called into question.
Investors, amid concerns about AI, especially neo-clouds, believe their value could be half that. So, what do you think about neo- clouds versus hyperscalers in terms of their security and role in the long term? Yes, I think hyperscalers are a whole different level. And this applies not only to their balance sheet.
In addition, look at the companies, their place and position in the strategic development and expansion of data centers. This is why I believe Microsoft, Amazon, and Google remain in a completely different stratosphere. Oracle is a good example of trying to become a hyperscaler, but it comes with challenges, right?
When you take on debt, you have to cut expenses. If you look at companies like CoreWeave or Nebius, they will become kind of champions among neo- clouds. But I think they need to be distinguished. And look, in the game of musical chairs, will anyone be left without a place? Of course. But I think it's still very early compared to where we are now .
If you look at demand, capacity, and where everything is heading, even if you exclude some of the data centers that won't be ready. rading AI assets faces a serious challenge. Nvidia CEO Jensen Huang defends the long-term prospects of artificial intelligence, even as investors question funding and spending within the AI bubble.
At the GTC event in Berlin, Huang addressed two major concerns looming over the tech sector. First, can the fastest-growing non- cloud companies afford further expansion? And second, can the memory chip industry eventually create more supply than the AI market needs? His answers were confident, but the market reaction suggests a more complex picture.
Because now they are asking not only about whether the demand for AI is real. They ask who can afford to meet this demand. How quickly will these companies be able to turn costs into profits and whether the current valuation is too much of a success story. Let's break down what's happening and what it could mean for Nvidia and the broader AI market.
Jensen Huang defended the neo-cloud business model . One of the biggest pain points in the AI market is the neo-cloud industry. Companies like CoreWeave and Nebius are building a business around renting computing power to customers who need access to powerful AI chips. Instead of building all the components of their own infrastructure, customers can rent computing power from specialized providers.
The opportunities here are huge. Training advanced artificial intelligence models and running AI applications requires enormous computing resources. But there is a certain nuance. These companies need huge funds before they can provide such computing power. They need to raise funding, purchase expensive graphics processors, build data centers, and get enough electricity to run the equipment.
And it will be a pretty complex financial equation. Demand may be high, but companies still need capital to grow. And as concerns about debt financing grow , investors are forced to question whether this business model can deliver attractive returns. Jensen Huang was asked directly if these companies have a fundamental problem.
His answer was clear. He believes that demand is extremely high, and the real constraint is access to working capital. In his opinion, new cloud providers are providing valuable services to customers in many areas of AI. The problem is that the amount of capital needed to support this growth exceeds the available funds.
Huang believes this gap will eventually close due to the strength of underlying demand. And this is an important distinction. It does not describe an industry that lacks customers. He describes an industry that needs enough funding to build capacity for these customers. However, there is another side to this argument.
High demand does not automatically guarantee business success. Companies still have to manage interest costs, debt repayments, equipment depreciation, energy costs, and the risk of falling utility prices over time. A business can have a queue of customers and still struggle financially if the cost of servicing those customers is too high.
That is why the market has become particularly sensitive to the financial structure of new cloud companies. Nvidia also has investment and commercial relationships within this ecosystem. And that makes the success of these customers important to the overall story of AI infrastructure. If new cloud companies continue to expand, they could become major buyers of Nvidia hardware.
But if financing becomes more expensive or harder to obtain, their procurement plans could come under pressure. The question is no longer whether the world needs more AI computing power. The question is whether the companies that create them will be able to finance their growth without dangerous financial risks.
The memory chip industry has faced its own challenge. The second important issue discussed in Berlin was memory. AI assistants need more than just powerful processors. They also require huge amounts of memory to transfer data quickly and support the efficient operation of increasingly complex models.
And this has made companies like Micron, Samsung, and SK Hynix important players in the AI infrastructure market. High-bandwidth memory , known as HBM, has become particularly important for advanced AI accelerators. As AI systems become more demanding, memory performance can become a major factor in overall computing speed.
But the success of the memory industry has given rise to new concerns. Could the industry eventually create too much capacity? A report from South Korea indicates the possibility of a memory oversupply around 2029. And this prospect is making investors wonder whether the current boom could eventually turn into another trading cycle.
Memory has historically been a cyclical business. When demand increases, manufacturers invest in additional capacity . If too many companies expand at the same time, supply can eventually exceed demand, and prices then fall. Profits are weakening and investors are overvaluing the industry.
Comments 0