저자: DanIves · 외국인
요약 보기
AI 투자가 2단계(인프라)에서 3단계(수익화)로 넘어갔다는 주장. 메모리(HBM) 수급은 2년간 12~15:1 수요초과가 이어지며 SK하이닉스가 대표 수혜주. 애플은 소비자 AI 관문, 팔란티어·스노플레이크는 기업 AI 수혜. 오라클은 부채·설비투자 부담으로 주가가 부진하나 다음 실적 발표(FY26 2Q)에서 AI 매출 전환을 보여주면 반전 가능. 나스닥 3만pt 전망 언급. 근거: HBM 수요/공급비 12~15:1, 오라클 FY25 잉여현금흐름 -237억달러, RPO 6380억달러(영상 수치 그대로), 2027년 공급 최대 난관(SK하이닉스 CEO 발언).
다듬은 글
Today we are going to break down the latest interview with Dan Ives, Wall Street's top tech analyst. In this interview, he explains the entire AI investment landscape—from hardware to software, and from Korean memory chips to the debt issues of tech giants— laying out a very complete logical chain. Let's take a look at his take on SK, what it might mean, what it might not mean, and maybe what it means for technology and the memory space in particular.
Yeah, I think it's a watershed day. I mean, not just for SK, but I think for Korean tech and in terms of what you see in the memory market. Because look, it speaks to someone like myself who spends so much time in Korea— equilibrium demand won't happen for another two years. So, we're going to continue to be in a 12-to-15-to-1 type of demand-supply environment.
That's what you're seeing now. The AI revolution continues to be in the third inning, one out, relative to this AI game. You were in the first inning not that long ago, Dan. Now you just said third inning. So what's still to come then? Take us to the rest of the game. What are the plays? What are the companies and stocks to watch?
The rest of the game, I mean, we'll talk about Apple in a second, but now the consumer AI revolution starts. It's going to go through Apple. They're going to be the collector. And then you look at the enterprise use cases. That's where software, we look at names like Palantir, Snowflake, and even though right now I've got to talk about Microsoft, Oracle, and others, it's time for them now to get into that game in terms of monetization.
It's a very, very important earnings season in terms of 2Q, which spills into cybersecurity as well. It speaks to our view. Oracle has been horrible since you remember— I know you remember the stock went up 40%on the back of that announcement— it's been horrible ever since the report in September. So we're obviously a ways away from that; we're through things.
What's going to be a catalyst here? Because it's pretty clear to me that Larry Ellison has pushed the chips . It comes down to perspective. They've got to show ultimate conversion RPO, not just in terms of open terms, but more and more of those deals. I think that's going to give the Street more comfort because look, it's a "New York City cab driver" scenario right now.
I mean, they are as much in the penalty box as anyone, but I think that's ultimately the wrong move because they will be able to monetize, and it's just like what Microsoft and these others are doing now. People are worried about Oracle's balance sheet. They're worried about the debt levels. I mean, that's really what it comes down to, I think, with Oracle.
Yeah. I think to that point, if you look at their debt load, it's still a small piece relative to their overall capital structure. They've already talked about the 45 to 50 billion and how they're going to raise it. Now they're going to have to do more. But I think if you look out, you've got to look at these stories over the next 12, 18, 24 months.
That's the key. But for all the ones that are ultimately funding the AI revolution, they're in the penalty box—Oracle, Microsoft, and others. That's why it's a very important earnings season to start to show that trend change. Let's first talk about SK Hynix. Ives mentioned on the show that this is a " watershed day." Why use such a heavy term?
We need to understand an underlying logic: AI training and inference require more than just GPUs. NVIDIA's chips are the engines of calculation, but data transport relies on memory chips. Especially HBM (High Bandwidth Memory). To use an analogy: if the GPU is the chef, HBM is the conveyor belt for serving the dishes.
No matter how skilled the chef is, if the conveyor belt isn't fast enough or abundant enough, the entire restaurant will be paralyzed. Equilibrium in the memory market won't likely arrive for at least another two years. Before that , the ratio of demand to supply will remain between 12: 1 and 15: 1.
What does this data imply? In the normal chip industry, a supply-demand ratio of around 1.1: 1 is already considered a tight shortage. A 12 to 15: 1 ratio means that for every unit of HBM capacity produced in the market, there are 12 to 15 units of demand waiting in line. Two days ago, the CEO of SK Hynix personally confirmed this assessment.
He stated that 2027 will be the most difficult year on the supply side; the total supply will be unable to keep up with the explosive growth of AI. So, Dan Ives calling this a "watershed moment" is actually not an exaggeration . It validates an investment logic: memory chip stocks—like SK Hynix, Samsung's semiconductor division, and Micron—are in a golden window of an extreme seller's market for the foreseeable future.
Then, the host of the show immediately followed up with a good question. "Ives, didn't you say earlier that AI was still in the first inning?""Why have we jumped to the third inning now?" Ives's answer is critical. The first inning was the infrastructure construction phase— buying GPUs and laying fiber optics.
The winners in that stage were those selling the 'shovels,' like NVIDIA, Broadcom, and TSMC. So, what is the third act? It is the stage where AI starts to monetize, specifically divided into two tracks. The first track is the consumer side, which is Apple. Davis says that Apple will become the toll booth for the consumer AI revolution.
Today, in 2026, this statement seems less like a prediction and more like reality. Just last month at WWDC 2026, Apple announced a brand-new OS; AI has been completely rebuilt to understand context, perform cross-app operations, and offer proactive recommendations. The Apple Intelligence integrated into the iPhone 17 Pro is already deeply embedded in everyday usage scenarios.
Why do we say Apple is a toll booth? Because it has 2 billion active devices; it doesn't need to build its own large models, it just needs to be the gateway, letting users access AI through Apple devices, then collect tolls via the App Store, subscriptions, and cloud services—it is the pipeline owner of the AI era.
The second track is the monetization moment for enterprise software companies; Davis named Palantir and Snowflake. Why them? Palantir's core capability is deploying AI models into real enterprise business workflows, not just giving you a chat box, but embedding directly into your supply chain decisions, customer analysis, and risk control systems.
Its AIP platform is already helping the U.S. military and major enterprises with practical AI deployment. As for Snowflake, it is the data infrastructure; for AI models to be effective, your enterprise data must first be clean, usable, and functional. Snowflake is the player that gets data ready for AI to use.
Both companies are seeing strong growth in 2026, and Snowflake and Palantir even announced a strategic partnership this year to unite their data pipelines. So, Davis's logic is that after making money in the first act by buying hardware stocks, the third act is for deploying capital into companies that use AI to make money.
The most brilliant clash in the interview was the discussion on Oracle stock; the host pointed out directly that since the earnings report in September 2025, the stock surged 40%, only to fall continuously ever since. That week in December 2025 was Oracle's worst week in 25 years. As of now in 2026, the stock has barely risen 3%, significantly underperforming the broader market.
Davis was very blunt, saying every investment bank analyst in New York is currently bearish on Oracle stock. So, what is Wall Street afraid of? The fear is that Oracle, in its pursuit of the AI wave, has borrowed too much money; let’s look at the data. In fiscal year 2025, Oracle borrowed a total of $ 43 billion in debt, plus $ 5 billion in equity financing.
Capital expenditures surged by 162%, reaching $ 55.7 billion. The full-year free cash flow was a negative $ 23.7 billion—negative, that is. And then, they stated they plan to raise another $ 40 billion, with $ 20 billion of that coming from issuing new shares. When the earnings report came out last month on June 10th, although revenue grew 21%, cloud revenue soared 47%, and RPO—remaining performance obligations —hit an astonishing $ 638 billion, the stock still dropped 10%after hours.
Why? Because investors aren't looking at whether revenue is good or bad, but at how much more you need to spend and borrow. What is Safra Catz's counter-logic? She made three points: first, Oracle’s debt, compared to its overall capital structure, remains at a manageable ratio. The $ 638 billion in RPO represents a massive amount of future revenue.
Bank analysts estimate that over 50%comes from OpenAI contracts, which aren't empty checks but cold, hard cash in advance. Second, many clients are already prepaying GPU fees to Oracle, or buying their own GPUs and handing them to Oracle for hosting. This effectively alleviates the capital pressure on Oracle to build its own data centers; the third point is also the most important.
You have to view this matter through a 12-to 24- month horizon. Safra Catz made it clear : all companies footing the bill for this AI revolution are currently in this "quiet period." Oracle is, and so is Microsoft. I think this perspective is correct. Microsoft went through a similar phase; when they first started pouring massive amounts of money into OpenAI and building data centers in 2023, Wall Street was also very worried .
But once the revenue growth proved the return on investment, the stock price took off. Microsoft’s latest quarterly report, for the third quarter of 2026, shows that enterprise AI is now deeply embedded in their product lines, and AI growth remains strong. Oracle is on the same path: invest first, withstand the skepticism, and then prove yourself through revenue growth.
The difference is that Oracle has a lower starting point, higher debt , and a tighter time window, which is why Dan Ives says the next three to six months are a critical period for them to prove themselves. To piece together the logic of Dan Ives 'entire interview , he actually provided a very clear framework.
First is underlying hardware and memory chips, such as SK Hynix and Micron, which will remain in a state of extreme supply shortage for the next two years, making them one of the AI sectors with the highest certainty. Second is the consumer-facing portal; Apple is the toll booth for AI to reach 2 billion consumers.
It doesn't need to build the base models; it only needs to perfect the last mile. Third, enterprise software companies like Snowflake are starting to truly help businesses make or save money using AI; they are the protagonists of this third phase. Fourth, the infrastructure builders like Oracle and Microsoft are currently being punished by the market, but if AI demand is real, then the monetization for these companies is just a matter of time.
The keyword is time. Wall Street's fear has never been that AI isn't viable, but rather, when will the massive amounts of money spent actually be earned back? Dan Ives' assessment is that the upcoming fiscal Q2 2026 earnings report will be a critical turning point. If companies like Oracle and Microsoft can demonstrate more AI revenue conversion and the realization of ARPU into actual earnings in their reports, then market sentiment will reverse.
Moomoo is a brand under a US-listed company and ha can click the link in the comments for a chance to get up to $ 100 in stocks, and enjoy up to 8.1%returns on idle funds. . Lastly, let me share my own view. Ives is one of Wall Street's most optimistic tech analysts, and he predicts the Nasdaq will reach 30,000 points.
He has always been a major bull, so you should take his words with a grain of salt. However, the logical chain in his interview today—from hardware supply gaps to consumer and enterprise AI, and then to the AI capital return cycle— is self-consistent. Especially with the SK Hynix CEO confirming the reality of the extreme supply tension just two days ago, it makes the foundation of this logic much more solid.
So, we are now entering a dense earnings season, and the numbers from Apple, Microsoft, and Oracle will be released one after another. We'll come back then to verify how the third game really went as Danfoss described; for more videos, click the link below to join the channel membership.
Comments 0