저자: FinVid · 외국인
요약 보기
영상은 나스닥이 1.5% 반등했음에도 전날 ETF가 큰 손실을 본 배경을 짚으며 시작한다. 헤지펀드 운용 축소 보도와 개인 투자자 매도가 겹쳤고, 데이터센터 정치화와 지역 반대 여론이 칩 수요 전망에 대한 의문을 키웠다는 해석이 제시된다. 진행자는 메모리 가격 상승과 공급 축소로 2027년 호황을 기대하는 시각을 소개하면서도, 실적 확인이 필요하다고 강조한다. 최종 소비자의 AI 도입은 계속되고 있으며, 하이퍼스케일러들은 데이터센터 투자를 줄이지 않고 있다는 주장도 담겨 있다. 주가 측면에서는 엔비디아가 선행 이익 대비 약 14배, 마이크론이 약 4배에 거래되며 밸류에이션이 비관적 심리를 반영한다는 분석이 나온다. 마이크론의 1,000억 달러 규모 자사주 매입 발표도 언급된다. 진행자는 AMD의 데이터센터 매출 97% 성장과 점유율 확대 가능성을 근거로 엔비디아 지분을 실적 발표 전에 줄였다고 밝힌다. 이어 TSMC 3D D램 보도, 마이크론 대만 공장 가동 정상화 공지, SK하이닉스 HBM4 수율, 마이크로소프트의 오라클 계약 종료와 GPU 수명 논란, 오픈AI 펀딩, 앤트로픽 IPO 일정 등을 정리한다. 아이런(IREN) 실적 분석도 포함되며, 연간 반복 매출 목표가 2027년 초로 밀린 점과 계약 단가 상승이 함께 다뤄진다. 마지막으로 컴퓨팅 부족이 2028년 이후까지 이어질 것이라는 낙관론과 피지컬 AI 성장 전망이 제시된다. 영상은 모든 예측이 가정에 기반하며 투자 판단은 본인 책임이라는 점을 반복해서 밝힌다.
다듬은 글
Dave, let's go back to the broader market. The Nasdaq is showing some recovery, up one and a half percent, but your ETF obviously suffered significant losses yesterday. What do you tell your investors when they are looking at the stock you own and are worried about the future of the business? Yes, we had a lot of conversations with investors yesterday, answering your question directly, because two things happened yesterday .
The first is a bit of technical leverage. Of course, there have been reports of hedge funds either closing down units or liquidating entirely, coupled with private investors making the largest number of sales since COVID. Then there are the fundamental questions you talked about regarding the politicization of data centers.
If we predict the need for additional chips for them, and they are not built or construction is postponed to the future, investors begin to overestimate the fantastic growth that these companies, in particular Micron and others, have demonstrated and that is projected in the future. Okay, what is your opinion on this ?
I mean, it's still early, and we heard Sarah talk about it too. At the moment, much remains unclear, despite the "blue wave" and many referendums held against data centers in certain areas to understand what it all means. How do you feel? Yes, that's interesting, right? The market has already started selling off, waiting for a "blue wave." What he didn't expect was a "blue tsunami," especially at the state and local government levels.
And we know how unpopular data centers have become, given the results of the votes against them. But interestingly, if you dig deeper and look at the policies that Texas intends to implement with their newly elected governor, it's actually not that different from what Abbott was talking about regarding the current moratorium.
And I think investors will need time to realize this and see where it all comes down to. But the fundamental story behind it has not yet been completely disproven. This raises some questions, and this uncertainty is the reason why investors first sell and then reassess the situation. OK. And finally, Dave, memory prices continue to rise, capacity is contracted.
So, what do you expect next year, in 2028? Has anything changed at all? Yes, that's interesting. The market has already begun to predict and anticipate that 2027 will be a great year with an eye on 2028. But it's interesting. If you look at the peak we saw in DRAM and memory stocks earlier this year, the market was really expecting incredible growth.
He now expects a 50-60% decline in revenue, which opens up opportunities for profitable investments as these stocks are trading at multiples we have n't seen in the last 5 years. Now the question arises: are investors ready to get into this today? We clearly see that it is. However, I would note that a market bottom is a process.
It doesn't happen overnight, but eventually, it will be a story that requires evidence. These companies, and Micron is reporting soon, as is Nvidia, will have to show stable demand for investors to continue to believe in this story, and not just wonder what 2029 or 2030 will look like . So, your thesis, your thesis about optimism remains completely valid?
Yes, because we always come back to final consumer demand. Are Citi, JP Morgan, Pfizer, Walmart, Unilever—all these companies implementing AI—going to slow down its implementation because of what happened on Tuesday? No. It does n't look like it. I mean, Jane Fraser didn't think so, and she said it was happening all over the world.
But is there a possibility that these data centers could be built outside the US? Certainly. And that, you know, that's always been an important part of that process. How they are built will change, where they are built will change, and how quickly they are built may also change. So it's all part of the process.
But none of these companies—Microsoft, Google, Amazon— are cutting their capital spending plans because of this. On the contrary, these companies will likely accelerate as they will want to take on more of this construction to ensure that everything is done correctly. Because, to be clear, people have no problem with data centers being built correctly.
They have problems with open water supply systems. They have problems with on-site methane generation . They have issues with projects that demolish 2,300 acres of farmland . This, in fact, is the real problem. If you build them right, as many of these large hyperscale companies have been doing for some time, this is not a problem.
No, but it definitely creates more friction, more regulatory risks, potentially longer timelines, and higher costs. With the increase in prices, everything is simple. If the only issue is that it costs more, then that's okay, because the construction part, if done correctly, is a very, very small fraction of the total costs.
Land , electricity, and frames—that's about 15 billion out of a budget of 60 billion. If it goes up by 10-20 percent, these companies will calmly cover it, because everything is getting more expensive anyway. I mean, memory prices are going up, connectivity prices are going up too. These are simply costs of doing business, and the profits these companies make remain high—we've seen the cash return on invested capital for hyperscalers.
Last quarter it was 30 percent. And that doesn't change. So do you see? The stock was up over 100 percent by June, and now it's up only 27 percent. The market sensed this in advance. I mean, yes, yesterday's falls in Micron and Nvidia stocks were stunning, of course; but they were, and perhaps are, the final chords of a correction, or even not a correction, but a " bear" market for these semiconductors, which now values them adequately for profits that may well decline in the next few years due to the slowdown in AI development.
But these valuations ...now we see Nvidia, it seems, with a multiple of 14 to forward earnings . And Micron, I think, with a factor of four to earnings. Maybe these profits will decrease, maybe that's good? But these multipliers now reflect sentiments that, in my opinion , are simply too pessimistic.
So I think that for certain stocks, now is the time to choose investments. But take Micron, for example. You know , you can look at this in two ways. You could say, "Yes, they fell by 50%." You could say, "They still grew 200% in a year." You could say, "Okay, we think data centers will only lose 10%," or you could say that eventually there will be competition, or maybe the losses will be more than 10%, but that's why stocks are still trading at a reasonable multiple—because people are very skeptical about real future earnings growth.
And it's just my feeling, having lived through the dot-com era, it's my feeling that we have n't gone far enough yet. And I think it's worth remembering some important calculations here: well, of course, the market looks great today. Green everywhere , right? But when stocks like Micron fell 18 or 17% yesterday, they have to rise almost 20% to make up for that.
They only grew by six. So I don't see today's movement as anything very encouraging in terms of changing that momentum. We all know I'm not a momentum trader, so maybe you shouldn't listen to me on this, but I think people are starting to get interested in valuations. The numbers are not the same. It doesn't feel right, you know, to achieve that kind of profit growth.
demand for memory, and it will be. There is a dramatic imbalance between supply and demand. And he will be. There is currently nothing to indicate that it will not be there. That's great, and it's true. But what will Micron's earnings growth rate be in 2027, 2028, and 2029 for sure? And no one can answer that right now.
This is where it gets tricky. want to talk about Micron, let's not forget that they are going to buy back $100 billion in shares. For a market capitalization that has become significantly cheaper—$750 billion. This is a huge amount of shares. This is a valid observation. so when we make these inevitable comparisons with the 1990s, we must distinguish where the situation is different.
Micron has no debt on its balance sheet. I mean, this is different from any other cycle that Micron has been through . There are no debts. They generate a ton of free cash flow . Yes, Jenny, there's a question about the growth rate, but at a price that's literally four times the earnings next year, that's fine with me.
You Uh , since I mentioned what happened with Nvidia this week, you cut the share. Tell me more. By the way, before the earnings reports, which seem to be on the 17th. Yes, I actually cut it for the third time this year, and I know you would assume that's probably because when the market seems to get frustrated with AI trading , the company that's at the center of all the AI-related spending and all the AI-related construction is going to be the first to come under pressure; but really it's because there was something about AMD's earnings report that I just didn't like.
So, AMD reported 97% year-over-year growth in data center revenue, which is a really big positive for the entire semiconductor industry and the overall state of the AI sector. But they also provided some projections that suggest they expect to capture an increasing share of AI infrastructure spending in the future, in an area that Nvidia has dominated.
They added a bit of detail about the predictions and contracts they have with model developers like Anthropic, OpenAI, hyperscalers, Microsoft, etc.; they have actually proven that they are capable of coming in and potentially crowding out Nvidia, biting off some of their market share. And I understand that $90 billion in revenue per quarter in data centers is Nvidia's figure, versus about $9 billion for AMD.
So I guess I'm comparing watermelon to grapefruit, but the very fact that they have the potential to become a significant competitor to Nvidia is a bit disconcerting to those who have been thrilled quarter after quarter with their controlled market share and consistently high gross margins. So, in agreement with you, I'm selling the stock before the earnings report in case the market decides again that it does n't like what it hears from Nvidia, because for the past three quarters after every report the market has been saying, "Yeah, that's great, but it's not good enough." What else can they tell us?
So I think this is a good time for risk management for those who, like me, have "taken too much risk" with Nvidia. I cut the position twice at the beginning of the year , but it continues to grow like a weed. So I think it's time to cut back while the market, as I said, is generally cooling down across the sector.
Kevin. I think the big difference when we talk about earnings is that Nvidia used to rise before the reports , and then fall after the news. If you look at Nvidia right now, Malcolm, it's a stock that's at best moving on inertia, or maybe even limping ahead of the report. I'm not against the idea of risk reduction.
I think it's fantastic. You are shorting the position again, and that makes sense. I would like to see Belsky reduce some of his position in Neo Cloud, but that's a conversation perhaps for another day. But I look at this NVIDIA reporting season in a completely different way than I did before. At least, I definitely hold the position, and we don't plan to reduce it .
You say you don't look at it the same way you used to, why? Because of market conditions? Yes, I think the whole trade has changed because of data centers. Let's say we're talking about 10% fewer data centers, 15% fewer data centers. It doesn't really matter. This changes the economics for each of these companies, regardless of their focus.
So for me you just have to lower your expectations. The talk about Micron was appropriate. They may be sold out by the 28th, but Jenny, you mentioned the 29th . I think peak sales are probably already here. So I still believe in NVIDIA. I still think it's cheap. We own it, but I'm not sure that after reporting it will have the same problems as before, in the quarters that we saw three, six, nine months ago.
And for me, that's a big difference between NVIDIA and some other names. And we will see how events develop . Yes, I apologize. Keep going, Malcolm. I'm just on Kev's point, I think we should at least consider what increased competition for NVIDIA would look like going forward. So the Helio system is what I mentioned earlier, allowing them to run software and hardware under one roof.
This seems very similar to what Nvidia offered with CUDA, which was what kept their customers in this stable ecosystem. So, with the companies that I just mentioned that they've already started contracting with and doing business with, they're ready to step in and confirm that there's potential for AMD to squeeze in and become a significant number two, Nvidia hasn't really had to worry about competition for the last four years since we've been talking about this AI trend.
And that's why I say that before their report, it's time to think about risk management. So, Nvidia is another company where the P/E multiple is, I would say, undemanding. And so even if growth were to slow for a company like Nvidia, I think the market, one, is partly anticipating that today, and two, you have this broader platform and a company that has a track record of investing and executing product cycles in challenging times.
You feel more confident choosing Nvidia, despite everything going on around you, right? I like Nvidia at its current level because it's not just about the silicon and systems, but also about their broader ecosystem . Millions of developers build their solutions on the Nvidia platform. And, in my opinion, that remains a decisive factor when I think about the next few years.
So, I hope you are doing well and staying calm in this market. Friday was a mixed day in the market, but a positive one for many AI infrastructure stocks, which fell sharply during Thursday's sell-off. On Friday morning, we received better-than-expected nonfarm payrolls data, which strengthened expectations of further Fed rate hikes.
The Information published an article stating that TSMC is considering expanding into advanced memory manufacturing and plans to manufacture 3D DRAM at the request of Nvidia. I must make a disclaimer: this source has previously been unreliable regarding some rumors about Nvidia, so I don't know if this story is true.
I'm just mentioning this so you're aware. If TSMC were to successfully commercialize 3D DRAM, the company could eventually find itself in competition with current DRAM manufacturers. However, if the report is true, we don't know what products this 3D DRAM would be intended for. It is important to note that 3D DRAM is essentially a method of manufacturing memory, not a specific type of product like HBM.
However, 3D DRAM could be used to increase the capacity of individual DRAM chips in an HBM stack, potentially increasing the amount of memory per GPU. Importantly, commercial production of 3D DRAM by TSMC is currently unconfirmed, and we do not know important details such as production cost, yield, and timeline.
And again, even if this is true, we don't know exactly what products Nvidia wants TSMC to produce 3D DRAM for. It can be used in many different products, so let's see what happens next . We have an update on the story I mentioned in the video on Thursday night. Bloomberg published an article stating that Micron has suspended work on a $4.5 billion memory plant in Taiwan due to a labor dispute.
The news added fuel to the tech sector's sell-off on Thursday, as some market participants mistakenly interpreted it as a sign that Micron would cut production due to concerns about a slowdown in AI development after the US midterm elections. As I said on Thursday, this doesn't seem to be true, but that was the immediate reaction of some market participants to the news.
And on Friday, we received an update to this story: Micron reported that its operations in Taiwan were continuing as usual. Micron did not directly comment on the Bloomberg article, but said in a statement that its operations in Taiwan are continuing as usual and the company is confident it can maintain stable operations without disruption to customer service.
In other news: SK Hynix's cumulative yield of HBM4 usable products has reportedly reached the mid-60% level. HBM4 is expected to surpass HBM3E in both shipments and revenue in the fourth quarter of 2026. This is, of course, good news for SK Hynix, as well as for NVIDIA , as SK Hynix is the largest HBM supplier to NVIDIA, despite Samsung gaining an initial lead in HBM4 over SK Hynix and Micron.
Recall that the NVIDIA Rubin architecture uses HBM4. Another story that attracted a lot of attention on Friday was a Bloomberg report that Microsoft is ending its deal with Oracle for AI computing after deciding not to renew the contract. The initial deal was announced in 2023 and involved Nvidia A100 GPUs.
Recall that this graphics processor was released six years ago . According to the report, the outdated GPUs no longer meet Microsoft's ever-increasing technical requirements, so the company has been gradually decommissioning leased capacity over the past few months. This news has caused market participants to question the useful life of Nvidia GPUs.
It is important that these GPUs were used to support the AI version of the Bing search engine. Interestingly, Microsoft's SEC filings indicate that the estimated useful lives of the company's servers and networking equipment range from two to six years. So this looks like a situation where Microsoft is choosing newer hardware because it is more attractive from both a technological and economic perspective.
This seems to be what is happening here, rather than a decline in demand for computing power. An Oracle representative reportedly said that demand for the company's AI infrastructure remains high, including for previous-generation GPUs. Recall, Oracle management reported during the last earnings report that GPU capacity whose lease term expired during the quarter was extended or resold at a 20 % premium to previous contracts.
Most of these GPUs were four years old or more. So even if Microsoft doesn't need six-year-old GPUs for relevant workloads, that doesn't mean Oracle can't find another customer willing to rent that capacity. If the world is limited in computing resources, developers will use whatever available power they can get their hands on .
Bloomberg reported on Friday that OpenAI has raised about $11 billion in new funding, the initial part of a round that values the company at $125 billion . The rest of the round could bring the total to $20 billion and is expected to close in the coming weeks. OpenAI originally planned to raise $30 billion or more at a valuation of $1.4 trillion .
If you invest in AI infrastructure stocks, you want to see how companies developing advanced models continue to demonstrate their ability to raise capital. With that in mind, The Information published an article on Friday stating that Anthropic plans to begin its IPO roadshow next week ahead of a potential public listing the week of November 16th.
In the short term, this is probably the most important event for AI stocks. I really wonder if we will see a pre-IPO sell-off in AI stocks similar to what we saw before the SpaceX IPO. But perhaps some of this sell-off has already taken place in recent weeks, given the recent decline in many AI stocks.
Anthropic's ability to raise capital and meet its spending commitments is critical to the entire AI supply chain. So when it comes to market sentiment on AI infrastructure stocks in the short term, investors want to see this IPO succeed, given all the recent concerns about cost and access to capital.
If you missed Thursday night's video, let me recap the Iron earnings, because I know many of you learned about Iron from this channel, and we learned a lot of important details on Thursday night. Iron is a relatively small company, but Iron's better-than-expected results and comments helped calm market participants' nerves, not only regarding Iron shares, but also AI infrastructure stocks in general.
I'll just quickly list the important points from both the earnings report and the conference call . Iron reported quarterly revenue of $228 million, versus the roughly $221.7 million expected. Quarterly revenue from AI cloud was $200 million. I was expecting a figure in the $160 million to $185 million range, so this is better than I expected.
Bitcoin mining revenue was just $28 million for the quarter as Iron retires mining equipment. $28 million is less than I expected. Overall, this is a positive from my perspective, as Iron exceeded my expectations for overall revenue, even though Bitcoin mining revenue was lower than I expected for the quarter.
The decrease in Bitcoin mining revenue is generally a positive factor , as we want Iron to quickly transition from mining to AI cloud, so this is a plus. AI Cloud gross margin for the quarter was 86%. The quarterly net loss was $263 million , driven by $148 million in depreciation and amortization charges, including costs for newly commissioned data centers and graphics processors, $106 million in non-cash stock-based payments, and $22 million in net finance costs.
Iron's target of over $4 billion in annual recurring revenue is fully contracted, a change from last quarter, and it is fully funded. After the end of the quarter, Iron expanded a contract with an unnamed company that deals with advanced artificial intelligence models. It's probably Anthropic, but we don't know for sure yet.
Iron also reported that contract prices continue to increase: by 11% for three-year agreements and by 12% for five-year ones. And this next point is very important. Iron signed a short-term contract with a revenue of $40 million per IT megawatt. This is a very strong indicator. It is important that this is revenue per IT megawatt, but we do not know which facility this contract applies to , so we do not know the power usage effectiveness (PUE) ratio.
But the main conclusion is that contract prices are rising. A quick note: If data center capacity does indeed become more scarce after the midterm elections, this will have a positive impact on the prices that companies like Iron can charge customers for capacity rental. Iron says its target of over $4 billion in annual recurring revenue by the end of 2026 is now being pushed back to January 2027 due to changes requested by the customer and delivery timelines.
And it could even stretch until February 2027 . During the reporting conference, management confirmed the rumors. Microsoft requested a change in the configuration of some of the network infrastructure after Iron handed over the Horizon 1 facility. Iron agreed to the requested change, which is why Iron is pushing back its $4 billion annual revenue target to January 2027, possibly February, depending on vendor timelines.
So, the rumors turned out to be true. What is important is that Iron built Horizon 1 according to specifications, and Microsoft accepted the object, meaning Iron performed the work according to the requirements. So it's not like Iron made a mistake on Horizon 1. Microsoft requested the change after they accepted the object.
Iron also mentioned signing a contract after the end of the quarter for approximately $22 million per megawatt of IT per year over a five-year contract , including an upfront payment from the customer covering 90% of the associated capital costs for the GPU. Iron also noted that it has $20.6 billion in total assets, including $6 billion in data center assets with a 20- year useful life.
And their assets and income are backed by contracts with 15 clients. Two-thirds of the order portfolio is with investment-grade counterparties. Iron also said it has $5.8 billion in cash, including $2.3 billion in restricted cash reserved primarily to fund GPU deployments. During the earnings conference call, we learned that Iron expects to recognize $250 million in AI cloud revenue in the company's next financial report.
This figure would have been higher if not for the delay associated with changes requested by the client. By June next year, Iron expects to have $25 billion in assets in the form of GPUs and data centers. Iron's CEO was asked about modular data centers and he said they are discussing it with Nvidia.
As a reminder, Jensen Huang recently announced Nvidia DGX Pods at the GTC conference in Berlin; These are essentially ready-made " AI factories" assembled at the supply chain stage and housed in a transportable container that serves as a shell for the data center. This eliminates the need to build a large facility to house such an AI factory.
As I said before, I believe Nvidia's DGX Pods are likely to be a big success and an important growth accelerator. However, it will take time for the industry to fully adopt them. Iron recently renegotiated contracts for H100 capacity at double the original price. Iron spent about $5 billion on capital expenditures during the quarter, and their capital spending forecast for the rest of the fiscal year is about $15 billion, lower than previous expectations.
When Iron was asked why expected capital expenditures were reduced, management responded that it was not due to a lack of demand, but solely due to the current state of capital markets. Iron seeks to find a balance between expanding to meet high customer demand and not wanting to raise funds at any cost.
They want to find the right balance. Market participants like it. At the same time, management has made it clear that there are signals of demand. They recently signed a short-term contract at a price of $40 million per megawatt, and Iron recently renegotiated contracts for H100 capacity at double the original price.
Management noted that shipments of Nvidia's Vera Rubin liquid-cooled systems are expected late next year, and supply of these systems next year will be extremely limited. Iron is optimistic about the economic outlook given this deficit. Sweetwater 1 and 2 are also currently under construction and are included as baseload in the distribution process, meaning that 2 gigawatts of capacity at these two sites in Texas are already secured.
Let me remind you that the new governor of Texas has stated that the construction of facilities that are already under development will continue. Sweetwater 1 and 2 are currently under construction. Overall, I would say this earnings report was much better than the previous one. Iron appears to be exercising restraint in raising capital and capital expenditures given current market conditions, and at the same time they are seeing stronger contract prices and increasing demand from customers.
As I said, quarterly revenue from AI cloud services was higher than I expected, and overall revenue was higher than I had anticipated, although Bitcoin mining revenue came in below expectations. These are all positive points. I previously stated an expected price range of $100 to $150 per share for the first half of 2027 and said I would adjust that range over time as needed.
The reason for such a wide range was Iron's large $6 billion share offering program . $6 billion is a lot for Iron. And after the last earnings report, I narrowed my expected price range to $90-$120 per share and told you I plan to narrow it even further when Iron issues additional shares. I have also repeatedly emphasized that the expected price range is not guaranteed.
It was just a general range of what I thought was reasonable. I also kept repeating that there are several assumptions built into this expected price range. These assumptions were: first, the absence of a market downturn. Second, Iron will reach its ARR goal on time. Third, the world will remain limited in computing power.
And fourth, the absence of significant capital dilution beyond the $6 billion program . I have repeatedly said that if at least one of these assumptions does not hold true, then you should not expect a price range in the specified time frame. So, two of these assumptions were not true . Of course, we have seen a recent sell-off in the tech stock market, focused largely on AI infrastructure stocks, especially in the Neo Cloud segment.
This is the first assumption that did not come true. We have seen a noticeable decline in the market. The second assumption that didn't pan out is that Iron won't hit its ARR goal by the end of the year . Again, this was because Microsoft demanded changes after Iron delivered Horizon 1 to specifications.
So technically it's not Iron's fault. However, the assumption was not confirmed. Therefore, the expected price range should not be expected in the appropriate timeframe. Considering the share dilution during the quarter and the fact that the $4 billion ARR target was pushed back to early calendar year 2027, I get a price range closer to $70–$95 per share in the first half of 2027.
But there are a few problems with this range . First, Iron reported that there was even more dilution of the stock after the end of the quarter . If I take this additional dilution into account, I get a price range of $67 to $91 per share in the first half of the 2027 calendar year. We should also consider that Iron is likely to issue more shares within its program in the coming days, and this will narrow the range even further.
Additionally, we must acknowledge that capital market conditions have changed markedly in recent weeks, and there has been a significant sell-off in AI infrastructure stocks. Simply put , two assumptions built into the initial expected price range failed to materialize: one was the market downturn, and the other was that Iron did not meet its annual ARR target on time.
Therefore, the expected price range should not be expected in the relevant time frames. I consider a range of $67 to $91 per share in the first half of the 2027 calendar year to be reasonable . But this range does not take into account the additional blurring that will occur in the future . I expect to narrow this range even further in the future as Iron issues more shares under its ATM program.
I said earlier that if Iron meets its ARR target by the end of the year, the earnings report published in May 2027 could be the first to really attract the market's attention in a positive way. With the ARR target pushed back to early 2027, I now believe that the August 2027 report is likely to be the first to truly capture positive attention from market participants.
I currently continue to be bullish on Iron and expect to likely remain so until at least the first half of the 2028 calendar year, depending on what happens. Postponing the ARR target is just a matter of time, so I'm not worried about it. Hopefully, in the future, Iron will be able to rely more on asset-backed financing, larger prepayments from customers, and revenue growth.
So yes, I'm still optimistic overall, despite having lowered my expected price range for the first half of 2027. As I mentioned earlier, I believe the sell-off in some AI infrastructure stocks is excessive and fails to take into account important context regarding the development of AI. However , I don't know where the bottom is for stocks and how long it will last.
If you need help with your situation, consider reaching out to a certified financial professional who can help with your unique circumstances. As I often say , remember to be calm in this market. Remember to take a long-term view and not make any hasty or irrational decisions. Looking ahead, we have Nvidia's earnings report on November 17th.
If you're new to the channel, I want to make sure you have at least a basic understanding of the main long-term thesis. So let's look at this now. I don't know what will happen in the short term , but in the long term I'm very confident that Nvidia will be worth much more in the coming years than it is today.
When Jensen was on Lex Friedman's podcast not long ago, he was very seriously considering the possibility of Nvidia becoming a $3 trillion revenue company in the near future. If this happens in the coming years, it is quite possible that Nvidia's market capitalization could one day reach tens of trillions of dollars .
It may sound crazy, but that's exactly what Jensen means when he raises the question of the possibility of turning Nvidia into a $3 trillion revenue company . I think the question at that point is what multiplier the market will be willing to give Nvidia. I don't know the answer to this question, but I truly believe that Nvidia will be worth significantly more in the years to come than it is today, purely due to the fundamental growth of the business.
From everything I see, the world is still limited in computing power, and I expect it to continue until at least the end of the 2028 calendar year. In a world of computing resource shortages, developers will use whatever available power they can get their hands on. There is no GPU today that is idle due to lack of demand, as was the case with fiber optics, which was unused due to lack of demand at the peak of the dot-com bubble.
Back then, companies were laying fiber optics in the hope that use cases and demand would emerge over time. Today we see the complete opposite. As I have said many times, when market participants compare this AI revolution to the dot-com bubble, they ignore the fact that this time the internet already exists.
This means that mass adoption of the technology and the development of new use cases on a large scale are possible immediately. We don't have to wait years for this to appear. It's already here. The world is limited in computing power, which means there is not enough supply to meet demand. New capacities are used immediately after commissioning.
Hyperscalers monetize capacity as soon as it becomes available . Each of the hyperscalers mentioned supply constraints during their recent earnings reports. In addition , many cloud providers are expanding to meet already contracted demand. They don't build blindly in the hope that demand will eventually appear.
No, they are expanding because they have signed contracts and, in some cases, significant prepayments from their clients. This AI revolution is fundamentally different from the dot-com bubble. And 2026 will be a pivotal year for the AI industry due to the rapid adoption of agent-based AI and the proliferation of agent-based systems in leading global enterprises.
The revenues of leading AI labs are now growing rapidly. Agent coding and the implementation of agent systems in large enterprises are new use cases that significantly increase the demand for inference, which , in turn, increases the demand for computing power. The rapid adoption of agent-based AI is why we are seeing a tipping point in demand for inference.
That is why we are seeing rapid growth in the revenues of leading AI labs. I wish Anthropic and OpenAI were public companies so the public could see their profits grow. I think the growth in revenues of leading laboratories may be the first evidence that will attract the attention of market participants and make them realize that AI infrastructure will provide a clear return on investment.
I believe that the growth in revenue of industry leaders will also reassure investors about Nvidia's long-term growth, since the revenues of these labs directly depend on computing power. If they had more computing power, they would have higher revenues. It's really that simple. The problem is not demand.
The problem is that there is not enough supply to meet this demand. As I said before, I expect the world to experience a shortage of computing power at least until the end of the 2028 calendar year, and possibly longer. So regardless of what happens in the short term, it is important for long-term investors to focus on fundamentals, remain forward-looking, and remember that we are only in the early stages of large-scale implementation of agent systems.
This will significantly increase the demand for computing. And after that, the next surge in demand for computing will likely be fueled by physical AI. We are no longer talking about digital agents performing digital tasks. With physical AI, we are talking about physical AI agents that perform physical tasks in the real world.
Nvidia's CFO called physical AI, and I quote, "a multi-trillion dollar opportunity" and the next stage of growth for Nvidia. This industry will fundamentally change society, and Nvidia has positioned itself to benefit enormously from it. Nvidia sells equipment for data centers where models are trained.
They offer Omniverse, where models are trained and tested, and Nvidia sells hardware that allows for real-time inference on devices using Nvidia AGX, allowing robots to intelligently interact with the real world even when they are not connected to a data center. Note that Nvidia is taking a holistic platform approach to physical AI, positioning itself as the fundamental foundation for it all.
Over 3 million developers are already working on Nvidia's robotics stack, and it's not getting enough attention. In terms of ramping up production, Blackwell Ultra has already been implemented and remains in high demand. Vera Rubin is already being delivered to customers. Nvidia Grock 3 LPX is now in full-scale production.
Subsequently, we expect the launch of Rubin Ultra in 2027 and Feynman after that in 2028. We have a clear product roadmap for data centers through 2028, and Jensen believes that spending on AI infrastructure will reach $3-4 trillion annually by the end of the decade. This means that Jensen expects the demand for AI to grow and the overall market for it to expand, which is what underlies it all.
I don't think we're getting close to any bubble-bursting event . With that in mind, I seriously believe that Nvidia has a lot of potential ahead of it, and I think that in the coming years this company will be worth significantly more than it is today . At least, that's my view on this situation. A quick note before I finish.
All selections on this channel are edited by Finvid with original structure and commentary. From time to time, the same montages appear elsewhere on YouTube. If you're looking for the original version, it's always here on this channel. Thank you for watching Finvid. I am grateful for your support. Remember the calm in this market.
Remember to keep a long-term perspective and don't make any hasty or irrational decisions. With all that said, I hope you have a great rest of your day, and I'm interested in hearing your thoughts on Nvidia in the comments below. Please like this video so more people can see it . And while you're there, please consider subscribing.
It's free , and you can always change your mind. Thanks for watching, and I hope to see you in the next video.
댓글 0