How Capital Markets Are Price-Discovering the "Supply-Demand Bottleneck" (Part 1)

US Market Audit Memo
The "Raw Materials" and "Pickaxes" of the AI Market: Capital Ripple Effects Across Upstream, Midstream, and Downstream

Dialogue Log (US Market, as of June 26, 2026, Close)
■You
Please provide an analysis of last night's US market movements.
■AI
I will organize last night's US market performance in the form of an audit memo (US Market, as of June 26, 2026, Close).
Conclusion
Last night's US stock market was characterized by 'minor index declines, but risk-off sentiment concentrated in tech and semiconductors.' On the surface, the Dow fell 0.09%, the S&P 500 fell 0.05%, and the Nasdaq fell 0.24%, all representing only minor losses. However, the underlying reality was that concerns over rising AI infrastructure costs and speculation regarding a delay in OpenAI's IPO pressured semiconductor and memory stocks, while defensive sectors like healthcare and certain mega-caps provided support, preventing a broader collapse in the indices.
Summary of Major Indices
Based on closing prices, the S&P 500 was at 7,354.02, a minor decline of 0.05% from the previous day. The Nasdaq Composite was at 25,297.62, down 0.24%, led by tech. The Dow Jones Industrial Average was at 51,876.11, a minor decline of 0.09%. These figures are consistent with CNBC's closing reports, and I judge them to be highly accurate.
On the other hand, the weekly view shows a different picture. The Dow is up less than 1%, indicating the relative strength of non-tech stocks. The S&P 500 is down about 2% for the week, showing overall weakness. And the Nasdaq is down over 4% for the week, revealing that the damage to tech is quite deep. In other words, while the movement was minor last night alone, it is important to note that the pain in the Nasdaq is quite significant on a weekly basis.
What Caused the Decline — Organizing the Causality
Breaking down the background of the decline by factor, it looks like this:
First, concerns over rising AI infrastructure costs. The market is wary of rising costs for AI data centers, memory, and storage, which has led to valuation compression for semiconductor and AI-related stocks.
Second, speculation regarding a delay in OpenAI's IPO. Following reports from The New York Times, the view that OpenAI's IPO timing would be pushed back spread, worsening sentiment for AI-themed stocks in general.
Third, the persistence of PCE inflation. May's PCE was 4.1% year-on-year, and core PCE was 3.4%, levels that keep expectations of interest rate hikes within the year alive, acting as a headwind for high-P/E stocks.
Fourth, the decline in crude oil. A drop in crude oil is usually a supportive factor for the stock market, but last night it was not enough to offset tech anxieties.
In summary, the sequence was that rising AI-related costs and worsening IPO speculation triggered a 'theme collapse,' and the persistence of PCE inflation amplified the fragility.
The Center of the Decline — Semiconductors and Memory
The center of last night's decline was clearly in the semiconductor and memory sector. Micron fell over 6% based on CNBC and 7% based on Investopedia amid volatility following its earnings, as the side effects of AI memory supply and demand were feared. Intel fell 3% following the sell-off in semiconductors in general. Arm fell nearly 4% as this theme's high valuation was reassessed. Marvell fell 5% as expectations for AI infrastructure were unwound. Furthermore, the sell-off spread to memory and storage in general, with Western Digital falling 13%, Seagate 12%, and Sandisk 10%.
What is important here is that the fact that 'AI demand is strong' itself has not been denied. Rather, the essence of last night was that demand is so strong that peripheral costs are rising, and the 'benefits' and 'side effects' of the AI boom have begun to be recognized simultaneously.
Why the Indices Did Not Completely Collapse
There are several supporting factors for why there wasn't a total sell-off. First, some mega-caps rebounded. Microsoft rose 6%, Apple 3%, and Amazon 2%, with Tesla and Meta also rising. Next, there was a shift of funds to defensive sectors. Healthcare rose over 7% for the week, real estate and utilities were up around 3.5%, and consumer staples were also firm. In addition, the decline in crude oil slightly eased macro anxieties through lower energy prices.
In other words, last night can be summarized as a day when the market's lead shifted temporarily from tech to defensive sectors, rather than a total sell-off.
Organizing Interest Rates, Inflation, and Crude Oil
The 10-year Treasury yield fell slightly from around 4.38% to just under 4.40%, but it is not at a level that would act as a tailwind for stocks. PCE inflation is 4.1% headline and 3.4% core, so expectations for interest rate hikes have not disappeared, which is a headwind for high-P/E stocks. Crude oil fell to around $72 for Brent and around $69-$70 for WTI, partially easing inflationary pressure.
Given this, it is more natural to view it as 'an environment where it is difficult to maintain high AI valuations due to the persistence of inflation' rather than 'stocks were sold because interest rates surged.'
My Analysis
From here on is my own interpretation.
First, even if the surface is a minor decline, the reality is a 're-valuation of the AI market.' The decline in the indices last night was limited, but the content was quite clear: a correction of overheating in AI infrastructure-related stocks. It is appropriate to view this not as a 'collapse of AI demand,' but as an unwinding of parts where 'demand is strong, but expectations for prices and financing were too far ahead.' It is not a move to deny the medium- to long-term AI theme, but a phase of short-term valuation correction. The basis for this lies in the fact that while semiconductor and memory stocks plummeted, some large-cap stocks like Microsoft and Apple were bought.
Second, the macro environment has not returned to 'complete peace of mind.' Figures of 4.1% PCE and 3.4% core are not levels at which the Fed can easily turn dovish. Therefore, the environment remains one where high-growth, high-P/E AI stocks are prone to price swings on even minor bad news. Last night's decline was a market driven more by the combination of interest rates and valuations than by corporate earnings themselves.
Third, it should be viewed that funds 'shifted' rather than 'fled.' Defensive sectors such as healthcare, real estate, and utilities were strong, and some mega-caps also rebounded. This is a rotation from some growth stocks to relatively safe areas rather than a capital outflow from the entire market. Therefore, it is appropriate to evaluate last night as 'a day when sector rotation became clear' rather than 'the first day of a total collapse.'
Checkpoints for Tonight and Beyond
I will list the points to watch in the future. First, whether there is an autonomous rebound in semiconductors. If Micron, Arm, Marvell, etc., do not recover, it is a signal that the AI adjustment will be prolonged. Next, the trend of long-term interest rates. If the 10-year Treasury settles around 4.4%, the downside room for tech will shrink. Regarding crude oil, the focus is on whether it will rise again due to the Strait of Hormuz or Middle East tensions; if it rises again, it will be a re-alert for inflation. Finally, the continuity of defensive sectors. If healthcare and utilities continue to rise, it can be judged that the shift of funds is in full swing.
In a Nutshell
Last night's US market was a day where 'indices were slightly down, but the content was a continued adjustment of AI and semiconductors. Funds fled to defensive sectors.' In the short term, the sluggishness of the Nasdaq's recovery is the most important factor, and unless that improves, it is better not to view it as 'no wind because the indices are slightly down.'
Next, I can delve into either 'spillover to Japanese stocks,' '5 stocks to watch tonight,' or 'a detailed version centered on NVIDIA, Micron, and Apple.'
■You
Please select 5 stocks to watch in the US market tonight and summarize their materials, points of interest, and risks.
■AI
I will organize this as '5 stocks to check first at the next session' based on tonight. To conclude, it is most efficient to view the current market through three pillars: ① AI semiconductor supply-demand and cost issues, ② AI investment financing anxiety, and ③ return to defensive sectors.
Showing the overall picture first, in order of priority, the most important are Micron and NVIDIA, followed by Apple, Oracle, and Moderna. Micron is a stock that 'had explosive earnings but fell sharply the next day,' reflecting the tug-of-war between the authenticity of AI demand and the correction of overheating. NVIDIA is at the center of the AI market, and the point to watch is that its recent recovery is sluggish. For Apple, the focus is on whether it can pass on the high cost of memory. Oracle is a proxy indicator reflecting concerns about debt expansion for AI investment. Moderna is a symbol of the rotation into healthcare. I will look at them individually below.
(1) Micron
In terms of materials, Micron exceeded market expectations in its recent earnings with revenue of $41.46 billion and adjusted EPS of $25.11, and further indicated revenue guidance of $50 billion for the current quarter. While the strength of AI memory demand was confirmed, the stock price plummeted again the next day due to concerns about AI infrastructure costs, making it a typical example of 'earnings are strong but the stock does not rise.'
The point of interest is whether the earnings evaluation will become dominant again. I will watch to see if the 16 long-term contracts and $22 billion in commitments will be re-evaluated as a source of peace of mind, and if Micron will react first if the overall semiconductor environment recovers.
The risk is that because the stock price has risen significantly in the last year, the 'high expectations' risk of being sold even on good earnings is large. Also, the longer the AI memory shortage lasts, the more the increased cost burden on the customer side could become a headwind for the entire market.
In a word, it is the most important stock to confirm tonight that 'AI demand is not dead yet.'
(2) NVIDIA
In terms of materials, amid the recent tech sell-off, while some mega-caps rebounded, NVIDIA, along with Alphabet, was on the side with a weak recovery. The market is looking at how much the rise in AI infrastructure costs will be tolerated, rather than NVIDIA's own earnings.
The point of interest is that if NVIDIA can rebound, it will be a signal of improved sentiment for the entire AI theme, while if it remains weak, the recovery of the entire Nasdaq is likely to be limited. It is also necessary to confirm whether it is linked to peripheral stocks such as Micron, Arm, and Marvell.
The risk is that because it is at the center of the AI theme, it is prone to the triple whammy of interest rates, valuations, and supply-demand adjustments. And it is sensitive to news like speculation about a delay in OpenAI's IPO, which 'lowers the heat of AI expectations.'
In a word, it is the 'thermometer of the AI market' that should be watched before the indices.
(3) Apple
In terms of materials, Apple officially announced price increases for the MacBook and iPad. The background is the surge in memory and storage prices due to the expansion of AI data centers. It fell over 6% the previous day, but rebounded 3% the next day, and the market is playing tug-of-war over whether to view this price increase as 'bad news' or a 'gross margin defense measure.'
The point of interest is whether the price increase will be evaluated as leading to the maintenance of profit margins, and whether the shift to AI-compatible terminals such as 12GB RAM will justify the rise in ASP (Average Selling Price). If Apple holds up, the view that 'large consumer-facing companies can absorb even high AI costs' will spread.
The risk is that if sales volume falls due to the price increase, it will turn into a concern about sales deceleration rather than gross margin defense. Also, if the trend of passing on AI costs continues, it will be a headwind not only for Apple but for hardware in general.
In a word, it is a stock to watch for the 'costs of the AI boom' rather than the 'benefits of the AI boom.'
(4) Oracle
In terms of materials, Oracle has fallen 19% in the last week, its largest decline since 2001. The focus is not on earnings themselves, but on the massive debt and capital increases to fund AI infrastructure investment. Debt at the end of the period is about $130 billion, and a financing plan for an additional $40 billion is indicated for fiscal year 2027.
The point of interest is that if the stock price stops falling, it will be a sign that anxiety over financing for AI infrastructure investment has subsided, while if it continues to fall, the view that 'demand is strong but profitability and finances cannot hold' will strengthen. The key is which will win: expectations for large-scale demand related to OpenAI or concerns about financial burdens.
The risk is increased debt, dilution, and prolonged low-margin investment, and it may be treated as a 'stock with AI investment financing anxiety' rather than an 'AI winner.'
In a word, it is a representative stock to watch for the 'credit risk aspect' of AI capital investment.
(5) Moderna
In terms of materials, Moderna's stock price rose significantly in the second half of June as the FDA accepted the review of the company's new influenza vaccine. In addition, in the recent market, healthcare has become a destination for funds fleeing from tech, and Moderna is a symbolic stock for that.
The point of interest is whether the rotation into healthcare will continue in a phase where tech adjustments continue, whether the progress of the FDA review will connect to earnings expectations without ending as a short-term material, and whether it can maintain relative strength even if the entire market is weak.
The risk is that biotech stocks are news-driven, and the rebound after digesting the material is fast. If sector rotation stops, the healthcare advantage is also likely to run its course.
In a word, it is a stock to measure strength as an 'alternative to tech.'
Order of Observation
I will list 3 points to check first. First, look at the opening and initial movement of Micron and NVIDIA to judge whether the AI theme will continue or be re-adjusted. Next, look at Apple's reaction to confirm whether it can pass on high AI costs. Then, look at the relative strength of Oracle and Moderna to measure whether AI financing anxiety or the return to defensive sectors is dominant.
My Conclusion
The 5 stocks tonight are not just individual materials, but a combination that represents the three themes of the market. Micron and NVIDIA confirm the strength of AI demand, Apple looks at the success or failure of passing on AI costs, Oracle looks at the financial sustainability of AI investment, and Moderna looks at the continuity of the flight to safety. If you look at these 5, you can judge with quite high accuracy whether tonight's US market is a 'resumption of the AI market' or a 'deepening of selection in the AI market.'
If necessary, I can next organize the buy/sell scenarios for these 5 stocks in 3 patterns: 'bullish, bearish, and neutral.'
■You
In other words, the demand for pickaxes is high. This part is solid. However, the raw materials for pickaxes have become too expensive. Therefore, the companies selling the raw materials for pickaxes are taking capital away from the companies making the pickaxes. This also gives benefits to the companies making the pickaxes, although it takes time, right?
■AI
Yes, that understanding is quite correct. However, to be precise, it is closer to say that 'the total amount of benefits increases, but the timing of the distribution of benefits is shifted.' The current structure is that the raw material side takes it first, and there is a possibility that it will circulate to the product side later.
Organizing the Structure
It becomes easier to understand if you divide this AI market into three layers. The upstream is the 'pickaxe raw material sellers,' i.e., the supply side of memory, storage, and semiconductors, which is the layer that can immediately monetize price increases. The midstream is the 'pickaxe manufacturing,' i.e., terminal manufacturers and server/AI infrastructure builders, which first bear the cost increase and then pass on the price later. The downstream is the 'gold diggers,' i.e., AI service operators and app companies, which get the biggest fruit at the end if they can monetize.
In the current market, it is a fact that the upstream, like Micron, is taking profits the fastest. In fact, Micron's revenue is more than four times higher than the previous year, and its gross margin has also improved significantly, so it is receiving the fruits of supply-demand tightness first. On the other hand, Apple has raised prices for MacBooks and iPads in response to the surge in memory, which is a move where the 'midstream is late to go and take it back.'
Response to Your Point
First, regarding the point that 'raw material companies are taking capital away from manufacturing companies.' In the short term, that is correct. When supply and demand are tight, raw material prices rise first, and the manufacturing side is exposed to an increase in procurement costs. What the market is wary of right now is exactly this: the expansion of AI infrastructure is causing memory and storage prices to soar, and that cost could squeeze the profit margins of Apple and AI-related companies in a broad sense.
Next, regarding the point that 'it still gives benefits to manufacturing companies.' This is also Yes, with conditions. This is because high raw material prices usually prompt an increase in final product prices, and if demand does not collapse, the manufacturing side can also pass on the price. Apple's price increase is exactly that; even if it is disliked in the short term, it could lead to gross margin defense in the medium term. Also, if supply is fixed, like Micron's long-term contracts, it becomes easier for the customer side to make business plans based on 'high but predictable costs.'
Under What Conditions Will 'Benefits Come Later'?
There are several conditions for benefits to circulate to midstream manufacturing companies. First, being able to pass on prices. If this can be done, profit margins can be protected. In the current market, Apple falls into this category. Second, demand does not fall. It is a situation where it sells even if prices are raised, and AI demand is still strong for now. Third, having long-term contracts. There is a prospect of costs, and Micron is expanding contracts here. Fourth, supply enhancement progresses. This eases future raw material price increases, but it takes time to realize. Fifth, being able to withstand financing. On the point of whether capital investment can be continued, Oracle is a weak part here.
In short, whether 'the midstream can take back the profits sucked up by the upstream first' is determined by price-passing power and capital strength. Apple is relatively strong. On the other hand, in cases like Oracle, where massive investment is run on debt, financial anxiety is more likely to appear in the stock price before the benefits come.
Looking at the Time Axis
Winners shift according to the time axis. In the short term, raw material companies that can immediately raise prices due to supply-demand tightness win. In the medium term, strong manufacturing companies that can recover investment through price increases or high value-added products win. In the long term, final monetization companies that have created high-profit businesses using AI win.
Therefore, your statement that 'this part is solid' can be said to be quite solid regarding demand itself. Micron's figures are symbolic, but AI-related demand is strong as actual demand. However, the stock market is now looking quite strictly at the next point of discussion: 'who will take the fruits of that demand?'
Conclusion
If I organize your hypothesis into one sentence, it becomes this: 'The demand for AI-oriented pickaxes is real and solid. However, at the present time, the upstream is taking excess profits first due to raw material constraints, and the benefits to the midstream and downstream will ripple later through price passing and supply enhancement.'
In other words, winning demand and winning the stock price are not simultaneous right now. The order is: first the upstream wins, then strong manufacturing companies, and finally platform companies that can monetize.
If necessary, I can next organize this structure by breaking it down into groups of stocks and the order of buying in the upstream, midstream, and downstream in the form of a matrix.
■You
In the current market, please divide US stocks into 'companies selling raw materials' and 'companies making pickaxes,' list 3 stocks to watch for each, and summarize the buy materials and risks.
■AI
I will organize this in the form of an audit memo as 6 US stocks divided into 'raw materials' and 'pickaxes' for the AI market.
Definition of Premises
Here, for convenience, I define 'companies selling raw materials' as companies that supply upstream materials and capacity for AI infrastructure, such as memory and storage. I define 'companies making pickaxes' as companies that use those materials to sell servers, cloud capacity, and AI terminals. This is not an official accounting classification, but a practical classification for reading the current market.
Conclusion First
The strong arguments in the current market for companies selling raw materials are supply-demand tightness, pricing power, and long-term contracts, and representative stocks are Micron, Sandisk, and Seagate. The strong arguments for companies making pickaxes are price-passing power, capturing AI demand, and resistance to capital burdens, and representative stocks are Dell, Oracle, and Apple. The center of gravity of the market at the present time is still dominated by the upstream (raw material side). However, the market's view right now is that 'strong pickaxe manufacturers' with price-passing power can take it back later.
Companies Selling Raw Materials — 3 Stocks to Watch
Micron is the main player in AI memory. Buy materials include the fact that revenue for the recent quarter was $41.46 billion, exceeding the market expectation of $35.84 billion, and the revenue outlook for the next quarter is $50 billion, also exceeding expectations. Furthermore, there are 16 long-term contracts and $22 billion in commitments, and the pricing power and visibility of demand as an upstream player are quite strong. The risk is that despite this, the stock price has plummeted recently due to concerns about rising AI infrastructure costs. In other words, it is in a state where 'earnings are strong, but expectations are stronger,' making it the group of stocks most prone to volatility in the short term. Monitoring points are the maintenance of guidance, the ratio of long-term contracts, and whether it can become the leader of the recovery for semiconductors as a whole.
Sandisk handles flash storage supply for AI. Buy materials include, based on Reuters, that storage demand for AI data centers is strong, and because the amount of data storage required by AI systems exceeds supply, it is in a structure where it is easy to raise prices. Data center segment revenue was more than triple at $1.47 billion, and the quarterly revenue outlook was $7.75-$8.25 billion, exceeding the market expectation of $6.49 billion. The risk is that the rise since the beginning of the year is large, and even in the Reuters article, it was said that 'there was not enough wow factor to maintain momentum,' and it is falling after hours even on good earnings. In other words, actual demand is strong, but expectations for the stock price are ahead. Monitoring points are the continued growth of data center revenue, maintenance of prices, and whether the disappointment sell-off after earnings will subside.
Seagate handles large-capacity storage in the AI era. Buy materials include the company's claim that 'HDD is necessary for large-capacity storage even in the AI era,' and it is aiming for 100TB HDDs by 2030. In addition, management has stated that while showing supply visibility to customers for 4-5 quarters ahead, 'demand is much stronger than that visibility,' which shows bargaining power as an upstream supplier. The risk is that the CEO himself has admitted that it takes time to expand factories and set up new equipment, so supply constraints could be prolonged. Also, competition with SSDs and the prolongation of supply shortages are sparks for stock price volatility. Monitoring points are the capacity roadmap, the continuation of data center demand, and the progress of supply expansion.
Companies Making Pickaxes — 3 Stocks to Watch
Dell handles the assembly and sale of AI servers. Buy materials include the fact that AI server revenue recently increased 757% year-on-year to $16.1 billion, total company revenue increased 88%, and adjusted EPS also significantly exceeded expectations, causing the stock price to rise 32.76% in one day after earnings. This is a representative example that 'even with high raw material costs, strong assembly and sales companies can absorb it with revenue.' The risk is that because the rise is too sudden, expectations are quite high, and if AI server demand slows down even a little, valuation adjustments are likely to appear, as is typical for hardware stocks. Monitoring points are the continuation of AI server orders, server gross margins, and linkage with NVIDIA supply.
Oracle handles the supply of AI data center and cloud capacity. Buy materials include the fact that AI demand itself is very strong, and according to CNBC, 71% of analysts recommend buying, so the demand signal itself is still considered strong. In other words, 'pickaxe demand' exists. The risk is that the stock price has fallen 19% in the last week. The reasons are $130 billion in debt, CAPEX of about $56 billion in fiscal year 2026, a free cash flow deficit of about $24 billion, and further large-scale financing plans. Demand is solid, but the capital burden being too heavy is the biggest risk. Monitoring points are financing conditions, the pace of CAPEX, and the speed at which large-scale demand such as for OpenAI converts to earnings.
Apple is a representative of price passing in AI-compatible terminals. Buy materials include the fact that it officially announced price increases for the MacBook and iPad, showing that it is one of the few companies that can pass on high material costs to customers. Furthermore, the shift to AI-compatible terminals with higher memory configurations is a tailwind for ASP increases in the long term. The risk is that the stock price fell over 6% immediately after the price increase announcement. The market was more wary of 'demand deceleration' than 'gross margin defense.' In other words, even if price passing is possible, if volume falls, profit improvement will be limited. Monitoring points are sales trends after price increases, replacement demand for AI terminals, and the presence or absence of further price revisions.
Practical View — Which Should Be Bought First?
The side that is advantageous changes depending on the phase of the market. In the early stages of supply-demand tightness, 'companies selling raw materials,' where prices rise first and profit reflection is fast, are advantageous. In the medium term, when price passing begins, 'strong pickaxe companies' that can recover gross margins if they have brand power or order-taking power are advantageous. In phases where capital investment fatigue appears, companies expanding with debt are at a disadvantage, and companies with light finances are advantageous. The current market is still leaning toward 'upstream advantage.' However, companies that can return with 'revenue growth or price passing,' like Dell and Apple, can become rotation destinations from the upstream-only focus. On the other hand, Oracle has strong demand, but the heaviness of its finances is the biggest point of discussion.
My Final Organization
The 3 that are currently the most 'solid' among companies selling raw materials are Micron, Sandisk, and Seagate. The 3 that I 'want to verify' the most among companies making pickaxes are Dell, Oracle, and Apple. In a word, if you want to go and get it, it's the upstream Micron, Sandisk, and Seagate; if you want to see the possibility of being rewarded later, it's the midstream Dell and Apple; and Oracle is where demand exists but capital efficiency is the biggest point of contention.
If necessary, I can next organize these 6 stocks into 3 categories: 'suitable for short-term trading, suitable for medium-term holding, and pass.'
■You
Some AI say, 'This is not just cyclical rotation, but there is a possibility that the market has begun to seriously doubt the ROI of AI infrastructure investment for the first time.' Personally, I cannot agree with this. In my private opinion, AI demand is solid, so pickaxe manufacturing companies jumped for the time being. However, AI demand is larger than imagined. Therefore, as actual demand, the raw materials for pickaxes themselves are in short supply. And capital first puts capital into pickaxe raw material companies to resolve that supply-demand gap—the essence is that they themselves want to make money. That will, as a result, resolve the supply-demand gap.
■AI
Yes, as far as I can see from the public information available at this time, your organization is more consistent with the current price movements. At the very least, I see it as one step too early to generalize the recent decline as 'the market has begun to seriously doubt the ROI of AI infrastructure investment.' Rather, what is happening now is not ROI denial, but a re-evaluation of supply constraints and profit distribution.
Conclusion
I will show my judgment for each point of discussion. Regarding 'Is AI demand solid?', Yes. This is because Micron's revenue, guidance, and long-term contracts are too strong. Regarding 'Is the recent rotation cyclical?', partly cyclical, but the content is a re-evaluation of supply and demand. Even within tech, profit distribution is changing between upstream and midstream. Regarding 'Is ROI concern the center of the market?', it exists in part but is not the center. It is still a problem of 'raw material prices, capital burden, and time lag,' not 'demand collapse.' And regarding 'Is your hypothesis valid?', quite valid. It is consistent with the flow of capital gathering in the upstream first to fill supply constraints.
(1) Why is it still difficult to see 'full-scale ROI skepticism'?
If ROI skepticism were in full swing, a chain of 'worsening demand outlook' → 'order slowdown' → 'lower guidance' should be widely visible. However, what can be confirmed recently is the opposite: Micron has shown revenue of $41.46 billion, a revenue outlook of $50 billion for the next quarter, 16 long-term contracts, and $22 billion in commitments. This shows that 'AI demand is not weaker than expected,' but 'demand is so strong that supply and prices cannot keep up.' In Sandisk and Seagate, the strength of storage demand for AI data centers, supply shortages, price increases, and visibility of the future have been repeatedly shown. It is more natural to read this as a bottleneck of raw material/capacity shortages rather than ROI denial.
(2) Translating your hypothesis into market structure
It can be organized as a time lag in profit distribution. In the leading phase, the entire server/finished product/AI-related sector rises due to expectations for AI demand, and pickaxe manufacturing companies win. In the phase where constraints become apparent, pricing power shifts to the upstream due to shortages of memory, storage, and capacity, and raw material companies win. In the adjustment phase, capital for production increases heads to the upstream, and expectations for resolving supply and demand are formed. This is a phase where the capital market is the protagonist. And in the next phase, strong manufacturing companies in the midstream take it back through price passing and normalization of supply.
In this sense, your explanation that 'pickaxes jumped first, but demand was too large, exposing raw material shortages, and capital is flowing to the raw material side' makes a lot of sense. In fact, Apple has begun to pass on memory price surges to consumer prices. This is a move where the midstream is not just swallowing high raw material prices and ending, but going to take it back with a time lag.
(3) What is the market really doubting, if not 'ROI'?
I see that what the market is doubting at the moment is, more accurately, the following three points. First, ROI itself—the doubt that AI investment might not be profitable. However, this has not yet become comprehensive. Second, the timing of recovery—the doubt that even if it is profitable, it might take a long time to recover. This is quite strong. Third, capital efficiency—the doubt that debt or dilution might be too heavy before recovery. Oracle is typical.
What is visible, especially in Oracle, is not demand anxiety but anxiety over capital burden. Based on CNBC, the stock price has fallen 19% in the last week, and debt, CAPEX, and financing plans are the biggest points of discussion. This is not a story of 'there is no demand for AI,' but 'the cost to meet demand is heavy.' In other words, it is closer to say that the market is not doubting 'gold will not come out even if you dig,' but 'will the financing not become difficult before you dig?'
(4) Why your view is strong
Checking for consistency with facts, materials supporting your hypothesis line up. Micron's explosive earnings support the strength of demand. The increase in long-term contracts is evidence that it is actual demand, not a temporary theme. Apple's price increase is evidence that high raw material prices are rippling to the midstream. Seagate's supply constraint comments show that bottlenecks actually exist. Sandisk's high outlook shows that storage shortages are pushing up prices.
Therefore, it is more consistent with public information to read the current price movement as 'AI investment is working too well, and the scarcity of upstream materials is soaring' rather than reading it as 'AI investment might be a waste.'
(5) However, your hypothesis also has conditions
I need to add one supplement here. It does not necessarily mean that if capital enters the upstream, the supply-demand gap will be resolved cleanly. Several conditions are intertwined. Regarding the lead time for production increases, it will go well if supply catches up in a few quarters, but it will fail if the enhancement is too slow. Regarding midstream price passing, it succeeds if gross margins can be maintained in the Apple style, and fails if it cannot be passed on and squeezes profits. Regarding capital market support, it succeeds if capital increases and corporate bonds are absorbed, and fails if the cost of capital soars in the Oracle style. Regarding the continuation of final demand, it succeeds if AI investment continues, and fails if investment is shelved due to macro deterioration.
In other words, your hypothesis is quite strong as an 'explanation on the demand side,' but in the stock market, you must also look at 'the speed of supply expansion' and 'the financial resistance of the midstream.'
Final Evaluation
Your explanation is quite close to the essence at the present time. To be more precise, the current market is in a phase where supply constraints of upstream materials and reorganization of capital allocation are beginning because the strength of AI demand is greater than expected, rather than having begun to fully deny the ROI of AI infrastructure investment itself.
And as a result, your view that first the upstream makes money, then capital flows to the upstream, which fills the supply-demand gap, and after that strong companies in the midstream take back profits—is quite persuasive.
If necessary, I can next organize the order in which funds ripple from the upstream to the midstream in the form of 10 US stocks based on your hypothesis.
■You
Based on your hypothesis, please create a comparison table of 10 US stocks arranged in the order in which funds ripple from upstream → midstream → downstream.
■AI
The premise is your hypothesis that AI demand is solid, but now it is not ROI collapse, but capital is moving ahead to resolve upstream bottlenecks such as memory and storage. Based on that premise, I will list 10 US stocks in the order in which funds are likely to enter. This time, I am intentionally organizing it by focusing on the 'bottleneck axis (memory/storage → server/cloud → AI monetization).'
Conclusion First
The layer where funds are most likely to enter is the 1st to 4th, which are the direct beneficiaries of raw material shortages. The layer where it is next likely to ripple is the 5th to 7th, which is the layer that uses raw materials to make AI equipment/capacity. The layer that is likely to be evaluated last is the 8th to 10th, which turns AI into actual revenue, ARR, and advertising revenue.
10 Stocks — Order of Ripple
1st place is Micron (upstream). It is the main player in memory shortages and the first beneficiary of supply-demand tightness, so funds enter first. Buy materials include revenue of $41.46 billion, a revenue outlook of $50 billion for the next quarter, 16 long-term contracts, and $22 billion in commitments. The main risks are volatility due to expectations being ahead and high-price caution. Indicators to monitor are guidance, long-term contract ratio, and gross margin. The certainty is high.
2nd place is Sandisk (upstream). It is a direct beneficiary of flash shortages for AI data centers. Buy materials include data center revenue being more than triple, the revenue outlook for the 4th quarter being $7.75-$8.25 billion, exceeding expectations, and having pricing power due to supply shortages. The main risks are disappointment sell-offs after a surge and overheating of expectations. Monitoring indicators are data center revenue, ASP, and comments regarding inventory/supply. The certainty is high.
3rd place is Seagate (upstream). It handles large-capacity storage demand in the AI era, and because it takes time to supply, it is a stock with high scarcity. Buy materials include supply visibility to customers for 4-5 quarters, the 100TB HDD roadmap, and data center storage demand. The main risks are the length of the lead time for production increases, competition with SSDs, and the prolongation of supply constraints. Monitoring indicators are capacity roadmap, shipment outlook, and price difference between HDD and SSD. The certainty is high.
4th place is Western Digital (upstream). It is an upstream stock for storage demand re-evaluation, linked to the same theme as Seagate and Sandisk. Buy materials include stock price rise due to the resurgence of AI storage selection, and management's explanation that 'data volume increases with AI inference.' The main risks are disappointment sell-offs if the decisive factor is not as strong as Micron or Sandisk. Monitoring indicators are storage demand, nearline/HDD orders, and price revisions. The certainty is medium to high.
5th place is Super Micro Computer (midstream). It is a layer that buys raw materials and assembles 'pickaxes,' and the strength of demand first appears as a shortage of working capital. Buy materials include acquiring $39 billion in AI server orders in a few weeks and showing revenue growth of over 100%. The main risks are dilution due to $7 billion in financing, high material costs, and China-related regulations. Monitoring indicators are order backlog, financing conditions, and memory procurement costs. The certainty is high.
6th place is Dell (midstream). It is a representative that turns materials into actual server revenue, and it is a layer where funds are likely to head next after the upstream. Buy materials include AI server revenue increasing 757% to $16.1 billion, total company revenue increasing 88%, and a significant upside in EPS. The main risks are the reaction after the valuation rise and the business cycle of hardware. Monitoring indicators are AI server orders, gross margin, and the growth of ISG (Infrastructure Solutions Group). The certainty is high.
7th place is Oracle (midstream). It is a layer that makes 'capacity itself,' and although demand is strong, the capital burden is heavy, so the stock market's evaluation is late. Buy materials include the AI demand signal being strong and the analyst buy recommendation ratio being high. The main risks are about $130 billion in debt, a surge in CAPEX, FCF deficit, and the risk of additional financing. Monitoring indicators are CAPEX, financing, and the speed of cloud order revenue conversion. The certainty is high.
8th place is Microsoft (downstream). From here on is the 'layer that turns AI into revenue,' and among the downstream, monetization is most visible. Buy materials include annual AI revenue of $37 billion, Azure growing 40%, and 365 Copilot having over 20 million seats. The main risks are $190 billion in CAPEX, gross margin squeeze, and competition after OpenAI's non-exclusivity. Monitoring indicators are Copilot seats, Azure growth, and annual rate of AI revenue. The certainty is high.
9th place is ServiceNow (downstream). It is a stock that works when enterprise AI enters the phase of full-scale introduction, and funds are likely to enter as it moves from verification to actual operation. Buy materials include raising the annual subscription outlook, strong AI demand, expansion of Now Assist introduction, and upward revision of the AI revenue target to $1.5 billion. The main risks are AI substitution concerns for software as a whole, M&A burden, and geopolitical factors. Monitoring indicators are ARR, Now Assist multi-product introduction rate, and number of large-scale projects. The certainty is high.
10th place is Meta (downstream). It is the 'final monetization' layer that comes last, and it is in the stage of spreading AI to advertising improvement, subscriptions, and future cloud. Buy materials include trial introduction of AI subscriptions, exploration of new revenue sources other than advertising, and medium- to long-term growth room through AI investment. The main risks are that about 98% of revenue is currently advertising, AI monetization is still in the experimental stage, and the heaviness of CAPEX. Monitoring indicators are AI subscription sign-ups, improvement in advertising unit prices, and ratio of new revenue. The certainty is medium.
How to Read
The most 'solid' right now is the 1st to 4th layer. This is the layer where funds enter for supply shortages themselves, and especially Micron and Sandisk have demand strength directly linked to revenue, guidance, and pricing power.
5th to 7th are the 'demand exists, but capital burden increases' layer. Super Micro and Dell buy upstream materials and turn them into actual server revenue, but there are cases like Oracle where the correctness of demand and the correctness of the stock price do not match. The reason is that even if demand is strong, debt, dilution, and CAPEX are heavy.
8th to 10th are the layer that looks at 'can they really take it all at the end?' Microsoft already has an annual rate of $37 billion in AI revenue, which is quite advanced among the downstream. On the other hand, ServiceNow is the next best if enterprise-oriented full-scale introduction grows. Meta is positioned as 'AI revenue other than advertising' still being in the verification stage.
3 Classifications for Practical Use
The leading fund inflow zone is Micron (MU), Sandisk (SNDK), Seagate (STX), and Western Digital (WDC), which are the main players in resolving bottlenecks. The ripple zone is Super Micro (SMCI), Dell (DELL), and Oracle (ORCL), which is the layer that turns upstream materials into equipment, capacity, and finished products. The recovery confirmation zone is Microsoft (MSFT), ServiceNow (NOW), and Meta (META), which looks at whether AI investment really turns into revenue and profit.
Final Word
If you follow your hypothesis, the current market is still using the 1st to 6th as the main battlefield, 7th is the watershed, and 8th to 10th are the 'layer that goes to take it back later.' What is especially important is that the further downstream you go, the more 'speed of monetization' is questioned rather than 'strength of demand'.