Pre-Market Briefing on Aug 26: Super Micro Computer Gains 9.35% Against the Trend, While Industry Average Drops 1.71%

On August 25th, the US stock market closed with the three major indexes rising synchronously, with the Nasdaq leading the way by 0.66% and closing at 26,151 points. The most memorable sector that day was not the Constituent Stock index, but SMCI (Supermicro Computer) - it rose 9.35% in a single day and closed at $38.46, while the average value of the Nasdaq computer hardware industry on the same day was -1.71%. A sector of stocks went in the opposite direction to its own industry.Looking upstream, the five companies in the AI power supply chain have seen their highest price divided by the lowest price in the past 52 weeks, rising from 1.53 times to 5.29 times. However, there are five completely different temperaments for the same theme. Looking ahead, at 12:30 UTC tonight, there will be PCE price data for July, and at 20:20 UTC, NVIDIA will release its financial report. The expected daily fluctuation given by option pricing is ± 5.3%, which is the lowest on this week's financial report list, but converted into an amount, the market value of about 270 billion US dollars is still fluctuating overnight.The data in this article is based on the closing of the US stock market on August 25th.

 

On August 25th, the three major US stock indexes closed in the same direction. The Nasdaq Composite Index rose 0.66% to close at 26,151 points, the S & P 500 Index rose 0.30% to close at 7,676 points, and the Dow Jones Industrial Average rose 0.30% to close at 53,577 points. The driving force in the closing session came from the news of the easing of the geopolitical situation, and the yield of government bonds fell synchronously.

 

Three indexes rose simultaneously, which is not common in the middle of a financial report-intensive week. The more common pattern is that funds switch positions between growth and defense, causing the index to diverge. The lack of differentiation on this day indicates that the driving force comes from a factor that applies to all sectors - a decline in yield - rather than an independent story of a single sector.

 

Apart from the index, there are three things worth noting separately.

 

The first is NVDA's own performance. It closed up 2.19% to $213.05 the day before its earnings report. Its 52-week high was $236.54, which means it is about 10% away from its highest price in a year before the most important earnings report is released. On the same day, there was also news that an employee of the company was charged with illegally exporting AI chips in Taiwan. This news did not have a discernible impact on the stock price that day.

 

The second is the counterexample of INTU. The company's stock fell 3.37% to $369.92 on the day, and fell about 9% after the financial report was released after hours. Its fourth-quarter revenue and earnings per share both exceeded market expectations - according to the simplest understanding, this should be a good financial report. However, the stock price's reaction was downward, due to the revenue guidance for fiscal year 2027: the company gave $233 to $23.50 billion, which was lower than the market's expectation of about $23.70 billion.

 

The third one is Zoom (ZM) on the same night. Its performance also exceeded expectations, and it also fell after hours, down about 3.5%, because the growth rate of the core online business continued to normalize.

 

Put these three things together and you get a clue that keeps popping up in today's report: During earnings week, the market usually pays a much lower price for what has happened than for what will happen next. Exceeding expectations is an assessment of the past quarter, and guidance is a commitment to the next quarter; when the two directions are not aligned, the market almost always chooses to believe the latter. This clue will appear again in the NVIDIA foresight in the sixth section of this article.

 

 

The star of the day on August 25th was SMCI, an AI server manufacturer with a market value of $24.90 billion. It closed at $38.46, up 9.35% in a single day.

 

First, let's look at the position of this gear in five dimensions.

 

Among the five dimensions, the two corners on the right are full: the relative strength of the same industry is 100 points, and the ranking of the same industry is 100 points. The trend position is 48 points, just stopping in the middle of its own one-year range. The weakest dimension is the industry valuation temperature, which is only 36 points, and the volatility control is 41 points.

 

This shape says a rather specific sentence: It is the strongest in the industry, but its own price has only returned to half of the one-year range, and the valuation of the industry it belongs to is not cheap. Strength and position are two things - a stock can be "the strongest in the industry" and "far from its high point" at the same time, and these two judgments are not contradictory to each other.

 

 

Closing at $38.46, 9.35% in a single day, corresponding to an increase of about 2.10 billion dollars in market value in one day. This number is consistent with the market value of 24.90 billion and the increase of 9.35%.

 

It is worth noting the trading volume. 47.78 million shares were traded on the day, which is about 0.9 times the 20-day average- lower than the recent average level . That is to say, this is not a breakthrough in volume, but a rise in volume: the price was pushed up by 9% without obvious incremental capital intervention. This usually means that the seller's willingness was relatively low on the day, rather than the buyer's strength being particularly strong. The meaning of these two explanations for the subsequent trend is not the same, and it is worth continuing to observe in the following trading days.

 

 

From a fundamental perspective, two sets of numbers drove the day.

 

One is orders. Supermicro received more than $60 billion in new orders in the fourth quarter, and the backlog of orders reached a historic high. The second is guidance. The company's revenue guidance for the 2027 fiscal year is $650 to $72 billion. In addition, AI solutions account for about 60% of the revenue in the fourth quarter, and the company expects this proportion to rise to over 80% as the backlog of orders is delivered. The full-year revenue for the 2026 fiscal year is $39.10 billion, a year-on-year increase of 78%.

 

Timing is also one of the reasons. August 25th is the trading day before NVIDIA's financial report. It is understandable for funds to concentrate on AI server hardware on this day.

 

 

On that day, the average value of the Nasdaq computer hardware industry was -1.71%, while 9 out of 11 companies on the same industry list of Supermicro rose, with Supermicro itself ranking first. The fact that the industry average was negative and most of the companies on the list rose indicates that the industry category also includes a large number of other companies that fell that day.

 

The list is not consistent internally. Western Digital (WDC) and Seagate Technology (STX) - both hard drives - rose 3.53% and 3.40% respectively; while SanDisk (SNDK), which also belongs to storage but makes flash memory, fell 0.83%. On the same AI server chain, hard drives and flash memory went in two different directions that day.

 

This detail is important because it rules out the easiest explanation. If "some big news related to AI drives the whole chain" that day, then hard drives and flash drives should not go their separate ways. They are separated, indicating that the driving force is hierarchical and structural, rather than a unified news event.

 

In terms of supporting roles, Robinhood (HOOD) closed up 8.17% on the same day. The driving factor is that Bitcoin rose to $77,000, retail trading volume increased, and brokerage fees directly consumed this amount.

 

 

Supermicro released two big numbers at the same time on this day: new orders in a single quarter exceeded $60 billion, and the full-year revenue guidance for fiscal year 2027 was $650 to $72 billion. Putting these two numbers together, it is easy to trigger an intuitive comparison - is the order received in a quarter almost equivalent to the goods to be sold in a whole year?

 

 

Backlog refers to orders that have been accepted but not yet delivered. Revenue is only counted after the goods are actually shipped and recognized according to accounting standards. The two measures two completely different time points in the industry chain: the former is "how much received", and the latter is "how much is expected to be handed over".

 

So when you see a large order number, the correct next step is not to compare it with revenue, but to ask three questions: How long is the delivery cycle? Which quarters will it spread over? Is the production capacity, supply chain, and customer payment rhythm stable during this period? Only when these three questions are answered can the order number be translated into expected revenue.

 

In practice, there are three most common forms of mixed use: directly treating the order amount as the revenue for the next year; dividing the order amount by the revenue to obtain a "coverage multiple" and inferring growth based on this; or using the year-on-year growth rate of the order amount as a leading indicator of revenue growth without making any delivery assumptions. All three methods will misread a number about "demand" as a number about "revenue".

 

For Supermicro, the company itself has made it clear: AI solutions accounted for about 60% of the fourth quarter revenue, and it is expected to rise to over 80% with the backlog of order delivery-- This sentence itself describes an unfinished delivery process The actual shipment rhythm in the following quarters will be gradually revealed in the financial report, which is where the orders will be turned into revenue.

 

This week, the US stock primary school is dismantling the AI electricity industry chain. The entry point is simple: on the same chain, the highest price in the past year divided by the lowest price, from 1.53 times to 5.29 times.

 

First, remove the chain.

 

The market value of the five companies differs by more than 30 times, but they sell things in a row along the same current: first there is a man-made unit (GE Vernova), then someone generates electricity (Constellation), then someone delivers electricity to the entrance of the data center (Eaton), and finally someone divides the electricity and takes away the heat in the computer room (Vertiv). Oklo stands at the forefront of this chain, making small modularization reactors, and the first power plant has not yet been built.

 

There is a key difference hidden in this ranking: The further downstream, the more customers are concentrated in the data center; the further upstream, the more dispersed customers are. Eaton sells power equipment to data centers, as well as factories, airports, ships, and cars; Vertiv's data center power supply and thermal management, the customer is almost the data center itself. The difference in customer concentration will eventually be reflected in the amplitude of the stock price.

 

List the 52-week interval multiples of these five companies, the order is quite clear.

 

 

Oklo's is 5.29 times, meaning it is more than five times as expensive as it is at its cheapest during the year. Eaton's is 1.53 times, meaning it is only 50% off throughout the year. The widest one swings about three and a half times as much as the narrowest one.

 

 

The calculation method is only one step: the highest price in the past 52 weeks divided by the lowest price in the past 52 weeks. Three times, which means that the most expensive time of the year is three times the cheapest time.

 

It is not a rise or fall . The rise or fall has a direction, but there is no interval multiple. It only answers with one sentence: How much has the market's view of this company swung this year? A stock that has fallen throughout the year can have a very high interval multiple.

 

It is not volatility either . Volatility measures the daily jitter, interval multiples the distance between the two extremes in a year, closer to the "narrative amplitude" - the difference between the two prices given by the market to the same company at its most optimistic and pessimistic times.

 

 

First, it measures the amplitude, not the direction. Eaton's 1.53 times increase does not mean that it has risen less, but rather that there are other businesses following it when it falls - about a quarter of its sales in the previous quarter came from aviation and vehicles, which is not in the AI power supply chain at all. When the AI narrative cools down, this quarter of revenue will not follow suit, so its lower edge is supported.

 

Second, the larger the multiple, the more it relies on expected pricing. Oklo had a net loss of $48.5 million on revenue of $1.20 million last quarter. Under this financial structure, the range of 5.29 times is almost entirely determined by the variable "when can monetization". Its valuation is essentially a discounted string of future cash flows, so when interest rates move, it moves first.

 

Third, and the one that can be most easily transferred to other tracks: When you see a popular track, first rank the interval multiples of each company, and then ask - If this story is not told next year, how much revenue will it have left? The company with a larger multiple and less remaining revenue, what you get is the elasticity of the story, not the growth of the industry. These two things can be worth holding at the same time, but they require completely different positioning management methods.

 

Therefore, "AI electricity" has never been a transaction. It is a bet with different properties shared by five parties using the same wire.

 

 

Tonight at 20:20 UTC, NVIDIA (NVDA) will release its financial report after the close of the US stock market. On the same day, there will also be Saifushi (CRM), CrowdStrike (CRWD), and Synopsys (SNPS) after the close of the market.

 

NVIDIA's uniqueness lies in its location. It is not just a company that sells chips, but the upstream gate of global AI computing power - the capital expenditures of cloud vendors, model companies, and sovereign AI projects, almost all of which have to be replaced by its acceleration cards and entire cabinets. Therefore, its quarterly data center revenue is actually the common denominator of the entire AI industry chain.

 

 

Four quarters, from $41.10 billion all the way to $75.20 billion. The quarter-on-quarter growth rate increased from 5% to 21% - not linear growth, it is accelerating .

 

To read this line, we must only look at the data center and not the total revenue. The reason is that the data center already accounts for about 90% of its total revenue, while the remaining parts such as games and cars are small in size and have a stable pace. Putting them into the total revenue will only dilute the signal. Only by looking at the data center alone can we read the true slope of AI capital expenditure.

 

The total revenue guidance given by the company in the previous quarter was $91 billion, with a fluctuation of 2%. Based on the proportion of 90% of the segments, Data Center needs to stand at about $83 billion to meet the standard. It should be noted that 83 billion this number is a reference level derived from the total revenue guidance, not the segment guidance given by the company When reading, it should be separated from the four quarters that have been announced.

 

 

First, look at the guidance, not this season. The digital market for this season has been priced repeatedly, and what really determines the direction after hours is the guidance for the next quarter given at the same time - its starting point is more important than whether this quarter exceeds expectations. This has already appeared once in the first section of this article: Caijie's revenue and earnings per share exceeded expectations, and the stock price still fell about 9% after hours, because the guidance for the next year was lower than expected.

 

Second, look at the gross profit margin and growth rate. The non-GAAP gross profit margin in the previous quarter was 75.0%. If the gross profit margin declines this quarter and the quarter-on-quarter growth rate of Data Center drops back to single digits, it indicates that the new platform's climb is eating up profits and growth at the same time. Looking at either of them alone is not enough to make the following judgment, it only makes sense to look at both together.

 

 

In addition to the financial report, three economic data will be released at 12:30 UTC tonight: the July PCE price index, the second quarter GDP revision, and the July durable goods orders. This week's macro focus will be on Friday's Jackson Hole annual meeting, where the new Fed president will give his first public speech.

 

Comparing NVIDIA to the entire week's financial report list, you will see a counterintuitive ranking.

 

 

The largest one tonight had an expected amplitude of ± 5.3%, which is the lowest in the whole game. OKTA's ± 12.6%, RBRK's ± 12.2%, and AFRM's ± 10.8% are all more than twice as much.

 

The reason is not difficult to understand: the $5 trillion plate is already difficult to push. But In terms of amount, this "minimum" is still about $270 billion in market value swinging overnight - larger than the total market value of several other companies on the list combined. Percentages and amounts are two completely different intuitions given here.

 

Regarding this indicator, there is one sentence that must be made clear: Implied volatility is inferred from the option price, only talking about amplitude, not direction. Numbers represent deep market disagreement on the results, not bullish or bearish. Any use of ± 5.3% as a rise or fall prediction is a misuse of this indicator.

 

The rest of the week's rhythm is: after-hours on Thursday, there are MRVL, WDAY, Autodesk (ADSK), AFRM, RBRK, and ULTA; before-hours, there are Best Buy (BBY), DG, DLTR, and Bilibili (BILI), as well as initial jobless claims data; there are no key financial reports on Friday, but there are August Chicago PMI and final consumer confidence values at the University of Michigan.

 

 

Not contradictory. The range of companies covered by the industry average is much larger than the 11-tier peer list used by Today's Star. The 9-tier list closed up, with Supermicro ranking first, while the average of the entire Nasdaq computer hardware industry is negative, indicating that there are many other companies in this category that fell that day. The two numbers represent different samples and can be established simultaneously.

 

No, and these two numbers should not be divided or subtracted together. Accepting orders refers to orders that have been accepted but not yet delivered, and revenue only counts after the goods are actually shipped and recognized. The two have different calibers and measure different points in the industry chain. The correct reading is to look at them separately: accepting orders reflects the intensity of demand, while revenue guidance reflects the company's own judgment of delivery rhythm.

 

Supermicro traded 47.78 million shares on the same day, about 0.9 times the 20-day average volume, lower than the recent average level. The price rose by 9%, but the volume of participation did not increase synchronously, which usually means that the seller's willingness to place orders on the day was low, rather than the buyer's strength being particularly strong. The meaning of volume increase and volume decrease in the subsequent trend is different, so it is worth noting separately instead of just looking at the increase.

 

The three quantities are completely different. The rise and fall have a direction, answering "whether it has risen or fallen, how much it has risen or fallen"; the volatility measures the daily fluctuation range; the interval multiple (52-week highest price ÷ 52-week lowest price) measures the distance between the two extremes in a year, without direction, closer to "how much the market's view of this company has swung". A stock that has fallen throughout the year can also have a high interval multiple.

 

Because this track accounts for a different proportion of each company's revenue. About a quarter of Eaton's sales in the previous quarter came from aviation and vehicles, which had nothing to do with this chain. Therefore, when the narrative cooled down, it had other businesses to follow, with a range of only 1.53 times. Oklo's revenue in the previous quarter was $1.20 million and a net loss of $48.5 million, with almost no ready-made revenue. The price was entirely determined by the expectation of "when can it be monetized", so the range was pulled up to 5.29 times. The number of times is the amplitude, and the revenue structure explains the reason.

 

No. Implied volatility is inferred from the option price, only talking about amplitude, not direction . It expresses the depth of the market's disagreement on the results: the larger the number, the greater the difference in views between the bulls and bears on the price drop point after the financial report. It is neither a prediction of rising nor falling, nor a target that can be directly operated.

 

Because the data center already accounts for about 90% of its total revenue, the rest of the games, cars, and other parts are small in size and have a stable pace. Putting them into the total revenue will only dilute the real signal. Only by looking at the data center line can we read the slope change of AI capital expenditure - from 41.10 billion to $75.20 billion in the past four quarters, the month-on-month growth rate has increased from 5% to 21%. This acceleration itself is key information.

 

Disclaimer: This article is compiled and written by the MEXC RealStocks team . The data in this article is based on the closing of the US stock market on August 25, 2026, and the financial report and forward-looking data are as of pre-market on August 26, 2026. The content is a compilation of public market information, and the individual stocks are the subject of public discussion, which does not represent the recommendation or opinion of MEXC and does not constitute any investment advice. More US stock content: @MEXC | @Alpha_MEXC | @MEXC_Research