CoreWeave Rises 11.72% Amid Signs of Growing Demand for AI Cloud Computing

The previous trading day was September 8 (Tuesday). All three major indices declined: the Dow Jones Industrial Average closed at 52,786.07, down 1.18%; the S&P 500 closed at 7,674.13, down 0.58%; and the Nasdaq Composite closed at 26,423.69, down 0.31%. The stock of the day was CoreWeave (CRWV), which rose 11.72% to close at $99.83—while its peer Nebius (NBIS), which actually had contracts traded that day, gained 7.73%. After-hours results for AeroVironment (AVAV) will be released at 20:00 UTC tonight, marking its first year-over-year comparison since completing its merger and consolidation in May 2025. Today’s U.S. Stock Academy breaks down the rare earth magnet supply chain: China extracts 60% of the world’s magnetic rare earths, yet refines 91% and produces 94% of sintered magnets. All data in this article is based on U.S. market closing prices as of September 8, 2026, and all times are listed in UTC.

September 8 was the first trading day after the Labor Day long weekend, and none of the three major indices closed higher. The Dow Jones Industrial Average fell 1.18% to 52,786.07, the largest decline among the three; the S&P 500 dropped 0.58% to 7,674.13; and the Nasdaq Composite fell 0.31% to 26,423.69, the smallest decline.

Throughout the day, three key themes dominated: oil prices nearing $100 per barrel, the 10-year U.S. Treasury yield rising to around 4.8%, and the escalation of the U.S.-Canada tariff dispute. While these three factors are not the same event, they all point in the same direction—rising costs and higher discount rates. As a result, indices with higher weightings in large-cap stocks were hit hardest, leading the Dow Jones to lead the decline while the Nasdaq suffered the mildest drop.

Beneath the index, the divergence among industries is more telling than the index itself. Chip stocks rose against the trend: Intel closed up 9.05% at $104.47, as the market recognized that server chip supply shortages have restored its pricing power; Qualcomm rose 3.17% to $174.09 after securing a partnership with Amazon to develop custom AI chips. Meanwhile, software stocks declined on the same day: ServiceNow fell 4.99% to $134.21, while Salesforce and Intuit each dropped about 4%, as investors worried that next-generation general-purpose models could directly displace vertical software applications. On the same trading day, hardware stocks rose while application-focused stocks fell—this contrast reveals more about where capital was flowing than the index’s minor fractional movements.

Additionally, GameStop’s common stock closed down 1.41% at $18.89; its second-quarter earnings report, revealing revenue of $790.2 million and an operating profit of $160.2 million—both record highs for the period—was released after hours. After-hours figures do not affect the day’s closing price, so please treat these two events separately.

No U.S. economic data is being released tonight. The macroeconomic focus this week rests on the final two days: Thursday at 12:30 for August PPI and initial jobless claims, and Friday at 12:30 for August CPI. The Fed’s September interest rate meeting is scheduled for next Tuesday through Wednesday. This means tonight, market attention is not on macroeconomic factors, but rather on individual company earnings reports.

CoreWeave (CRWV) is an AI cloud computing company with a market capitalization of $54.5 billion. On September 8, it closed at $99.83, rising 11.72% in a single day and adding approximately $5.7 billion to its market value. Trading volume reached 50.88 million shares, 1.79 times the daily average. The stock price is currently near the 40% level of its 52-week range.

On this day, the news was particularly notable: Palantir named Nebius its preferred sovereign AI infrastructure partner, excluding CoreWeave. However, the market interpreted this partnership as a signal of demand across the entire GPU rental space—resulting in Nebius, which received the endorsement, rising 7.73%, while CoreWeave, which did not, rose 11.72%.

The five-dimensional rating clearly describes the nature of the day. The peer relative strength and peer ranking both scored 100 out of 100, meaning no other asset in the same category was stronger that day. The industry valuation temperature at 49 and trend position at 42 are both slightly below average. Volatility control scored 0—the only zero among the five dimensions.

A score of zero is not a scoring error. It corresponds to Beta 7.41—meaning when the market moves 1%, this stock moves an average of 7.4%. Dividing +11.72% by 7.41 gives approximately 1.6% on a market-adjusted basis. Viewed this way, the day’s gain is not as impressive as it appears. High-Beta stocks inherently experience amplified movements; comparing them directly to low-Beta stocks is unfair. This adjustment should always be the first step when interpreting any single-day return.

Another aspect to note is that this endorsement does not disclose the amount or committed capacity. Without specified figures, it is more akin to a qualification than an order—actual usage will only be visible in future financial reports. This is not negative news, but it determines how long this announcement can support pricing.

On the same AI computing power chain, the rewards are split two ways on the same day.

The six companies on the same AI compute chain moved together, resulting in a clean chart for the day: CoreWeave (CRWV) rose 11.72%, Nebius (NBIS) rose 7.73%, IREN rose 5.04%, Super Micro Computer (SMCI) rose 1.69%, while Micron (MU) fell 1.61% and NVIDIA (NVDA) fell 2.01%.

The four companies above are engaged in renting computing power and assembling servers, while the two below sell chips. Those renting computing power rose, while those selling chips fell—on the same day, the same blockchain was split in two. This isn’t because demand for chips has weakened, but because the announcement specifically pointed to demand for computing power rental, not new chip procurement commitments. The focus of the message determined who was revalued.

For this reason, this day cannot be described as "the AI sector collectively rose." CoreWeave's sector averaged only a 0.26% gain that day; if the entire sector had been up, it would not be accurate to say CoreWeave outperformed by 11.5 percentage points. Sector averages and intra-sector divergence represent two different levels of information—combining them can lead to reversed conclusions.

The standout was IREN, which rose 5.04% that day. Its business involves leasing out data centers previously used for mining to AI training providers, so it moved in line with AI cloud stocks rather than cryptocurrency prices—while COIN fell 3.09% and MSTR fell 4.40% on the same day. A company’s classification within a sector doesn’t always align with what drives its price movement on a given day.

A company's news being interpreted by the market as evidence of demand across the entire sector is called a read-through. CoreWeave’s rally on September 8 was a classic example of a read-through: the news originated at Nebius, but the pricing impact landed on CoreWeave.

The issue is that the reliability of peer readings varies significantly. Some readings reflect genuine new demand, while others are merely spillover sentiment. To distinguish between the two, look at three key numbers.

First, does the announcement include figures for amount and capacity? A partnership that specifies the contract amount, committed capacity, and delivery timeline can be used to estimate the incremental impact on the entire sector; a partnership that merely states “preferred partner” cannot. Today’s answer for this item is not disclosed, so the first item is left blank.

Second, has the trading volume of the second company increased? If only the price moves but volume stays flat, it suggests that only a small amount of capital is driving the move; if volume clearly expands, it indicates that new capital has genuinely entered and is re-pricing the asset. Today, CoreWeave traded 50.88 million shares, 1.79 times its daily average—reverting to an average of approximately 28.42 million shares daily—this is concrete evidence.

Third, did the price increase exceed the sector average? If the entire sector rose together, that’s a market- or industry-wide movement and has little to do with this news. But if this asset significantly outperformed the rest, then the market has likely priced in this news. Today, CoreWeave outperformed its sector by 11.46 percentage points—this point is also valid.

Two out of the three items are real, while one is left blank—this is the complete picture of the reading on September 8: funds have indeed come in, but the underlying partnership has not yet yielded quantifiable results. What needs to be verified going forward is the actual usage metrics in future financial reports.

AeroVironment (AVAV) will release its earnings after hours at 20:00 UTC, with a conference call at 20:30. Chewy (CHWY) and Signet (SIG) will also report before the market open on the same day, but AVAV is the main focus tonight.

It is the primary supplier of small military drones in the United States, and after acquiring BlueHalo in May 2025, it expanded into space, cyber, and directed energy capabilities. Positioned at the consumables end of the supply chain, it sells low-unit-cost equipment that is replenished in batches and scales with operational intensity, rather than large, decade-long platform contracts. As a result, its revenue rhythm more closely resembles that of a consumables business than a project-based engineering endeavor. This positioning means that quarterly fluctuations in its financials are naturally more pronounced than those of major platform providers.

First, look at the absolute amounts, not year-over-year comparisons. The total revenue for each quarter of fiscal year 2026 (which ends on April 30, 2026) was $455 million, $473 million, $408 million, and $642 million, respectively. These quarters did not show a steady upward trend: revenue dropped to $408 million in the third quarter, then surged to a record high of $642 million in the fourth quarter. The delivery pace varied significantly between quarters, making it easy to be misled by focusing on any single quarter.

This season ends tonight on August 1, compared to $455 million in the same period last year. Whether revenue can hold above this line matters more than how it looks versus the year-over-year figure—the reason will be explained in the next section.

The company's own full-year fiscal 2027 guidance is $2.125 billion to $2.225 billion, representing approximately +10% year-over-year. At the lower end of the range, this equates to +7.5%, and at the upper end, +12.5%.

The triple-digit growth is driven by consolidation; guidance is only set at 10%.

When you lay out the numbers side by side, the discrepancy is stark: the year-over-year revenue growth for the four quarters of fiscal year 2026 was 140%, 151%, 143%, and 133%, respectively, while the midpoint of the company’s full-year fiscal year 2027 guidance implies a growth rate of only 10%.

The sole reason for a one-order-of-magnitude difference is the consolidation base effect. The acquisition closed on May 1, 2025, and the acquired company was only included in the consolidated financial statements from that date onward. Therefore, in each quarter of fiscal year 2026, the numerator includes the new business while the denominator does not—this is precisely how the three-digit figures arose, and it has nothing to do with the pace of organic growth. In accounting, this phenomenon is called the consolidation base effect, and it occurs after any large acquisition, lasting exactly one year.

This quarter, ending August 1, marks the first year-over-year comparison on a like-for-like basis since full consolidation: both numerator and denominator now include new businesses, revealing the true organic growth rate. The company has set its full-year target around 10%.

Therefore, the figures 133% and 10% cannot be side-by-side compared as a “growth collapse.” They are calculated differently: the former is year-over-year growth excluding the acquired company, while the latter is the midpoint of the full-year guidance, including both numerator and denominator. When viewed together, the only correct interpretation is that “the base effect has ended,” not that “growth has disappeared.”

How to analyze tonight’s results? There are two key lines to watch. First, revenue must exceed the $455 million mark—the level from the same period last year and the first clean year-over-year baseline; falling below it would indicate weakening organic growth. Second, check whether there have been any updates on two metrics: Can gross margin hold at last quarter’s 32% (compared to just 21% and 22% in the prior two quarters)? And has the company retracted its prior guidance that free cash flow for this fiscal year will be negative, with growth skewed toward the second half?

On Wednesday, we break down the rare earth magnet sector. Let’s start with a set of numbers: China mines 60% of the world’s magnetic rare earths, but refines 91% and produces 94% of sintered magnets. From mining to magnet production, China’s share increases by 34 percentage points.

Putting these three numbers together reveals the conclusion: what’s scarce isn’t the ore, but the two steps required to turn it into a magnet. Rare earth elements are not rare in the Earth’s crust—they can be mined all over the world; the challenge lies in individually separating the seventeen chemically similar elements, then pressing the purified metals into sintered magnets suitable for motors. These two processes are the bottleneck.

Following this chain, five companies are positioned at five distinct points.

MP Materials (MP), with a market capitalization of approximately $9.9 billion, operates at the upstream mining stage. It is the only active rare earth mine in the United States, producing 50,700 tons of rare earth concentrate and 2,599 tons of praseodymium-neodymium oxide in 2025—doubling year-over-year from approximately 1,300 tons last year—and has produced its first neodymium-iron-boron magnets in Texas.

Energy Fuels (UUUU), with a market cap of approximately $3.7 billion, specializes in midstream separation and purification. Its core business is uranium; at its White Mesa facility in Utah, rare earth elements are co-separated during the processing of monazite, and last year it produced dysprosium oxide at 99.9% purity. It is the most direct example of the phrase “process-limited, not ore-limited.”

US Rare Earths (USAR), with a market cap of approximately $2.3 billion, operates at the downstream magnet manufacturing stage. It focuses solely on the final step: pressing rare earth metals into sintered neodymium-iron-boron magnets. Its Oklahoma production line began operations in March this year, with a year-end target capacity of 600 tons per year.

Ramaco Resources (METC), with a market cap of approximately $700 million, is positioned at the site of a new mine under development. Its core business is metallurgical coal, and its Brook Mine in Wyoming extracts rare earth elements alongside coal seams—but commercial production is not expected until 2027, meaning this segment currently generates no revenue.

General Motors (GM), with a market capitalization of approximately $77.6 billion, stands at the end of the chain. Electric vehicle drive motors require magnets, and GM is the first long-term buyer to sign on to this U.S. magnet production line.

The closer a trade is to delivery, the narrower the price range.

Calculating the ratio of 52-week high to 52-week low for these five companies yields a very orderly ranking: General Motors (GM) at 1.69x, Uranium Energy Corp (UUUU) at 2.61x, MP Materials (MP) at 2.65x, American Rare Earths (USAR) at 3.84x, and Metcalf Energy (METC) at 6.75x. The difference between the highest and lowest is a factor of 4.

First, let’s clarify what this number is not: it is not the 52-week gain, not a valuation multiple, and not Beta. It is the width of the price range over one year—the ratio of the highest price to the lowest price. The wider the range, the more frequently and by how much the market has re-priced it over the year.

Let’s explain why they’re valued this way. General Motors at 1.69x because its current sales directly represent its revenue—the market’s assessment simply extrapolates from already realized sales figures. LaMaco Resources at 6.75x because its mine won’t begin trial production until 2027—the market’s valuation is entirely based on future, unrealized events, and any news regarding timelines, costs, or policies can completely rewrite its estimates.

So this principle can be directly applied to other sectors: the closer a business is to delivering actual products, the narrower the price range; the closer it is to undeveloped capacity, the wider the range. This isn’t saying one end is better—it’s saying that when evaluating companies at either end, you need to prepare for different levels of volatility.

From here, you can identify three questions to understand a supply chain, asked in sequence. First, ask whether this stage sells raw materials or processing services—minerals are found everywhere, but production lines capable of separating seventeen elements individually are rare. Second, ask whether its revenue is being realized today—companies already delivering products see their stock prices follow orders, while those still building factories see their stock prices follow expectations. Third, ask who must buy it—the magnets ultimately end up in motors, with buyers being automakers, robotics companies, wind power providers, and defense contractors.

Linking the three questions together yields a transferable criterion: understand a blockchain by starting with “which component is hardest to replace,” rather than “which component is most upstream.”

MP Materials deserves a separate mention because it is the only company on this chain attempting to complete the entire process—from mining to magnets—making it the sole such company in the United States. In 2025, rare earth concentrate production reached 50,700 metric tons, and praseodymium-neodymium oxide output hit 2,599 metric tons, both doubling year-over-year. In the fourth quarter, the company produced its first magnets using commercial equipment in Texas.

It also features an unusual arrangement: the price floor has been signed, but production capacity is still on the way. A ten-year agreement with the U.S. Department of Defense sets a price floor of $110 per kilogram for praseodymium-neodymium oxide—effectively locking in part of the revenue floor in a highly volatile upstream commodity. However, on the other side, the new magnet plant is not scheduled to begin production until 2028, when total U.S. magnet capacity will reach approximately 10,000 metric tons.

On one side is a locked-in price floor; on the other, production capacity that won’t come online until two years from now. Together, these factors explain why its one-year price range is 2.65x—wider than General Motors, which only sells vehicles, but narrower than Lamaco Resources, which won’t begin pilot production until 2027. Its position lies squarely between the two.

Also, note that the 600-ton annual capacity target for US Rare Earths (USAR) represents only 6% of the approximately 10,000 tons of total U.S. capacity expected once new facilities are fully operational. Most of the capacity currently being discussed in this sector is still in the future.

Question: CoreWeave didn’t win the contract, so why did it rise more than Nebius?

Answer: The market interpreted this partnership as a demand signal for the entire leased hashrate sector, not just a single company’s order. This phenomenon—where news about one company is treated as evidence for an entire sector—is called peer sentiment. To assess its value, look at three metrics: whether the announcement includes amount and capacity (not disclosed today), whether the second company’s trading volume has increased (1.79x), and whether the price rise exceeded the sector average (outperformed by 11.46 percentage points).

Question: Can September 8 be described as "the AI sector rising collectively"?

Answer: No. On that day, the CoreWeave sector averaged only a 0.26% increase, while Micron (MU) and NVIDIA (NVDA)—both on the same AI computing chain—fell by 1.61% and 2.01%, respectively. On that day, the segment benefiting from renting computing power rose, while the segment selling chips declined.

Question: What does Beta 7.41 mean, and how do I use it?

Answer: Beta measures how much a stock moves, on average, for every 1% movement in the market. A Beta of 7.41 is extremely high. When comparing price changes, first adjust for Beta: divide +11.72% by 7.41, which equals approximately a 1.6% move relative to the market. High-Beta stocks inherently experience amplified price swings; comparing them directly to low-Beta stocks without this adjustment is unfair.

Question: AVA's quarter-over-quarter growth was 133%, yet the company only provided a full-year guidance of 10%—is growth collapsing?

Answer: No, the two figures are not on the same basis. The 133% represents year-over-year growth for Q4 of fiscal year 2026, when the denominator did not yet include the acquired entity merged in May 2025; the 10% is the midpoint of the full-year fiscal year 2027 guidance, which includes the acquired entity in both numerator and denominator. The difference arises from the consolidation base effect, which lasted exactly one year. This quarter marks the first year-over-year comparison on a consistent basis after one full year of consolidation.

How do you read the ratio of "52-week high divided by 52-week low"?

Answer: It is the width of the price range over a year, not an increase rate, valuation multiple, or Beta. A higher value indicates that the market has adjusted its pricing for it more frequently and by larger margins within that year. The pattern is: the closer a project is to delivering actual production, the narrower the range (e.g., General Motors at 1.69x); the closer a project is to undeveloped capacity, the wider the range (e.g., Ramaco Resources at 6.75x).