Last night's CPI data was unremarkable, but the market is already looking for the next sector to support the AI narrative

Last night's CPI data was unremarkable------3.4%, in line with market expectations, and did not cause much of a stir at the macro level.

However, last night’s U.S. stock market was not lacking in highlights. The most attention-grabbing focus was on where the next trading sector that can carry the AI narrative will be after the storage chip narrative has temporarily cooled down. Apart from the optical communication sector we discussed earlier, last night several companies' earnings reports were telling the market: Neocloud (next-generation AI cloud) is becoming a strong candidate.

1. Doubling Revenue and Tripling Order Book

In last night's U.S. stock market, two focus companies in the Neocloud sector, CoreWeave (CRWV) and NEBIUS (NBIS), delivered impressive earnings reports, with the two companies ultimately rising 19% and 34%, respectively.

First, let’s look at CoreWeave. The quarterly revenue was $2.58 billion, doubling compared to the same period last year. Even more astonishing is the backlog of orders: it reached $104.2 billion, expanding two and a half times within twelve months; and the third quarter has just begun, with new contracts adding another $25 billion.

Nebius's growth rate is even more significant: the quarterly revenue was $582 million, a year-on-year increase of 454%; its AI cloud business has reached an annualized scale of $3 billion; on the operational side, EBITDA has achieved a positive number for the first time.

The earnings reports of the two companies point to the same fact: the existing computing power capacity has been completely booked by customers, and even so, it is still in short supply.

2. The Wall of This Business is Not Chips, But Electricity

It is worth noting that these new cloud vendors are shifting their barrier-building strategy from "stockpiling chips" to "stockpiling electricity."

The logic is easy to understand: when GPUs are in short supply, placing orders at a premium can always solve the problem; however, substations, transmission capacity, and grid connection permits cannot be hastily acquired with money. The chip shortage is a commercial issue, while the electricity shortage is a physical issue.

Nebius's recent transaction prices are the best illustration: short-cycle, fast-delivery computing power is priced at $40 million to $50 million per megawatt; while the long-term contract price is only $20 million to $25 million per megawatt------the pricing for short-cycle orders is about twice that of long-term contracts.

Even more interesting is the company's sales strategy. For the capacity in 2027, customers are willing to buy the entire order now, but management intentionally holds back a portion from signing. It’s not that they can’t sell it, but as electricity becomes scarcer, the capacity reserved for later becomes more valuable------the pricing power must be in their own hands.

3. Burry's Short Formula is Being Disproven Item by Item by Reality

These two earnings reports also conveniently dismantled a short seller's platform.

According to public reports, short-selling institution Michael Burry previously questioned: some manufacturers depreciate GPUs over five to six years, while the actual usable lifespan of GPUs may only be two to three years, thus the profits in the sector may be overestimated.

However, the details disclosed by the two companies during the conference call almost disproved this premise item by item.

First, the assets have not depreciated. Nebius publicly auctioned computing power capacity for the first time, and the hammer price was 15% higher than the company's historical highest bid. Something that, according to short sellers, "only has two to three years of residual value," is being bid up by buyers.

Second, the lifespan far exceeds assumptions. CoreWeave confirmed it signed a new A100 computing power contract, with an ideal price, extending the term all the way to 2029. This is a chip that was launched in 2020 and is already three generations behind------according to the short seller model, it should have been scrapped long ago, yet in reality, it has secured a full nine-year commercial contract. The company also revealed that the pricing for older products is not inferior to, and even stronger than, several years ago, with the old fleet basically booked out.

One analyst described the current computing power shortage: customers are no longer picky; unable to get the top-spec new cards, they are taking all the old models.

4. The Key Short Sellers Missed: Renewal Revenue Equals Pure Profit

If the first two points only overturned the "lifespan assumption," then the third point directly pierced through the short seller's valuation model.

These old clusters that have been renewed have already paid off all asset-side loans during the initial contract period, and depreciation has also been fully accounted for. This means that every revenue generated from renewals only faces costs of electricity and minimal operational expenses------there's almost nothing else to consume it.

A contract signed for the A100 until 2029 reveals two cards at once: the commercial lifespan of old chips is much longer than what short sellers think; and these assets, which are counted as "zero" in the short seller model, are continuously generating cash close to pure profit. Burry was wrong not just about the depreciation period but also missed calculating the second segment of revenue, which is the most lucrative part of the entire business model.

5. Final Thoughts

Putting together the recent market trends, the power transition within the AI sector has become quite clear: the expectations for storage have peaked, optical communication is taking over with performance slope, and now Neocloud has presented "sold-out capacity, electricity pricing power, and the collapse of short seller logic" as triple validation.

For investors, the upcoming stock selection criteria can actually be condensed into three sentences: Do you have electricity? Is the order visibility high? Can old assets continue to generate money? Those that meet these three criteria are the hard roles in the new narrative; those that can only tell stories and burn capital as pseudo-computing power targets will be accelerated in elimination during this round of "accounting-style" market.

In the second half of AI trading, being loud is useless; you need to have real assets in hand.

Risk Warning: This article is contributed by a third-party author and represents the author's personal views, not the stance or views of BIT. The market data, company earnings information, and analytical judgments mentioned in the article are sourced from public channels, and BIT does not guarantee their accuracy, completeness, or timeliness, nor does it bear any responsibility for any losses arising from the use of this article's content. The content of this article is for reference only and does not constitute any investment advice, offer, or suggestion to buy or sell any financial products. The market has risks, and investment requires caution; past performance does not represent future returns. Investors should independently assess their own risk tolerance and be responsible for their investment decisions.