Meta「Selling Computing Power」 Drives AI Hardware and Software Stocks Up Against the Trend

TL;DR

· Bloomberg reported that Meta is planning its cloud infrastructure business, intending to sell AI computing power and model access.

· In trading on July 1, META surged nearly 9%, while AI cloud targets like CoreWeave and Nebius came under pressure, and the hardware chain also faced strain.

· Market funds are shifting from a pure AI infrastructure expansion bet to a focus on capital expenditure return, software monetization, and cash flow certainty.

· Related Tickers: META, NVDA, AMD, MU, MRVL, TSM, CoreWeave, Nebius, PLTR, CRM, NOW, MSFT.

Bloomberg reported on July 1 that Meta is planning a cloud infrastructure business, with an internal project named Meta Compute, involving the sale of AI computing power and model access. Meta has not commented on the report, so this is still a reported plan and not an officially announced new business.

The most direct market reaction to this news appeared in the divergence between different AI sectors. On July 1, Meta's stock price surged nearly 9%, while AI cloud companies like CoreWeave and Nebius faced pressure, and some software stocks strengthened. Related hardware targets also experienced pressure, but public attributions were not consistent. A more prudent interpretation is that the market is starting to reshuffle the AI chain.

Over the past two years, the smoothest logic in AI trading has been that big companies continuously buy GPUs, build data centers, and companies selling chips, memory, network equipment, and computing power leasing have continued to benefit. Meta's new signal adds another layer of constraint to this logic. Big companies are still buying computing power, but they may also be willing to sell surplus computing power at times. Therefore, the market is not only looking at who is selling shovels but also starting to see who can fully utilize and lease out the shovels and turn the investment into returns.

Meta Surges, Trading a Capital Expenditure Recovery Story

Ordinary investors can understand this as: Meta has spent a huge amount in the past to buy "shovels," and now it may rent out some shovels to make money. Here, capital expenditure includes investments in chips, servers, data centers, and power.

The surge in Meta's stock price does not mean the market believes it will soon challenge Amazon Web Services, Microsoft Azure, or Google Cloud. The more direct reason is that the massive AI investment now has an understandable recovery story. As long as idle resources can turn into external revenue, investors' concerns about excessive capital expenditure will decrease.

According to media and market data, Meta's previously stated 2026 capital expenditure expectation is in the range of $125 billion to $145 billion, significantly higher than in 2025. Meta has not stopped expanding its AI infrastructure. The issue is that such a large investment, if only supported by a long-term AI product narrative, will eventually prompt the market to inquire about how the profit statement will catch up.

The sale of compute power has filled this gap. It does not require immediate monetization of Meta's AI applications, nor does it require Meta to win the cloud computing war instantly. It first addresses a more practical question: can the already built infrastructure increase utilization and generate revenue.

This is also the core of the divergence in trend between META and some AI infrastructure targets on that day. For Meta, externalizing compute power clarifies the path to capital expenditure return; for external compute power suppliers and hardware chains, this means that a new variable has emerged in the previously unidirectional demand narrative.

AI Cloud Targets Under Pressure, Scarcity Premium Being Repriced

DA Davidson's Gil Luria likened this pattern to SpaceX selling excess launch capacity. The essence is not a change in core business, but turning the already invested infrastructure into an external source of income. He also mentioned that this could affect the growth stories of companies like CoreWeave and Nebius in the AI cloud sector.

This explains why an increase in Meta's stock price and a decrease in AI cloud companies' stock prices can coexist. For Meta, cloudification may be asset optimization. For specialized AI cloud providers, a financially stronger, more hardware-acquiring large customer could turn into a new competitor.

Companies like CoreWeave and Nebius were previously priced by the market based on the scarcity of AI compute power. Customers needed GPU clusters, traditional cloud providers could not fully meet the demand, so professional AI cloud providers commanded a premium. If more major companies start commercializing their in-house compute power, the supply boundary will expand, and the scarcity valuation will be compressed.

This does not mean that AI cloud demand will disappear. Model training and inference will still consume a significant amount of resources, and customers may not be willing to rely entirely on potential platform competitors like Meta. But the market will first tackle a question: if more large-scale sellers emerge on the supply side, how much pricing and growth premium can professional compute power lessors maintain.

Therefore, the pressure on CoreWeave and Nebius does not simply come from "AI no longer needing compute power," but from a change in the competitive structure. In the past, the market bought into the scarcity of compute power supply, but now the market is starting to consider: if major companies' self-built compute power becomes external supply at certain stages, the valuation logic of professional AI cloud providers will need to be discounted again.

Hardware Stocks Pullback, Reflecting Procurement Slope Scrutiny

The pressure on hardware chain is more prone to misinterpretation. The market is not implying the sudden disappearance of demand for companies like Nvidia, TSMC, Micron, but rather reassessing the demand growth slope. Over the past two years, the strongest valuation anchor for AI hardware stocks has been the large factories' continued acceleration of procurement. As long as the models get bigger and the inferences increase, it seems like there is never enough computing power.

Meta's reported plan to sell excess computing power disrupts the one-sided nature of this narrative. The large factories are not only buyers of computing power but may also become sellers. Capital expenditure now needs to answer not only "how much was purchased" but also utilization rate, depreciation, customer demand, and pricing.

The impact on hardware and supply chain targets such as NVDA, AMD, MU, MRVL, TSM is more reflected at the valuation level. Nvidia and AMD represent GPU and accelerated computing, TSMC corresponds to advanced process manufacturing, Micron involves storage demand, Marvell is related to data center networking and custom chips. They all have benefited from the expected expansion of AI infrastructure in the past, and as the market begins to discuss the external sale of excess computing power, funds will first question the speed of future order upgrades.

Current evidence is more supportive of "asset utilization improvement" rather than "significant slowdown in new purchases". Meta is still in a high capital expenditure environment, and there is no collective signal from other tech giants to indicate a clear reduction in AI infrastructure plans. This pullback in hardware stocks this time seems more like a loosening of the valuation anchor, rather than a reversal of the order cycle.

This loosening is particularly sensitive to storage, network chip, and manufacturing segments. Their stock prices often reflect expectations of order upgrades in the next few quarters in advance. Once investors suspect that large customers will optimize their procurement pace, improve asset utilization, prices will move first, even if actual orders have not been revised downward yet.

In other words, the trading of hardware stocks is not about the disappearance of current demand, but about marginal expectation changes. As long as the market shifts from "computing power is always insufficient" to "computing power also needs to yield returns", the valuation of the hardware chain will transition from a narrative of supply shortage to a test of order visibility, customer procurement pace, and price elasticity.

Software Stocks Strengthen, Funds' Preference Shifts to Cash Flow Certainty

While some AI cloud and hardware targets were under pressure, some software stocks strengthened on the same trading day. This differentiation indicates that the market has not entirely withdrawn from the AI theme but is reselecting directions within the AI chain with clearer risk-reward profiles.

Software and platform targets such as PLTR, CRM, NOW, MSFT have a different valuation anchor compared to hardware chain. Hardware companies rely more on capital expenditure cycles, order upgrades, and production capacity, while software companies are more easily evaluated by the market based on application deployment, customer budgets, subscription revenue, and profit margin framework. When investors start to question the AI infrastructure's return, software assets that can clearly explain the path to monetization are relatively more likely to attract fund attention.

This does not mean that software stocks are not under valuation pressure, nor does it mean that all AI software companies can realize their growth. More precisely, in the trading triggered by this Meta news, there has been a shift in market preference: from "who provides computing power" to "who can monetize AI into revenue and efficiency".

Microsoft possesses attributes of both cloud, software, and AI platforms, making its position more complex in this round of reassessment. It is both an AI infrastructure investment side and a software and cloud service monetization side. In contrast, pure hardware chains and specialized AI cloud vendors are more directly exposed to procurement cycles and computing power supply changes.

The strength of software stocks essentially reflects the market's increased demand for AI investment returns. In the past, funds could be used to first buy infrastructure expansion; now, funds are more willing to see how AI enters enterprise workflows, how it forms chargeable products, and how it improves the bottom line.

The Divergence Lies in "Old Card Rental" or "Procurement Slowdown"

The reason why this transaction is important is that it accommodates two explanations at the same time. The more positive explanation is that Meta is entering a phase of AI infrastructure financial discipline. The huge capital expenditure not only serves internal models but also needs to be turned into sellable cloud products. This is favorable for Meta's investment return narrative.

Another explanation comes from some investors and industry observers: Meta may simply be taking out old cards or temporarily idle computing power to sell, which does not mean it is not short on computing power. It can sell some resources while continuing to purchase new-generation hardware and expand data centers. The two things are not in conflict.

This distinction is crucial for investment decisions. Selling existing assets mainly impacts Meta's income statement and capital expenditure recovery narrative. Reducing new purchases is what will truly impact revenue expectations along the chip, memory, server, and data center chain.

Therefore, the more reasonable conclusion for now is that the market has preemptively punished the narrative of "permanent computing power shortage," but there is not enough evidence to prove that AI orders are entering a downturn cycle. For the hardware chain, this is a round of expectation adjustment, not a confirmation of demand collapse.

For the software chain, this round of differentiation actually reinforces another theme: AI investment cannot only stay in the construction phase, eventually it must enter the product, customer, and revenue stage. Those who can prove this point will be more likely to receive a relative premium in the market rotation.

Capital Expenditure and Order Realization Determine the Depth of Reassessment

What can change the assessment in the future is not what Meta's cloud business is called or whether it is packaged into a product similar to AWS, but two harder variables: whether Meta will reduce subsequent capital expenditure or GPU procurement pace, and whether other major players will follow suit in selling their own AI computing power.

If Meta sells part of its computing power while maintaining high capital expenditure, this is more like increasing asset utilization. It enhances Meta's investment return story but may not necessarily hurt long-term hardware demand. A hardware shakeout is more likely a phase-wise expectation correction.

If more major players commercialize "excess computing power" and simultaneously slow down new purchases, AI infrastructure will enter another phase. The industry is not ceasing development but transitioning from grabbing resources to accounting for them. At that point, the valuation anchor for shovel sellers will shift from supply shortages to order visibility, pricing, and utilization.

This is also key to observing the hardware chain such as NVDA, AMD, MU, MRVL, TSM, and AI cloud targets like CoreWeave, Nebius. The market will continue to inquire: Are large customer purchases still on the rise, is computing power rental pricing stable, and is new supply compressing the premium of professional cloud providers.

Meanwhile, the performance of software and platform targets such as PLTR, CRM, NOW, MSFT will also serve as a reference for AI trading style changes. If funds continue to prefer the software side, it indicates a growing market emphasis on "AI application monetization." If hardware rebounds, it means investors are reconfirming the sustainability of AI infrastructure demand.

Meta has not yet proven that AI capital expenditure has peaked, but it has prompted the market to start trading a new constraint. AI infrastructure not only needs to be purchased quickly but also used fully, sold efficiently, and eventually reflected in returns.

Welcome to join the official BlockBeats community:

Telegram Subscription Group: https://t.me/theblockbeats

Telegram Discussion Group: https://t.me/BlockBeats_App

Official Twitter Account: https://twitter.com/BlockBeatsAsia