GoBull Research: Meta Begins Selling Compute Power, AI Craze Enters Profitability Phase

GoBull Research's assessment on this matter is straightforward:

Meta selling compute power doesn't signal the end of the AI boom. It indicates the AI trade is entering a new phase of accountability.

Over the past two years, the market's focus when buying AI infrastructure has largely been on one thing: who can secure GPUs.
Now, the question has shifted: once GPUs are acquired, can they generate profits?

Meta's reported move to sell excess AI compute has excited the market because it offers investors a new narrative:

Meta's AI datacenter investments may not be just costs, but could become rentable assets.

This is positive for Meta ($META).
For pure-play compute cloud companies like CoreWeave ($CRWV) and Nebius ($NBIS), it's less comfortable.

GoBull Research's US stock takeaways:

• Meta ($META): Slightly bullish. The market finally sees a path to recoup AI capex.
• CoreWeave ($CRWV), Nebius ($NBIS): Slightly bearish. Previously valued for "compute scarcity," their largest customers and hyperscalers may now sell compute directly.
• Nvidia ($NVDA): Not immediately bearish. As long as GPUs remain highly utilized, the order logic holds. But the market will increasingly ask: are customers making money with these cards?
• Amazon ($AMZN), Microsoft ($MSFT), Alphabet ($GOOGL): Neutral to slightly positive impact. More competition, but AI cloud demand is validated.
• Power, cooling, datacenter engineering chains: The thesis remains intact, but blind buying is over. Future focus will be on orders, margins, and true utilization rates.

In a nutshell:

The AI US stock theme persists, but capital will shift from "who has compute" to "who can turn compute into cash flow."

What Happened

On 2026-07-01, multiple media outlets citing Bloomberg reported that Meta is preparing to launch a cloud computing/AI compute business, potentially to sell excess AI processing power.

This business could take two forms:

1. Directly renting out raw AI compute;
2. Providing model or API access on Meta's infrastructure.

Following the report, Meta's stock surged. Business Insider noted Meta's stock rose nearly 10% intraday; WSJ reported an 8.8% gain for the day. Meanwhile, MarketWatch stated CoreWeave and Nebius fell about 13.9% and 17%, respectively.

This reaction is telling.

The market isn't suddenly convinced Meta's models are superior. The excitement stems from Meta potentially finding a way to turn its massive AI capex into a business.

Here’s the key distinction: Meta hasn't officially announced a full-fledged business yet. What's clear is that Zuckerberg previously stated if the company ends up with excess compute, selling or renting it out is an option to consider. This report has brought that possibility to the forefront.

Why This Matters

Meta's previous issue was: the ad business is great, but AI spending is too intense.

An internet advertising company suddenly starts spending like a utility, cloud provider, or datacenter operator — investors will naturally get nervous. Especially when Meta's large language models aren't universally recognized as top-tier, the problem becomes more acute:

What exactly are you getting for all that money?

Selling compute provides an answer.

Not a perfect one, but enough for the market to rally.

It reframes Meta's AI infrastructure from "internal R&D cost" to "commercializable asset." If the company's own models and products can't consume all the compute, sell it to others. If demand picks up, reclaim it for internal use.

It's akin to installing a safety valve on AI capex.

This is the true significance of the news: it's not that Meta is abandoning frontier AI, nor is it suddenly becoming AWS. It's more like Zuckerberg telling the market:

The market fears Meta's spending, and Meta has now found a revenue outlet for its GPUs.

Impact on the AI Market

1. The Narrative of Compute Scarcity Begins to Crumble

The most valuable aspect of AI compute in the past was its scarcity.

Whoever could get Nvidia GPUs had bargaining power. Whoever could put GPUs in datacenters could raise funds. Companies like CoreWeave thrived on this dynamic.

But if Meta starts selling excess compute, things will get complicated.

Today, it's Meta.
Tomorrow, it could be more hyperscalers.
SpaceX has already been referenced by the market.

This doesn't mean cloud computing prices will crash immediately. High-quality AI clusters require more than just GPUs; they need power, networking, storage, scheduling, stability, and customer support. But the narrative of "only a few have GPUs" is definitely less compelling than before.

Going forward, cloud computing companies can't just tout their GPUs. They must prove:

• Lower costs than big platforms;
• Diverse customer base;
• Renewable contracts;
• Sustainable gross margins;
• Funding not solely reliant on the market's belief in perpetual GPU scarcity.

Any of these points failing would hurt their valuation.

2. Neoclouds Face the Toughest Challenge

AI neoclouds like CoreWeave and Nebius aren't without value. They address real needs and have played a crucial role during AI infrastructure shortages.

But the stock market values marginal changes.

Meta's move makes the market consider an uncomfortable question:

If hyperscalers build too much capacity and then sell the excess, how much premium will neoclouds retain?

CoreWeave's past story was strong: acquire GPUs, sign long-term contracts, raise funds, and continue expanding.
This story fears two things the most:

First, GPU rental prices falling.
Second, customers discovering hyperscalers can sell too, with more stability, lower costs, and better services.

So it's not that CRWV and NBIS fundamentals immediately turn bad. Demand remains. But the valuation logic has changed.

Previously, the market was willing to pay a "scarcity premium" for their compute power.
Now, that premium will be discounted.

3. AI Application Layer Might Actually Benefit

If more hyperscalers sell excess capacity, smaller AI companies and the application layer will benefit.

The reason is simple: more suppliers mean easier price negotiations.

AI application companies' biggest pain point is the cost of inference and training. The more expensive compute power is, the harder it is for products to achieve healthy margins. As long as compute supply increases, even if it's just an expectation, pressure on the application layer will ease.

This segment is worth watching more closely.

The market used to love buying "shovel sellers." But if shovels start competing on price, those who can actually use them to dig up value will find themselves in a better position.

4. Nvidia Not in Immediate Danger, but Market Will Change Its Question

Meta selling compute power isn't immediately negative for Nvidia ($NVDA).

On the contrary, it indicates big companies are still serious about building AI infrastructure. If Meta didn't have sustained demand, it wouldn't spend so much on data centers and chips.

But Nvidia will face a tougher question down the road.

Previously, investors asked:

How many GPU orders are left?

Going forward, they'll ask:

Can these GPU customers make money with them?

These questions are different.

If Meta can sell excess capacity at high prices, Nvidia's demand thesis strengthens.
If Meta can only rent at low prices, or if a price war erupts, Nvidia's orders may hold, but its valuation will be pressured first.

So GoBull Research's take on NVDA is: short-term neutral, medium-term watch two metrics.

One is GPU rental prices.
The other is hyperscalers' capex return rates.

Impact on US Stocks

Meta ($META)

Meta is the most direct winner from this news.

It solves a narrative problem for them: previously, investors saw "Zuck burning cash again"; now, they can at least say "the infrastructure bought with this money can be sold to external customers."

This doesn't mean Meta's cloud business will necessarily succeed.

Meta lacks AWS's enterprise cloud sales system and Azure's enterprise software integration. Its strengths lie in consumer distribution, social data, open-source ecosystem, and hyperscale infrastructure. If it only sells raw compute power, margins may not look good. If it can package Llama / model hosting / API / developer ecosystem together, the story becomes bigger.

GoBull Research's conclusion on META:

This news supports valuation recovery, but future revenue disclosures will be key. Relying solely on "potential compute power sales" won't last long.

CoreWeave ($CRWV), Nebius ($NBIS)

These two are the most sensitive.

The market has already voted with its stock price. Meta up, CRWV and NBIS down, and this is no coincidence.

Their real problem isn't disappearing demand, but weakening bargaining power. AI compute demand remains strong, but once supply shifts from "few neoclouds" to "hyperscalers' excess capacity," valuations must be recalculated.

Key points to watch next:

• GPU rental prices: Will they stabilize or continue falling?
• Customer retention: Can they keep clients from migrating to hyperscalers?
• Product differentiation: Can they offer unique services hyperscalers can't match?

AI Application Layer ($AI)

This segment may benefit the most.

More supply means lower prices. If hyperscalers sell excess capacity, AI startups and application companies will face less pressure on compute costs.

This could accelerate AI adoption and innovation.

Investors should watch companies in this space for potential upside.

Nvidia ($NVDA)

Nvidia isn't in immediate danger, but the market will shift its focus.

Meta's move shows demand for AI infrastructure remains robust. However, the key question will evolve.

Investors will increasingly scrutinize whether Nvidia's customers can monetize their AI investments.

If hyperscalers can sell excess capacity profitably, Nvidia's demand outlook strengthens.
If they can't, Nvidia's orders may hold, but its valuation could face pressure.

GoBull Research maintains a neutral stance on NVDA for now, but advises watching GPU rental prices and hyperscaler capex return rates closely.

• Whether the backlog continues to grow;
• If gross margins are being squeezed by price competition;
• If the concentration of large customers is decreasing.

If any of these three deteriorates, the stock will struggle.

Amazon ($AMZN), Microsoft ($MSFT), Alphabet ($GOOGL)

Meta entering AI cloud sounds like added competitive pressure on the big three clouds.

But this isn't a major issue for AMZN, MSFT, and GOOGL.

Enterprise customers buy cloud services for more than just GPUs. They purchase databases, storage, security, permissions, networking, compliance, billing systems, operational expertise, and a suite of integrated workflows.

Meta can impact the edge pricing of AI compute, but it's unlikely to immediately capture the core enterprise cloud market.

What's more likely: the AI cloud market continues to expand, the big three continue to gain enterprise customers, while Meta and neocloud compete in specific model, developer, and bare compute markets.

This is neutral to positive for the big three. Price competition will occur, but demand is being validated.

Nvidia ($NVDA), Broadcom ($AVGO), AMD ($AMD)

Chip stocks shouldn't panic over this news in the short term.

Meta selling compute capacity implies they've already bought, and may continue buying, significant compute power. For upstream suppliers, this indicates ongoing demand for AI infrastructure.

However, the market's tolerance for "unlimited capex" should be tempered.

Going forward, the market won't just reward orders. It will scrutinize customer ROI.
The chipmakers who enable lower unit costs for customers will be more stable.

Nvidia remains central. Broadcom benefits from custom chips and networking. AMD still has room to catch up. The difference is that the next phase of semiconductor valuation will depend more on customers' actual monetization capabilities.

Power, Cooling, Engineering Chain

The AI infrastructure chain, including Vertiv ($VRT), Eaton ($ETN), GE Vernova ($GEV), ABB, Schneider, and Jacobs ($J), still holds.

AI data centers can't rely solely on GPUs. Power, cooling, transformers, liquid cooling, and construction are all hard constraints.

But this news also reminds the market: once data centers are built, they need users.

If local compute overcapacity emerges in the future, this chain won't collapse immediately, but valuations will become more discerning. Companies that can fulfill orders will remain strong, while those relying on AI hype will fall.

GoBull Research's Ranking

Based on this news alone, the ranking is:

First Tier: META.
The event directly improves the narrative, clearest short-term impact.

Second Tier: GOOGL / MSFT / AMZN.
AI cloud demand continues to be validated, large platforms retain ecosystem advantages.

Third Tier: NVDA / AVGO.
The logic holds, but enters a "customer ROI scrutiny period."

Caution Warranted: CRWV / NBIS.
Not that they can't rise, but the scarcity premium in their valuations needs recalculating.

Monitor Closely: VRT / ETN / GEV / J.
AI factory construction will continue, but focus should be on orders and margins, not just the story.

Three Scenarios

GoBull Research currently favors the first scenario.

Meta isn't immediately becoming the fourth major cloud. A more realistic path is: sell excess capacity first to find a recovery outlet for AI capex, then see if developers and model ecosystems can follow.

Final Conclusion

Meta selling compute capacity shouldn't be viewed as an AI bubble burst.

It's more like a watershed moment.

Previously, the market rewarded "buying GPUs aggressively."
Going forward, the market will reward "buying GPUs and making money."

This isn't the end for the US AI sector, but a weeding out.
Platform companies will be more comfortable.
Pure compute intermediaries will face more challenges.
Chip and data center chains remain in the game, but the market will start scrutinizing whether their customers are truly profitable.

GoBull Research interprets this news as:

The AI infrastructure trade isn't over, but the era of mindless GPU buying has passed.

Only four metrics truly matter here:

1. Whether Meta will officially announce Meta Compute or a similar business;
2. If GPU rental prices have started to decline;
3. Whether CRWV and NBIS's gross margins and backlog have changed;
4. If Nvidia's major clients' capex guidance has softened.

If all four metrics remain stable, the AI theme continues.
If GPU rental prices soften first, the market will reprice the entire AI infrastructure basket.

Sources

• Business Insider: Meta stock spikes 10% on reports of a plan to sell excess AI compute
• MarketWatch: Is Meta 'giving up' on cutting-edge AI? Wall Street is divided over potential cloud pivot
• MarketWatch: CoreWeave, Nebius shares tumble as Meta stands to become a fresh threat in the cloud
• WSJ: Meta Stock Soars Nearly 9% on Reported Plan to Build Cloud Infrastructure
• Investopedia: Meta Plans to Launch a Cloud Computing Business, Sending Stock Surging
• TechRadar: Meta cloud computing business 'definitely on the table', says Mark Zuckerberg
• Tom's Hardware / Financial Times data: Google, Microsoft, Meta, and Amazon capex spending to hit $725 billion in 2026

Disclaimer: This is market research and content analysis, not investment advice. Stock prices and news developments change rapidly. Make trading decisions based on the latest announcements, financial reports, and your personal risk tolerance.