CRWV and NBIS surge after earnings! Is AI computing power becoming the next major market trend?
Orders are piling up faster than they can be fulfilled! With Google leading the charge in 'outsourcing compute capacity' and NVIDIA continuing to double down, could Neocloud be poised for a re-rating?
Google Cloud was Google's strongest growth engine this quarter.
According to $Alphabet-C (GOOG.US)$ The latest earnings show that Google Cloud generated $24.8 billion in revenue in Q2, up 82% year-over-year; operating profit reached $8.8 billion, more than tripling year-over-year, and operating margin rose to 35.6% from 20.7% a year earlier.
Source: Google Earnings
However, beyond these already realized impressive profits, two key forward-looking indicators offer clearer guidance on the future direction of the technology hardware and computing power supply chain:
First is the explosive growth in order backlog. Google Cloud’s contract backlog surged by over $50 billion quarter-over-quarter, reaching a staggering $514 billion in total. This directly prompted the company to raise its capital expenditure ceiling further, revising its 2026 guidance upward to $195–205 billion.
Second is management’s signal of 'computing capacity overflow.' During the earnings call, Google executives candidly acknowledged a 'happy problem': the pace of building out internal data centers and computing infrastructure has temporarily fallen behind the surge in customer demand. To ensure it can smoothly onboard large cloud clients and multi-year contracts, Google will expand its use of third-party computing capacity in Q3 as a 'bridge strategy' until its own infrastructure comes online.
What does this mean? Although this strategy may introduce short-term cost volatility, it sends an extremely clear signal to the market: major tech giants’ computing capacity constraints are now spilling over externally.For emerging 'Neocloud' providers currently in their growth phase, this 'bridge' demand from tech giants represents a golden window for revenue breakthroughs and valuation re-rating.
Structural windfall from spillover computing power: How can Neocloud capture the tech giants’ 'massive fortune'?
A new round of the AI computing power arms race is reshaping the cloud computing landscape, with the 'Neocloud' sector experiencing a wave of catalysts. These companies have seen gains of 10%–30% over the past five days.
The core driver behind this latest rally in computing infrastructure is undergoing a profound structural shift:Demand for GPU computing power from cutting-edge AI models, AI agents, and embodied intelligence companies is growing exponentially; however, hyperscalers are showing signs of slowing in data center expansion due to constraints in power supply and hardware supply chains. This supply-demand mismatch has opened up a vast blue-ocean market for independent Neocloud providers.
Recent developments across the industry chain indicate that the Neocloud segment is now experiencing a double boost to both earnings and valuation—a classic Davis double play:
1. Exponential growth in order scale:
Strong market demand for pure-play computing power leasing is translating into massive long-term contracts for Neocloud companies.IREN recently secured a multi-year AI cloud computing contract worth $2.8 billion and aggressively raised its full-year annualized revenue target to over $4 billion, with an order lock-in rate of 85% providing a strong earnings safety cushion.
Similarly, Hut 8’s Texas facility is now fully leased and has signed a 15-year super contract with a top-tier enterprise client, carrying a base value of $9.8 billion. With the client’s computing power demand doubling to 704MW, the agreement’s potential total value could reach as high as $50 billion, dramatically expanding market expectations for its long-term cash flow.
2. Strategic bets by top-tier capital:
Beyond fundamental order support, strategic capital deployments by key players in the industrial chain further validate the sector’s strategic value. TakeNebiusfor example: the latest SEC filings show that NVIDIA has acquired approximately 9.3% beneficial ownership in the company through common shares and prepaid warrants, involving around $2 billion in capital. Backed by this deep strategic alignment and endorsement from a tech giant, institutions have issued extremely optimistic long-term forecasts, expecting Nebius to reach $30 billion in revenue by 2030.
Overall, prior to full capacity ramp-up by hyperscalers, Neocloud—thanks to its flexible deployment capabilities and focused AI compute services—is emerging as the optimal solution for alleviating global AI compute anxiety.
Which Neocloud companies are worth watching?
Previously,SpaceX Exposes High-Priced Compute Contracts, Nebius Announces GPU Price Hikes… Has the 'Turning Point' Arrived for U.S. Stock Compute Leasing Sector?we reviewed cloud service providers. Beyond hyperscale cloud vendors, they broadly fall into two categories: one group consists of 'new clouds' that genuinely sell GPU compute power and cloud services directly to customers; the other comprises HPC operators that own land, power, and data center infrastructure and lease these facilities to tech giants or new cloud providers.
1. New Cloud Service Providers: The 'Pure Water Sellers' of the AI Era
These companies are a new breed born alongside the demand for large model training, focusing exclusively on offering high-end GPU computing power leasing services. They are typically deeply integrated with hardware giants like NVIDIA.
company—not merely renting GPUs, but offering an end-to-end AI cloud platform covering compute, storage, networking, and model deployment. By Q1 2026, CoreWeave’s revenue backlog had reached $99.4 billion, active power capacity exceeded 1 GW, and the company signed new agreements with major clients like Meta and Anthropic. It also plans to co-develop over 5 GW of AI factories with NVIDIA. This indicates CoreWeave is evolving from a 'GPU lessor' into a specialized AI cloud platform capable of competing with traditional cloud giants.
However, CoreWeave is also a classic capital-intensive, highly leveraged company. Its massive order backlog requires upfront investments in data centers, GPUs, and network equipment. While revenue is growing rapidly, interest expenses and capital expenditure pressures are significant. Therefore, the key metric to watch isn’t demand—it’s whether the company can deliver on schedule, manage financing costs, and convert orders into free cash flow.
Compared to simply purchasing GPUs and leasing them out, Nebius can independently design servers, networking, storage, scheduling systems, and cloud software, making it closer to a vertically integrated AI infrastructure company. In March 2026, Nebius signed a new five-year AI infrastructure agreement with Meta, covering fixed dedicated capacity and additional purchasable computing power, with a potential total contract value of up to approximately $27 billion, to be deployed on the Vera Rubin platform.
Such long-term contracts not only lock in future revenue but also help Nebius secure financing and pre-procure GPU and power resources. Its greatest upside lies in the possibility that, as its cloud software and developer tools mature, Nebius may no longer be valued solely based on data center assets but could gradually command a valuation premium typical of AI platform companies.
IREN originally focused primarily on Bitcoin mining, but leveraging its own land, low-cost power, and data center development capabilities, it has quickly moved into AI cloud services. In addition to the aforementioned new contract, the company previously disclosed a multi-year GPU cloud agreement with Microsoft worth approximately $9.7 billion and announced in 2026 plans to expand its AI cloud capacity to 150,000 GPUs.
IREN’s strength lies in its control over the entire stack—from power and data center facilities to GPU clusters—giving it strong cost management capabilities. However, its transformation is still in a rapid expansion phase, and investors should monitor GPU delivery timelines, customer concentration, and whether its substantial capital expenditures can be successfully financed.
The company offers a range of services including GPU cloud, AI and HPC colocation, and private AI clouds, positioning itself more toward enterprise-grade and customized computing solutions. Compared to CoreWeave, Nebius, and IREN, WhiteFiber is smaller in scale; however, if it secures large clients or brings new data centers online, its revenue could show greater elasticity. Conversely, it faces higher risks related to financing capability, customer concentration, and delivery execution, making it a high-volatility investment.
II. 'Compute landlords' controlling power and data centers
Another more direct path to benefit is by leasing AI data centers to hyperscalers or neoclouds. These companies may not operate full cloud platforms themselves but possess the currently most scarce assets: land, power interconnection rights, and construction capabilities.
$Applied Digital (APLD.US)$ and$Hut 8 (HUT.US)$ Is a relatively representative 'AI data center developer' among them
Applied Digital is selling data center capacity to investment-grade hyperscalers through long-term leases of approximately 15 years. The company has announced multiple new campuses and leases for 2026, with contracted capacity exceeding 1 GW, gradually shifting its business model from highly volatile computing power services toward long-cycle rental cash flows.
Hut 8’s progress is even more aggressive. According to its latest announcement in July, the company’s AI data centers have signed IT capacity commitments totaling 949 MW, with base contract value reaching $26.6 billion; if all renewal options are exercised, the potential contract value from the Beacon Point campus alone could reach $50.2 billion.
The advantage of such companies lies in higher revenue visibility and stronger customer credit quality, without directly bearing the risk of rapid GPU price obsolescence due to technological iteration. The downside is the enormous upfront capital required for construction, with ultimate returns heavily dependent on construction costs, financing rates, and whether projects can be delivered on schedule.
A direct reflection of Google's spillover computing demand: $TeraWulf (WULF.US)$ and$Cipher Digital (CIFR.US)$
If we focus solely on the theme of 'Google expanding its use of third-party computing capacity' to identify relevant proxies,TeraWulf and Cipher Digital warrant close attention.。
Google previously provided credit support for certain HPC leases by Fluidstack and TeraWulf and obtained potential equity stakes in TeraWulf through low-strike-price warrants. This arrangement essentially allows Google to leverage its own creditworthiness to help third-party computing providers secure financing for data center construction, thereby locking in future infrastructure capacity ahead of time.
More recently, TeraWulf also signed a 20-year AI data center lease with Anthropic, covering approximately 401 MW of IT load and generating initial contract revenue of about $19 billion. This signifies that the company is no longer merely a 'Bitcoin miner转型 concept' but is now backed by long-term cash flows from major AI customers.
Cipher follows a very similar path. The company signed a 168 MW, ten-year AI colocation agreement with Fluidstack, generating initial contract revenue of approximately $3 billion; Google provided credit support for $1.4 billion of the lease obligations and acquired a potential equity stake of about 5.4%. Cipher has since secured a third investment-grade hyperscaler lease, indicating its AI data center transformation is evolving from a single project into a replicable model.
As for$Riot Platforms (RIOT.US)$ 、 $MARA Holdings (MARA.US)$ 、 $CleanSpark (CLSK.US)$ 、 $Bitdeer Technologies Group (BTDR.US)$ 、 $HIVE Digital Technologies (HIVE.US)$ 、 $Digi Power X (DGXX.US)$ Companies like these, although possessing certain power and data center resources, should not yet be valued as Neoclouds merely because they have announced they are 'exploring AI or HPC.'
Summary
Overall, the signal from Google about spillover computing capacity does not mean that all companies labeled as having data centers will benefit. What will truly be scarce in the future is not just a single GPU order, butimmediately available power, data centers that can be delivered on schedule, long-term customers with strong credit, and access to low-cost financing.Whoever can simultaneously control these four resources will be the most likely next winner in this AI computing arms race.
Risk Disclaimer: The above content only represents the author's view. It does not represent any position or investment advice of Futu. Futu makes no representation or warranty.Read more
Comments (31)
to post a comment
100
276