In June 2026, the AI computing power investment boom continued to heat up. In the “AI Factory” ecosystem vision proposed by Nvidia CEO Huang Renxun, three companies marked as “priority to track”—CRWV, NBIS, and IREN—happen to have a high overlap with the Q1 holdings of Situational Awareness LP, the Situational Awareness LP run by 25-year-old “Wall Street AI stock god” Leopold Aschenbrenner.
This phenomenon is not accidental. Demand for computing power from AI large-model training and inference is shifting from “chip procurement” to “computing power infrastructure as a service.” As a GPU supplier, Nvidia needs to support cloud platform companies that can deploy its chips at scale and deliver computing power to end customers. As a hedge fund focused on long-term AI trends, Situational Awareness LP’s holding logic points to AI cloud service providers with scalable delivery capabilities and healthy financial structures as well.
As of June 3, 2026, these three companies’ year-to-date gains are approximately 74% for CRWV, about 216% for NBIS, and about 73% for IREN. The market is voting with capital, validating the value of this consensus’ anchor.
What core business-model differences exist among the three AI cloud platform companies?
Although they all belong to the AI infrastructure track, CRWV, NBIS, and IREN have significant differences in business structure.
NBIS (Nebius) is defined by the market as the “purest” Neocloud (new cloud) pure-play. Its core business is high-performance computing cloud services for AI training and inference needs, with customers mainly consisting of model training teams and AI application developers. NBIS’s strength lies in the fact that most of its revenue comes from long-term contracts with mega-scale customers, resulting in relatively high cash-flow certainty.
CRWV (CoreWeave) initially focused on managing Ethereum mining computing power, then fully shifted to AI rendering and general-purpose GPU cloud services. Its key feature is deep binding to Nvidia’s chip supply chain, enabling it to obtain batch allocations of H-series and subsequent B-series GPUs ahead of others. But CRWV’s high-growth profile comes with high debt leverage, and the market has been discussing its ability to keep funding.
IREN (Iris Energy) shows a “dual-track” characteristic: it has a Bitcoin mining business as a stable cash-flow source, while extending remaining computing capacity into the AI cloud services sector. This structure makes its computing power scheduling more flexible, but it also faces skepticism about “full-stack execution uncertainty”—whether it can simultaneously balance two business lines with very different technical routes.
What core competitive moat is behind NBIS being heavily held by multiple parties?
In Q1, Situational Awareness LP held more than $2 billion worth of NBIS shares, while Serenity—Reddit WSB’s veteran trader—listed it as the “highest-confidence choice” in the Neocloud track. NBIS’s differentiated advantages mainly show up in three areas.
- Revenue scale and customer quality. In the Neocloud track, NBIS’s revenue scale is among the leading positions, and its customer base is dominated by large AI labs and enterprise users. Contract terms are longer, reducing the risk of cyclical volatility.
- Avoiding competitors’ typical pain points. In a November 2025 analysis, Serenity pointed out that NBIS does not have the full-stack execution uncertainty like IREN and ORCL; does not carry high-interest debt like CRWV and APLD; and does not lack revenue certainty derived from mega-scale customer contracts like CLSK and BITF. This “no short-board” financial and operational structure makes it the preferred allocation for institutional capital in the Neocloud track.
- Strategic investment endorsement from Nvidia. NBIS is listed in Nvidia’s official investment portfolio, meaning it has certain advantages in GPU supply priority and technical synergy. In a market where GPUs remain in persistent supply shortage, supply-chain stability itself becomes a scarce competitive moat.
How are CRWV’s and IREN’s respective growth drivers and potential risks balanced?
CRWV’s growth core is its close cooperation with Nvidia. As one of the earliest cloud service providers to obtain capabilities to deploy large-scale H100 clusters, CRWV has accumulated real operational experience in high-performance computing scenarios such as AI rendering and drug discovery. However, its relatively high debt level on the balance sheet keeps the market attentive to its refinancing ability in a high-interest-rate environment.
IREN’s uniqueness lies in its “mining + AI” dual-business structure. Bitcoin mining provides a predictable USD revenue base, while AI cloud services offer high-growth upside elasticity. But this structure also brings complexity in technical routes and resource allocation: the ASIC devices used in mining cannot be directly repurposed for the GPU clusters required by AI cloud services, meaning the company must invest capital expenditures in two directions at the same time. There is still disagreement in the market over whether it can maintain execution strength on both fronts.
It’s worth noting that Situational Awareness LP held shares of both of these companies in Q1, but additionally allocated $141 million in call options for CRWV. This “stock + options” combination suggests it expects a higher sensitivity to short-term price volatility of CRWV, making it a more tactical allocation.
How should we understand Situational Awareness LP’s holding changes and hedging signals?
Based on the Q1 13F filing, Situational Awareness LP’s optical communications company LITE, which it held in Q4 2025, has been fully exited by Q1 2026. Meanwhile, its position in Intel (INTC) shows a “small equity stake + large put options” structure, with an overall shift toward hedging or a cautious stance.
These adjustments convey two signals. First, the investment focus for AI infrastructure is shrinking from “peripheral support” (such as optical modules and packaging materials) toward the “core computing power delivery layer.” The exit of LITE, along with increases in CRWV, NBIS, and IREN, reflects a trend of capital concentrating on the AI cloud platform “computing power outlet” segment.
Second, for traditional semiconductor giants like INTC, the fund believes uncertainty in AI computing-power competition is relatively higher. Therefore, it uses put options for risk hedging rather than fully liquidating. This approach keeps room for upside while limiting drawdown risk, which is typical of event-driven position management.
As for Synopsys (SNPS), Marvell (MRVL), Nokia (NOK), and other frequently discussed recent targets, none of them appeared in its main holdings list, indicating these companies are not currently within the fund’s long-term conviction range.
From “screw parts” to “infrastructure,” how is the AI investment logic evolving?
Serenity’s research framework offers a useful perspective: focusing on “key bottleneck points” in Nvidia’s AI chip industry chain. Companies in these “screw-part” categories usually have scarcity and pricing power, but the market upside ceiling is relatively clear.
Meanwhile, the market’s attention on AI cloud platforms such as NBIS, CRWV, and IREN suggests that investment logic is shifting from “selling shovels” to “running mines.” AI cloud platform companies do not directly produce chips, but they are a necessary pipeline for converting chip capabilities into customer value. As AI applications shift from the training stage toward the inference stage, the distribution of computing-power demand will become more decentralized, and the distribution value of cloud platforms will become even more prominent.
This evolution also explains why the three companies have such different valuation logic: NBIS receives a premium due to high revenue certainty and low-debt structure; CRWV prices in elasticity after risk discounts due to high growth and high leverage; and IREN faces a rebuild of its composite valuation model because of its dual-business structure.
What major disagreements does the market currently have about these three AI cloud platform companies?
Although all three achieved significant gains in 2026, the market’s disagreements about their sustainability still remain.
- Valuation vs. growth fit. NBIS’s market cap is $66 billion, and Serenity believes it could break through $100 billion in the long run. But critics argue that competition in the Neocloud track is intensifying, and traditional cloud providers (such as AWS and Azure) are also increasing investment in AI computing-power services, which could pressure mid-to-long-term profit margins.
- Safety of debt leverage. CRWV’s high-interest debt structure could become a factor suppressing the stock price when market risk appetite declines. Supporters argue that as long as the growth in GPU lease yields covers the interest cost, leverage is an effective tool to accelerate expansion.
- The primary/secondary relationship between AI and crypto businesses. IREN’s dual-business structure allows it to flexibly reallocate computing resources, but it also creates ambiguity about whether it is an AI company or a miner. Different investors, based on different valuation models, may have significantly different views on its reasonable price range.
These disagreements in themselves do not negate investment value; they mean the three types of targets are suitable for funds with different risk preferences. The core of the game is not “who is the best,” but “who best fits the current portfolio structure.”
What analytical framework should be built to track the AI cloud platform sector?
Based on the discussion above, a relatively complete tracking framework can include four dimensions.
- Supply-chain security: Does the company have stable GPU supply commitments from Nvidia or AMD? Does it have exclusive or priority supply agreements?
- Customer quality and contract structure: What proportion of revenue comes from mega-scale customers (such as annual contract amounts exceeding $50 million)? Are contract terms mostly multi-year?
- Financial health: Are the debt-to-asset ratio and the EBITDA interest coverage multiple in a safe range? Under assumptions of unchanged or rising interest rates, how much refinancing pressure does the company face?
- Computing power utilization and unit-economics model: Is the average utilization of the GPU cluster maintained above 60%? Is the gap between unit computing rental income and operating costs stable or expanding?
Across these four dimensions, the three consensus picks have their own strengths and weaknesses, which is also the logical basis for the different valuation premiums the market grants them. There is no perfect target—only the choice that best matches one’s own judgment.
FAQ
Q: Are CRWV, NBIS, and IREN crypto assets or traditional stocks?
These three companies are all traditional stocks listed in the United States. Among them, IREN also has Bitcoin mining operations, which makes it more closely linked to the crypto industry. Investors can look to Gate for trading opportunities around AI and crypto fusion themes.
Q: Do the historical gains of these three companies imply they are still worth关注 now?
Historical gains do not guarantee future performance. All three are in high-growth but high-volatility AI infrastructure tracks, so investors should make decisions based on their own risk tolerance and holding-period considerations. This article does not constitute any investment advice.
Q: What signal does Situational Awareness LP’s put-option holdings on INTC convey?
The fund adopts a structure of “small equity stake + large put options” for Intel (INTC), indicating it is cautious about the prospects of this traditional semiconductor giant in the AI competition, but it has not fully abandoned upside potential. This is a strategy expression dominated by hedging.
Q: What is the difference between the Neocloud track and traditional cloud services?
Neocloud (new cloud) specifically refers to GPU computing cloud platforms designed for AI training and inference needs, which are significantly different from traditional general-compute cloud services (CPU computing) such as AWS and Azure in terms of hardware architecture, scheduling systems, and customer base. The Neocloud track is still in a phase of rapid expansion, and the competitive landscape has not yet solidified.
Q: How to track AI infrastructure-related themes on Gate?
Gate provides trading services for a variety of AI + Web3 related assets. Users can indirectly participate in AI infrastructure themes by following targets such as AI track tokens and computing-power protocols. It is recommended to make comprehensive judgments by combining on-chain data and market sentiment indicators.
