Duan Yongping's Bottom-Fishing in CoreWeave Is Turning into a Battleground for Bulls and Bears

marsbitPublished on 2026-05-11Last updated on 2026-05-11

Abstract

CoreWeave's Q1 2026 earnings report has intensified the ongoing bull-bear battle over the AI infrastructure stock. While revenue doubled year-over-year to $2.08B and the firm's remaining performance obligation (RPO) surged to nearly $100B, its net loss more than doubled to $740M. The critical point of contention is profitability: while adjusted EBITDA margin was a robust 56%, the adjusted operating margin collapsed to just 1% due to soaring infrastructure and sales costs. A weaker-than-expected Q2 revenue guidance further triggered an 11.4% single-day stock drop. The bull thesis hinges on CoreWeave's massive order backlog, deep strategic ties with NVIDIA (as a customer, investor, and key supplier), and a diversified client base now including Anthropic and Meta. The bear case focuses on the "scale at all costs" model, where expanding revenue leads to wider losses, ballooning debt ($25B), and massive capital expenditures ($6.8B in Q1). Insider selling by executives contrasts with a notable new investor: Chinese investor Duan Yongping initiated a small position (0.12% of his portfolio) in Q4 2025 near the stock's lows. The coming Q2 report is seen as a key test for management's promise of a profit margin recovery.

Author: Deep Tide TechFlow

On May 8, AI cloud computing power supplier CoreWeave (CRWV) plunged 11.4% in a single day, closing at $114.15. This marked another "earnings day decline" since the company's IPO in March last year. However, unlike previous drops, this decline coincided with a more dramatic contrast: Duan Yongping, widely known in the Chinese-speaking community as a disciple of Warren Buffett, had just established his first position in CoreWeave in Q4 2025, with an allocation of approximately $20 million. Based on the position size and the Q4 average price, the entry point was close to CoreWeave's lowest level for the year in December 2025.

CoreWeave is currently one of the most divisive AI assets in the US stock market. On one side lies the narrative of being the "picks and shovels" seller with nearly $100 billion in order backlog and deep ties to Nvidia; on the other side is the financial reality of widening losses despite scale expansion and continuous insider selling. The Q1 earnings report acted like a prism, making this divergence exceptionally clear.

Q1 Earnings: Revenue Doubles but Losses Widen, Q2 Guidance Punctures Valuation

CoreWeave's Q1 revenue was $2.08 billion, a 112% year-over-year increase and a 32% sequential increase, surpassing the LSEG market expectation of $1.97 billion. However, adjusted loss per share was $1.12, worse than the expected loss of $0.90. Net loss expanded to $740 million, more than doubling from $315 million in the same period last year.

What truly ignited the sell-off was the forward guidance. The company provided a Q2 revenue range of $2.45 billion to $2.6 billion, with a midpoint of $2.53 billion, far below the market expectation of $2.69 billion. Simultaneously, the lower limit of full-year 2026 capital expenditure was raised from $30 billion to $31 billion, which CFO Nitin Agrawal attributed to rising component costs.

The fragility of the profit structure was exposed. Q1 adjusted EBITDA reached $1.16 billion (a 56% margin), seemingly impressive, but adjusted operating profit was only $21 million, with the operating margin compressed to 1%. The reason lies in technology and infrastructure costs soaring 127% year-over-year to $1.27 billion, and sales & marketing expenses skyrocketing over 6 times year-over-year to $69 million. Revenue is growing, but costs are growing faster.

CEO Michael Intrator emphasized on the earnings call: "We have reached hyperscale." He disclosed that the company now has 10 customers committed to spending over $1 billion each. Compared to 2024 when 62% of revenue relied on Microsoft as a single customer, concentration risk has significantly improved. Intrator also expects CoreWeave's annualized revenue to exceed $30 billion by the end of 2027.

Bull Narrative: $100 Billion Order Backlog, Deep Ties with Nvidia

The core supporting the bull case is the order backlog. As of the end of Q1, CoreWeave's remaining performance obligations (RPO) amounted to $99.4 billion, a net increase of approximately $33 billion sequentially and nearly quadrupling year-over-year. Intrator stated that new contracts signed in Q1 alone exceeded $40 billion.

The client list is also reshaping market perceptions. In Q1, Anthropic was added as a customer, providing computing power for its Claude series models; a $2.1 billion AI cloud agreement was signed with Meta; trading firm Jane Street committed to approximately $6 billion in orders and separately completed a $1 billion equity investment. Nvidia again purchased $2 billion worth of CoreWeave Class A common stock this quarter. This global largest GPU supplier is simultaneously an investor and a major customer of CoreWeave. This triple-role binding relationship is referred to as Nvidia's "favored son."

Regarding financing structure, CoreWeave completed an $8.5 billion investment-grade HPC (High-Performance Computing) secured delayed draw term loan (DDTL) in Q1, priced below 6%, hailed by management as "groundbreaking." Year-to-date, the company has secured over $20 billion in debt and equity financing, with the weighted average cost of debt decreasing by approximately 80 basis points. S&P Global Ratings concurrently revised CoreWeave's credit rating outlook from "stable" to "positive."

Bear Logic: Less Profitable as it Scales, Debt Snowball Growing Larger

However, another set of numbers in the earnings report is causing anxiety. Q1 capital expenditure reached $6.8 billion, and the company expects Q2 capital expenditure to further climb to between $7 billion and $9 billion. The Q2 interest expense guidance range is $650 million to $730 million, reflecting the rapid swelling of debt.

The total debt is already staggering. As of the end of Q1, CoreWeave's total debt stood at approximately $25 billion. Relative to the company's current annualized revenue level, this leverage is significantly higher than that of traditional cloud service providers. Morgan Stanley data shows that CoreWeave's debt financing for the full year 2025 amounted to about $11.8 billion, far exceeding its equity financing of about $1.5 billion. The company's core expansion tool is the DDTL, a "book first, finance later" model that uses order contracts as collateral to secure bank financing for GPU procurement.

The sharpest criticism comes from profit quality. Although management repeatedly emphasized the 56% EBITDA margin, the adjusted operating margin was only 1%. The "real" gross margin after deducting technology and infrastructure costs was about 4%, showing compression both sequentially and against market expectations. Intrator attributed this on the earnings call to the temporary effect of rapid scaling, where the dilution effect on margins from newly added capacity is immense when a company expands dramatically from a 1-gigawatt operational scale. He promised this was the "margin low point" and that margins would gradually recover in future quarters.

But the market is currently unwilling to pay for this promise. While analysts from Morgan Stanley and Jefferies gave positive reviews, CoreWeave has experienced short-term pullbacks after each earnings report previously, with this decline being one of the steeper post-earnings drops.

Insider Selling Continues, Mirroring Duan Yongping's Bottom-Fishing

Around the earnings release, CoreWeave insiders maintained their selling pace. CEO Mike Intrator sold 307,693 shares in late April; co-founders Brian Venturo and Chen Goldberg also recorded sales; institutional shareholder Magnetar Financial had previously sold over $300 million worth. Latest disclosures show that another major shareholder recently sold approximately 1.2 million shares.

This forms a stark contrast with Duan Yongping's position-building in Q4. According to a 13F filing disclosed by H&H International Investment in February 2026, Duan Yongping established his first position in CoreWeave of 299,900 shares in Q4 2025, at a time when the stock had retreated more than 65% from its highs, and market concerns over its debt structure had peaked.

It's noteworthy that CoreWeave accounts for only 0.12% of Duan Yongping's total H&H holdings, constituting a "small pilot" position. In the same quarter, Duan massively increased his stake in Nvidia by over 1110% and established new positions in Credo Technology (high-speed connectivity) and Tempus AI (AI healthcare). These three new AI positions combined account for less than 0.3% of his portfolio. This indicates Duan's true heavy bet is Nvidia itself, with CoreWeave being more like a small extension of his bet on the downstream AI computing power supply chain.

The Crucial Question Now: Inflection Point or Trap?

Intrator posed a somewhat emotional rhetorical question during the Q&A session: "I've always felt that everyone is staring at the stock price tree and missing the entire forest."

This statement aptly summarizes the current bull-bear standoff. Bulls see a forest of nearly $100 billion in contract backlog, a diversified customer base, a triple-binding relationship with Nvidia, and an upgraded credit rating. Bears see a tree of 1% operating margins, widening net losses, aggressive capital expenditures, and persistent insider selling.

CoreWeave's stock price is still up nearly 80% year-to-date and has gained over 200% since its IPO. But when the bull case for a stock rests on a distant narrative and the bear case on current numbers, every earnings report becomes a battleground for these two narratives. Duan Yongping previously stated in an interview with Fang Sanwen: "AI is a massive revolution brought about by a qualitative leap in computing power, with an impact potentially exceeding that of the internet and the industrial revolution. There is a clear AI bubble now, and 90% of companies may be eliminated, but the survivors will become the next-generation giants." His 0.12% light position itself acknowledges the uncertainty of this gamble.

The next test is already clear: the Q2 earnings report. If the operating margin does not recover as promised by management, the credibility of the "forest" narrative will face a genuine stress test.

Related Questions

QWhat were the key highlights and concerns from CoreWeave's Q1 2026 earnings report?

ACoreWeave's Q1 2026 revenue was $2.08 billion, up 112% year-over-year and exceeding expectations. However, the company's net loss widened to $740 million and it provided weaker-than-expected Q2 revenue guidance. A major concern was the adjusted operating profit margin collapsing to just 1%, despite a high adjusted EBITDA margin of 56%, due to surging technology and infrastructure costs.

QWhat are the main arguments of the bulls (optimists) and bears (pessimists) on CoreWeave?

ABulls are optimistic due to CoreWeave's massive $99.4 billion Remaining Performance Obligations (RPO), deep strategic and financial ties with NVIDIA, a diversified and growing client base including Anthropic and Meta, and recent credit rating upgrades. Bears are pessimistic because the company is scaling unprofitably, with operating margins near zero, rapidly escalating capital expenditures and debt (now ~$25 billion), and persistent insider selling by executives and major shareholders.

QWhat is the significance of Duan Yongping's (段永平) investment in CoreWeave, and how does it contrast with insider activity?

ADuan Yongping, a well-known value investor in China, initiated a small position in CoreWeave (worth ~$20 million, ~0.12% of his portfolio) in Q4 2025 when the stock was near its yearly low. This 'dip-buying' action contrasts sharply with recent and ongoing insider selling by CoreWeave's CEO, co-founders, and major institutional shareholders like Magnetar Financial. Duan's position is seen as a high-conviction, speculative bet on the AI infrastructure theme, while insider sales raise questions about near-term confidence.

QHow is CoreWeave financing its aggressive expansion, and what are the associated risks?

ACoreWeave is financing its massive GPU procurement and data center build-out primarily through debt, specifically a novel 'Delayed Draw Term Loan (DDTL)' structure secured by its customer contracts. The company raised $8.5 billion in such debt in Q1 alone. The major risk is the rapid ballooning of its total debt to approximately $25 billion, leading to high interest expenses. This 'asset-light, contract-heavy' model creates significant financial leverage, making the company highly sensitive to execution risks and potential slowdowns in AI demand.

QWhat is the critical test for CoreWeave's investment narrative according to the article?

AThe next critical test is the Q2 2026 earnings report. Management, particularly CEO Michael Intrator, has promised that Q1 represented the 'low point' for operating margins and that profitability should improve in subsequent quarters. If the Q2 report fails to show a meaningful rebound in operating profit margins, the company's long-term growth narrative (the 'forest') will face severe credibility pressure from the current poor financial metrics (the 'tree'), potentially leading to further significant stock price volatility.

Related Reads

Morning Post | Trump Media Group Releases Q1 Financial Report; Top Three DeFi Applications Return Nearly $100 Million in Revenue to Token Holders in 30 Days; Michael Saylor Shares Bitcoin Tracker Info Again

**Title: Daily Briefing | Trump Media Group Releases Q1 Report; Top 3 DeFi Apps Return Nearly $100M to Token Holders; Michael Saylor Signals Potential Bitcoin Buy** **Summary:** Key developments in the past 24 hours include: * **Economic Outlook:** Goldman Sachs has pushed back its forecast for the next two Federal Reserve interest rate cuts to December 2026 and March 2027, citing persistent inflationary pressures from energy costs. This delayed timeline is expected to tighten liquidity flow into risk assets, including cryptocurrencies. * **DeFi & Revenue:** Data from DefiLlama shows that three leading DeFi applications—Hyperliquid, Pump.fun, and EdgeX—collectively distributed $96.3 million in revenue to their token holders over the last 30 days. This trend highlights a shift in the crypto community's focus towards real protocol earnings and sustainable economic models. * **Corporate Bitcoin Moves:** Michael Saylor, founder of MicroStrategy (note: referred to as 'Strategy' in the text, likely a typographical error), has signaled potential upcoming Bitcoin purchases by posting a "Bitcoin Tracker" update, following a pattern that typically precedes the company's official disclosure of new acquisitions. * **Market Integrity:** Prediction market platform Polymarket announced updates to address platform issues, including identifying and banning clusters of accounts involved in "ghost-fill" activities and implementing measures to prevent bulk account creation. * **Regulation:** The Bank of England Governor warned that stablecoin regulation could lead to tensions between US and international regulators. In South Korea, the National Tax Service has launched a pilot program to entrust seized virtual assets to private custody firms for management. * **Meme Token Trends:** GMGN data lists the top trending meme tokens on Ethereum (e.g., HEX, SHIB), Solana (e.g., FWOG, TROLL), and Base (e.g., SKITTEN, PEPE) over the past day. **Financial Note:** Trump Media & Technology Group reported a Q1 loss of approximately $4 billion, primarily attributed to unrealized losses on its Bitcoin and other digital asset holdings.

链捕手9m ago

Morning Post | Trump Media Group Releases Q1 Financial Report; Top Three DeFi Applications Return Nearly $100 Million in Revenue to Token Holders in 30 Days; Michael Saylor Shares Bitcoin Tracker Info Again

链捕手9m ago

Telegram Takes Direct Control of TON, Social Traffic Rewrites the Public Chain Narrative

Telegram founder Pavel Durov announced that Telegram will replace the TON Foundation as the core driver and largest validator of The Open Network (TON). Key initiatives include a sixfold reduction in transaction fees, performance upgrades, and improved developer tools within the next few weeks. This marks a strategic shift from Telegram merely providing user access to deeply integrating TON into its platform's core infrastructure. The goal is to transform Telegram's massive social traffic into sustainable on-chain activity. While viral mini-apps like Notcoin have demonstrated Telegram's ability to drive user adoption, TON aims to support frequent, low-value transactions inherent to social platforms—such as tipping, in-app payments, and game rewards. Ultra-low fees and sub-second finality (0.6 seconds) are crucial to making blockchain interactions seamless and nearly invisible within the Telegram user experience. However, Telegram's increased central role raises questions about network decentralization. Durov argues that Telegram's participation will attract more large validators, thereby enhancing decentralization. TON also offers high annual staking rewards (18.8%), aiming to retain capital within its ecosystem. The fundamental challenge for TON is no longer leveraging Telegram's user base, but becoming an indispensable, seamless infrastructure layer for Telegram's everyday applications—moving from an adjacent chain to an embedded utility.

marsbit11m ago

Telegram Takes Direct Control of TON, Social Traffic Rewrites the Public Chain Narrative

marsbit11m ago

OpenAI Post-Training Engineer Weng Jiayi Proposes a New Paradigm Hypothesis for Agentic AI

OpenAI engineer Weng Jiayi's "Heuristic Learning" experiments propose a new paradigm for Agentic AI, suggesting that intelligent agents can improve not just by training neural networks, but also by autonomously writing and refining code based on environmental feedback. In the experiment, a coding agent (powered by Codex) was tasked with developing and maintaining a programmatic strategy for the Atari game Breakout. Starting from a basic prompt, the agent iteratively wrote code, ran the game, analyzed logs and video replays to identify failures, and then modified the code. Through this engineering loop of "code-run-debug-update," it evolved a pure Python heuristic strategy that achieved a perfect score of 864 in Breakout and performed competitively with deep reinforcement learning (RL) algorithms in MuJoCo control tasks like Ant and HalfCheetah. This approach, termed Heuristic Learning (HL), contrasts with Deep RL. In HL, experience is captured in readable, modifiable code, tests, logs, and configurations—a software system—rather than being encoded solely into opaque neural network weights. This offers potential advantages in explainability, auditability for safety-critical applications, easier integration of regression tests to combat catastrophic forgetting, and more efficient sample use in early learning stages, as demonstrated in broader tests on 57 Atari games. However, the blog acknowledges clear limitations. Programmatic strategies struggle with tasks requiring long-horizon planning or complex perception (e.g., Montezuma's Revenge), areas where neural networks excel. The future vision is a hybrid architecture: specialized neural networks for fast perception (System 1), HL systems for rules, safety, and local recovery (also System 1), and LLM agents providing high-level feedback and learning from the HL system's data (System 2). The core proposition is that in the era of capable coding agents, a significant portion of an AI's learned experience could be maintained as an auditable, evolving software system.

marsbit1h ago

OpenAI Post-Training Engineer Weng Jiayi Proposes a New Paradigm Hypothesis for Agentic AI

marsbit1h ago

Your Claude Will Dream Tonight, Don't Disturb It

This article explores the recent phenomenon of AI companies increasingly using anthropomorphic language—like "thinking," "memory," "hallucination," and now "dreaming"—to describe machine learning processes. Focusing on Anthropic's newly announced "Dreaming" feature for its Claude Agent platform, the piece explains that this function is essentially an automated, offline batch processing of an agent's operational logs. It analyzes past task sessions to identify patterns, optimize future actions, and consolidate learnings into a persistent memory system, akin to a form of reinforcement learning and self-correction. The article draws parallels to similar features in other AI agent systems like Hermes Agent and OpenClaw, which also implement mechanisms for reviewing historical data, extracting reusable "skills," and strengthening long-term memory. It notes a key difference from human dreaming: these AI "dreams" still consume computational resources and user tokens. Further context is provided by discussing the technical challenges of managing AI "memory" or context, highlighting the computational expense of large context windows and innovations like Subquadratic's new model claiming drastically longer contexts. The core critique argues that this strategic use of human-centric vocabulary does more than market products; it subtly reshapes user perception. By framing algorithms with terms associated with consciousness, companies blur the line between tool and autonomous entity. This linguistic shift can influence user expectations, tolerance for errors, and even perceptions of responsibility when systems fail, potentially diverting scrutiny from the companies and engineers behind the technology. The article concludes by speculating that terms like "daydreaming" for predictive task simulation might be next, continuing this trend of embedding the idea of an "inner life" into computational processes.

marsbit1h ago

Your Claude Will Dream Tonight, Don't Disturb It

marsbit1h ago

Trading

Spot
Futures
活动图片