In late July 2026, Leopold Aschenbrenner’s Situational Awareness LP, a fund that had grown from roughly $383 million to an estimated $45 billion at peak, was forced to liquidate its entire public equities book in a single block trade to Citadel. The fund employed approximately 4x gross leverage and had concentrated its portfolio in AI infrastructure names: semiconductors, memory, data centers, power generation, and crypto miners pivoting to AI compute.
When AI infrastructure stocks declined 27% to 54% from recent peaks, and the fund’s short software positions rallied simultaneously, the losses cascaded into margin calls and forced deleveraging. The fund collapsed from $45 billion to approximately $10 billion in assets within days.
The event has been widely discussed as a leverage story or a cautionary tale about concentration. Both are true. But there is a deeper structural question that traditional portfolio analytics did not surface and that most post mortems have not addressed: what was the forensic risk profile of the individual holdings, and what patterns emerge when you score them across multiple dimensions simultaneously?
We applied SignalVest’s six dimension Red Flag Intelligence Scoring framework to the fund’s core holdings as disclosed in the Q1 2026 13F filing. The results reveal something that sector exposure analysis, factor tilts, and correlation matrices would have missed entirely.
The Forensic Risk Heat Map
We scored six of the fund’s core holdings, CoreWeave (CRWV), Bloom Energy (BE), Nebius Group (NBIS), SK Hynix, IREN, and Core Scientific (CORZ), across all six dimensions of our scoring framework: Capital Structure Risk, Regulatory Exposure, Insider Behavior, Narrative Fragility, Operational Signals, and Governance Integrity.
The composite scores:
Nebius Group (NBIS): 7.5/10 , High Risk
CoreWeave (CRWV): 7.4/10 , High Risk
IREN (IREN): 7.1/10 , High Risk
Bloom Energy (BE): 6.5/10 , High Risk
Core Scientific (CORZ): 6.3/10 , High Risk
SK Hynix: 5.8/10 , Elevated Risk
Five of six core holdings scored 6.3 or above. Four scored above 7.0. This is not a portfolio with one or two elevated names and several stable anchors. This is a portfolio where the forensic risk profile was uniformly elevated across nearly every position.
What the Dimensions Revealed
Capital Structure: Portfolio Average 7.0/10
Four of six holdings scored 7 or above on Capital Structure. CoreWeave ended Q1 2026 with over $17.3 billion in debt, up from $4.9 billion a year earlier, reporting $536 million in interest expense against $2.08 billion in quarterly revenue while spending $7.7 billion in capital expenditures in a single quarter. By mid 2026, total debt exceeded $35 billion. Nebius guided to $20 to $25 billion in 2026 capital expenditures while remaining unprofitable. IREN and Core Scientific, both transitioning from crypto mining to AI hosting, required continuous capital for infrastructure build out.
The common pattern: every one of these businesses depended on continuous access to capital markets at favorable terms. Their business models were not merely capital intensive, the capital structure was the business model. Any disruption to funding access, whether from rising rates, credit tightening, or a sentiment shift, would create simultaneous stress across the entire portfolio.
Narrative Fragility: Portfolio Average 7.3/10
This is the most significant finding.
Every holding in the portfolio was dependent on the same macro narrative: AI requires massive physical infrastructure buildout. CoreWeave’s valuation assumed multi year hypergrowth in AI compute demand. Nebius’s $46 billion contract backlog with Meta and Microsoft assumed sustained hyperscaler outsourcing. Bloom Energy’s re rating from niche fuel cell company to AI power play assumed accelerating data center electricity demand. IREN and Core Scientific’s crypto to AI pivots assumed the AI infrastructure trade had durable momentum. SK Hynix’s premium valuation assumed sustained HBM pricing power.
Each of these narratives is individually reasonable. But assembled into a single portfolio, they create something different: a single bet expressed through six vehicles. When that narrative wobbled in July 2026, there was no uncorrelated position to absorb the shock. The diversification was an illusion, visible in sector labels but invisible in the underlying narrative dependency.
This is exactly the kind of hidden fragility that standard portfolio analytics cannot detect. A sector exposure report would show holdings across Technology, Industrials, Consumer, and Energy. A factor analysis would show growth and momentum tilts. Neither would reveal that every position shared the same narrative foundation, and that the narrative was the load bearing structure for the entire portfolio.
Insider Behavior: Portfolio Average 6.2/10
Multiple insider behavior signals were active across the portfolio. Bloom Energy insiders sold approximately $59.8 million in stock over a three month period, with director Jeffrey Immelt selling $7.17 million and COO Chitoori Satish selling 20,000 shares. IREN’s board approved over 18 million restricted stock units for its co founder co CEOs, a governance event that triggered an immediate stock decline and investor backlash. The broader crypto miner sector exhibited insider selling patterns during the AI pivot rally.
The pattern across the portfolio: informed parties inside multiple holdings were reducing exposure or capturing value during the period of peak narrative premium. No single transaction is conclusive, but the portfolio level pattern is a signal.
Governance Integrity: Portfolio Average 6.7/10
Nebius operates under a dual class share structure that concentrates approximately 52% voting power in a family trust while holding only roughly 11% economic interest. The company qualifies as a Nasdaq Controlled Company with reduced board independence requirements. It also disclosed material weaknesses in internal controls over fixed assets and revenue recognition as of year end 2025.
IREN’s 18 million RSU grants to co founders, defended publicly by the independent chair rather than modified in response to shareholder concerns, signal a governance culture that may prioritize founder alignment over broader shareholder interests.
Bloom Energy amended its certificate of incorporation to add officer exculpation provisions during a period when a short seller report was publicly challenging the company’s disclosures, a pattern of expanding legal protections while transparency is being questioned.
Core Scientific emerged from Chapter 11 bankruptcy in January 2024 and is now aggressively deploying capital, a pace that tests governance oversight capacity regardless of board quality.
Operational Signals: Portfolio Average 6.3/10
Customer concentration was the recurring operational theme, and the dependencies were not merely parallel but layered. CoreWeave depends on a small number of hyperscaler and AI model developer clients. Nebius expects a substantial portion of future revenue from Meta and Microsoft. IREN and Core Scientific depend on CoreWeave as a key customer. Bloom Energy depends on Oracle and Brookfield. SK Hynix’s top 5 customers likely represent over 70% of HBM revenue.
The portfolio’s operational risk was not diversified across different customer bases. It was layered through the same ecosystem, where a disruption at one node (say, a CoreWeave stress event) would cascade to multiple holdings simultaneously.
Regulatory Exposure: Portfolio Average 5.8/10
SK Hynix carried the highest regulatory exposure. The company operates 30% to 40% of its DRAM and NAND production in China, and US export controls revoked its validated end user status in September 2025. Annual licenses granted in 2026 provided short term relief but not the commercial certainty the business requires. South Korea’s broader dependency on Middle Eastern oil imports through the Strait of Hormuz introduced energy security risk that contributed to the Kospi losing roughly a third of its value during the July selloff.
Nebius carried elevated regulatory exposure through its Dutch incorporation, historical Yandex provenance, and multi jurisdictional data center operations.
The Core Finding: Correlated Dimension Risk
The most important output of this analysis is not any individual score. It is the pattern across the portfolio.
When five of six holdings simultaneously score High or Critical on Narrative Fragility, and four of six simultaneously score High or Critical on Capital Structure, the portfolio is not diversified in any meaningful forensic sense. It is a concentrated structural bet on a single thesis, financed by capital structures that depend on that thesis remaining intact, and governed by institutions that are either too young, too conflicted, or too founder aligned to provide effective checks under stress.
Traditional portfolio analytics would not have surfaced this. Sector labels, factor decompositions, and correlation matrices operate at a different resolution than the forensic signals embedded in filings, insider behavior, governance structures, and narrative dependency patterns.
This is the gap that forensic scoring fills.
Detection, Not Prediction
It is important to state clearly what this framework would and would not have done.
It would not have predicted the specific timing of the unwind, the exact catalyst, the magnitude of losses, or the forced liquidation event. It would not have assessed the fund’s leverage (4x gross) or portfolio construction decisions, which were the proximate causes of the catastrophic outcome.
What it would have done is surface the structural fragility that made the portfolio vulnerable to exactly the type of stress event that occurred. An allocator reviewing this scoring output would have seen a portfolio where the forensic risk was uniformly elevated, the narrative dependency was correlated, the capital structures were fragile, and the insider behavior patterns suggested informed parties were already reducing exposure.
That is not prediction. It is early detection of structural risk, delivered before it became visible to consensus, which is precisely what forensic intelligence is designed to do.
Implications for Allocators
The Situational Awareness LP unwind is not primarily a story about AI, leverage, or a 25 year old fund manager. It is a case study in what happens when concentrated structural fragility meets an adverse catalyst.
For allocators evaluating exposure to concentrated funds, the lesson is that traditional portfolio analytics, sector exposure, factor tilts, correlation, are necessary but insufficient. A forensic screening layer that scores individual holdings across capital structure, regulatory exposure, insider behavior, narrative fragility, operational signals, and governance integrity, and then identifies correlated patterns across the portfolio, reveals structural vulnerabilities that exist in a different analytical dimension.
The signals were in the filings. The signals were in the insider transactions. The signals were in the governance structures. They required a framework that integrates all of them simultaneously.
This analysis is for informational purposes only and does not constitute investment advice. Investors should conduct their own due diligence and consult with financial advisors before making investment decisions.