Weekly NVIDIA Circular Financing Risk Monitor | 2026-06-15

📋 Weekly Summary

No major structural changes this week, but several signals warrant close attention: NVIDIA Q1 FY2026 record $81.6B revenue confirms demand remains strong, but OpenAI Q1 2026 Non-GAAP operating margin of -122% leaked data intensifies circular financing concerns; CoreWeave debt-to-equity at 738% (MRQ) with first Vera Rubin systems online; Microsoft Maia 200 claims performance superiority over TPU/Trainium, escalating the custom silicon arms race.

1️⃣ NVIDIA Related-Party Transactions

Key Points:

• NVIDIA’s 2026 Annual Report (published May 12) explicitly discloses: “Except as discussed below, there were no transactions with related persons since the beginning of Fiscal 2026 or currently proposed transactions.” At face value, NVIDIA’s related-party disclosure appears clean.
• However, NVIDIA’s role as a financial player continues to expand: direct investment in CoreWeave ($2B, January 2026), investment in Nebius ($2B), and ongoing vendor financing structures — these fall outside the narrow definition of “related-party transactions” but are critical links in the circular financing chain.
• Bloomberg’s January feature “A Guide to the Circular Deals Underpinning the AI Boom” documented NVIDIA’s transformation from chip supplier to financial linchpin in 2024-2025.

Risk Rating: 🟡 Moderate-High — Disclosure compliant, but structural role unchanged.

2️⃣ OpenAI Financial Health

Key Points:

• The Information (May 22) leaked OpenAI Q1 financials: $5.7B revenue, Non-GAAP operating margin of -122% — meaning the company lost $1.22 for every $1 earned. This excludes stock-based compensation (SBC), so actual GAAP losses are likely larger.
• Growth stall signal: ChatGPT WAU dropped from 920M in February to a Q1 average of 905M, suggesting declines in January or March. OpenAI had projected breaking 1B WAU in 2025 — still unmet.
• Burn rate is staggering: Sacra projects ~$27B cash burn in 2026 and ~$63B in 2027. Even with $122B raised, the runway is limited at current loss rates.
• Conversion rate: 55M paying users / 905M WAU ≈ 6%, improved from 2.6% last year, but heavily diluted by cheap Go subscriptions ($5-8/month) and free annual Go cards given to Indian subscribers.
• OpenAI is one of CoreWeave’s two largest customers — its financial health directly impacts GPU cloud demand sustainability.

Risk Rating: 🔴 High — Revenue growth consumed by loss velocity, deepening circular financing dependency.

3️⃣ GPU Cloud Provider Health

Key Points:

• CoreWeave’s debt risk is this week’s brightest red flag: Morningstar (June 3) assigned “No Economic Moat, Very High Uncertainty” rating with fair value estimate of $106 (near current price). Debt/equity at 738%, annual interest ~$2B.
• Microsoft + Meta’s $122B+ commitments equal ~90% of AWS’s TTM revenue — but these are long-term contracts. Core risk: if AI demand slows, can hyperscalers execute as promised?
• Hyperscaler motivation analysis (IO Fund, June 11 deep dive): The core purpose of hyperscaler-neocloud partnerships is moving AI capex off the balance sheet to opex — not inherently problematic, but if neoclouds can’t profit, the chain becomes “offloading risk to weaker players.”
• NVIDIA as “lender of last resort”: $2B into CoreWeave + $2B into Nebius, plus GPU-backed debt structures — NVIDIA is essentially financing its own customers to buy its own products.
• Lambda Labs: No major financing/operational news this week.

Risk Rating: 🔴 High — Extreme leverage, profitability distant, dependent on NVIDIA financial support.

4️⃣ Circular Financing Risk Signals

Key Points:

• IO Fund (June 11) — this week — published deep analysis “Nvidia, CoreWeave, and Nebius: Inside the Circular Financing of the GPU Boom” — the most important new analysis this week.
• Core circular chain: NVIDIA invests in CoreWeave/Nebius → These companies borrow GPU-backed debt to buy NVIDIA GPUs → Provide compute to hyperscalers → Hyperscalers pay → NVIDIA recognizes revenue → NVIDIA reinvests.
• Key question: If AI demand slows (OpenAI’s growth stall is an early warning), can neocloud long-term contracts cover their massive debt interest? CoreWeave’s $2B annual interest vs. projected $12.6B revenue leaves thin coverage.
• Telecom bubble comparison: LinkedIn’s Tomasz Tunguz (October 2025) compared NVIDIA’s $110B vendor financing to the telecom bubble — “NVIDIA’s vendor financing strategy totals $110 billion in direct investments plus another $15 billion in financial backstops.”
• UBS view: NVIDIA accounts for ~60-70% of total AI capex; its investment behavior effectively creates its own demand.

Risk Rating: 🔴 High — Circular structure shows no signs of abating; this week’s IO Fund deep dive further confirms risk.

5️⃣ Competitor Custom Silicon Progress

Key Points:

• Microsoft Maia 200 (launched January 26) is this week’s most notable competitive development — Microsoft claims “the most performant first-party silicon from any hyperscaler,” with FP4 performance 3x Amazon Trainium 3 and FP8 performance exceeding Google TPU. This directly challenges NVIDIA’s dominance in inference.
• Google TPU v6e Trillium is in full production; Spheron’s April analysis shows competitive inference cost/BFLOP, but the CUDA ecosystem moat remains NVIDIA’s greatest advantage.
• Amazon Trainium 3 is claimed by Microsoft to lag 3x on FP4 — if accurate, AWS’s custom silicon strategy faces pressure.
• Meta + AMD: Meta signed a 6GW AMD partnership, signaling hyperscalers’ unwillingness to be fully locked into NVIDIA.
• Key judgment: Custom silicon won’t displace NVIDIA short-term (ecosystem barriers too deep), but will progressively compress pricing power, especially in inference.

Risk Rating: 🟡 Medium — Competition intensifying but no immediate threat.

🎯 Overall Risk Assessment

💡 Core Conclusions

1. OpenAI is the weakest link in the circular chain: -122% operating margin + growth stall = this company is entirely sustained by fundraising. $122B raised has limited runway at ~$27B/year burn. If OpenAI growth slows or fundraising difficulties arise, the entire NVIDIA → Neocloud → Hyperscaler chain will face pressure.

2. CoreWeave’s debt structure is alarming: 738% debt/equity, $2B/year interest, near-zero operating cash flow — this is a company using GPU-collateralized loans to buy more GPUs. First Vera Rubin systems online is good news, but profitability remains far away.

3. NVIDIA’s dual role as “supplier + banker” amplifies systemic risk: Invest in customers → Customers buy products → Recognize revenue → Reinvest — this cycle runs perfectly in an upswing, but once AI capex growth slows, NVIDIA will face simultaneous revenue decline and investment write-downs.

4. Microsoft Maia 200 is the biggest long-term threat: Not because it can replace NVIDIA now, but because it represents hyperscalers’ serious intent to develop custom silicon. Meta + AMD’s 6GW contract points in the same direction.

5. No “black swan” events this week, but risk is cumulative — each additional round of circular financing increases the system’s fragility.

Data as of 2026-06-15 | Next Monitor: 2026-06-22