Nofinity Logo

Latest Analyses(7)

Wall Street is HIDING the AI Debt Bomb (2008 Again)
3 min read

Wall Street is HIDING the AI Debt Bomb (2008 Again)

The Core Issue: GPU Value Plummets vs. Long-Term Debt
  • GPUs lose value rapidly: An Nvidia H100 chip rented for ~$8/hour in early 2024 but fell to $2–3 by late 2025 – a 50–70% collapse in under two years. After three years, resale value drops to roughly half the original price.
  • Debt runs for decades: Financing for AI infrastructure has maturities of 5, 10, 15 years – one vehicle for Meta‘s Hyperion campus doesn‘t mature until 2049.
How Wall Street Hides the Debt
  • Off-balance-sheet financing: Companies like Meta, Oracle, and Anthropic use special purpose vehicles (SPVs) to keep debt off their books. Example: Big Net Investors LLC (Meta) – $27 billion in debt secured by lease payments but not on Meta’s balance sheet.
  • Private credit explosion: From virtually nothing to over $200 billion in outstanding loans to AI companies. The top five hyperscalers issued $121 billion in corporate bonds in 2025 alone – four times the prior annual average.
  • Securitization like 2006: Structured finance products (e.g., asset-backed securities) are issued where credit ratings depend on the paying tenant, not the underlying asset (GPU). Moody‘s downgraded a CoreWeave facility because the end customer was weaker.
The Most Leveraged Position: Bitcoin Miners
  • Bitcoin miners transform into AI landlords: They sign long-term leases (12–20 years) to survive, as mining costs ($75,000–$88,000 per Bitcoin) often exceed the price. They have disclosed over $70 billion in AI contracts and sold thousands of Bitcoins to fund the shift.
  • Dramatic leverage: Miners finance their conversion ($8–15 million per megawatt) with high-yield debt secured by GPUs that rapidly depreciate. Core Scientific posted a $347 million net loss in Q1 2026 despite growing AI revenue.
  • Chain reaction risk: If GPU values fall faster than debt can be refinanced, losses hit the miners first – and they are publicly traded, directly in portfolios.
Warning Signs and Parallels to 2008
  • Credit default swaps (CDS) spike: CoreWeave‘s CDS implied a 50% probability of default within five years in July 2026, while some bond tranches still carried investment-grade ratings. Michael Burry warned of “parabolic” CDS on Nvidia.
  • Moody‘s raises alarm: AI spending threatens credit quality at Microsoft, Amazon, Alphabet, Meta, Oracle, and CoreWeave. These six firms carry $460 billion in direct debt and $1.2 trillion in lease commitments.
  • US senators call for regulation: In January 2026, four senators asked regulators to examine opaque debt markets.
Bottom Line: No Sudden Bust, but a Gradual Shift
  • The structure ensures losses don‘t hit the big hyperscalers but trickle down to the weakest links – the miners and their investors.
  • The AI bubble doesn‘t need to burst; it‘s enough if hardware ages faster than debt is repaid. Then it‘s “look out below.”
Mercadolibre MELI - Subprime 62% Interest Credit Expansion Makes 40% Growth UNSUSTAINABLE
2 min read

Mercadolibre MELI - Subprime 62% Interest Credit Expansion Makes 40% Growth UNSUSTAINABLE

Business Model and Growth Engine
  • 146 million buyers across 18 Latin American countries
  • Integrated ecosystem of marketplace, logistics, advertising, and fintech (Mercado Pago)
  • Revenue growth of 48% per year over the past six years, driven by credit expansion
The Credit Business – Core Risk and Opportunity
  • 62% interest rates on credit cards, targeting subprime and underbanked customers
  • High default rates: 7% NPLs (15-90 days), 18% after 90 days
  • Loan loss provisions of around 20% of each credit; gross margin (NIMAL) declining to 20%
  • Regulatory risk: Brazilian interest cap at 100% of loan principal
Growth Sustainability
  • Current price-to-earnings ratio of 40 with 40% growth – cheap but unsustainable
  • Analyst reduces long-term assumption to 20% or even 12%
  • A credit slowdown would also dampen e-commerce and advertising
  • 82% of Brazilian households in debt, 29% of income for debt service
  • Risk of a credit cycle crash with potential government intervention
Valuation and Conclusion
  • After adjustment, intrinsic value is below 1,000
  • Currently too risky from a value perspective; reevaluation at a price of 1,000
  • Main risk factor: credit quality and expansion
Context
  • Michael Burry holds the largest position but is seen as a gambler rather than a value investor
  • Compared to Amazon, MercadoLibre barely dilutes shareholders
How a Base Token Could Help Decentralize Coinbase's Blockchain
2 min read

How a Base Token Could Help Decentralize Coinbase's Blockchain

Focus on Decentralization and Token Strategy
  • Background: Over a year ago, Coinbase announced the exploration of a Base token to further decentralize its blockchain.
  • Philosophy: Initially, they deliberately avoided introducing a token to prevent distraction from noise and incentives. The goal was to achieve product-market fit purely from a product perspective, without mixing in economic incentives.
  • Progress: As Base matured (growing in rank, enabling trading and payments), a token was seen as a powerful lever for decentralization, aligning participants, and driving technological innovation.
  • Current Stance: Coinbase continues to explore the token launch but remains unhurried. The priority is to accelerate growth without derailing product goals via premature decisions.
  • Outlook: A token remains a future possibility, but no further details or timelines are available.
Stocks are crashing! When will it stop? THIS is what matters now // BRIEFING
2 min read

Stocks are crashing! When will it stop? THIS is what matters now // BRIEFING

Market Driven by Sentiment: Panic vs. Fundamentals
  • The market is currently ruled by panic and psychology, not fundamentals. Recency Bias and Negativity Bias distort reality.
  • Bears vs. Bulls: While bears celebrate the end of the AI bubble, bulls point to rock-solid fundamentals (Nvidia growth +103%) and missed gains from panic selling.
AI Sector Under Fire
  • Anthropic IPO: Successful launch, then heavy selling. A sentiment killer, but not the end of the story.
  • Nvidia: Despite record revenue and beating all estimates, the stock falls. Fear of rate hikes and new regulations outweigh the stellar earnings.
  • Everything is dropping simultaneously (Stocks, Gold, Crypto) – a clear sign of a liquidity-driven, sentiment-based selloff.
The Key Variable: The Fed & Interest Rates
  • Markets are pricing in an interest rate shock of up to 100 basis points.
  • Hope: Inflation is cooling. Goldman Sachs forecasts point towards a soft landing.
  • Risk Scenario: Rate hikes combined with an AI cooldown = recession. The analyst doesn't see this as the base case yet.
Actionable Advice: Hold Cash & Watch for Opportunities
  • Keep your powder dry. Mario himself holds roughly 20% cash, parked at XTB earning 4.5% interest.
  • Single Stocks: Despite the crash, stocks like SK Hynix are still massively up on a yearly basis. SpaceX is touted as a long-term high-risk, high-reward independence play.
Bitcoin: The Lower Low
2 min read

Bitcoin: The Lower Low

Bitcoin: The Lower Low – A Market Analysis

The speaker analyzes Bitcoin's recent drop to around $55,000, placing it in historical context. He emphasizes this is a typical pattern for Q4 of a midterm year, similar to 2014, 2018, and 2022.

Key Takeaways:

  • Realized Price as Key Level: Bitcoin is approaching the Realized Price, which has been a floor in every bear market. Breaking below it would be historically normal and could set the stage for the next rally.
  • Parallels to 2018: The current move mirrors 2018 – both in timing and in the context of falling stock markets (NASDAQ ~20% down). Bitcoin behaves as a risk asset, not a safe haven.
  • Liquidity Risk: A deep stock market correction could ease liquidity conditions, ultimately helping Bitcoin. This happened in 2018 when expected rate hikes were abandoned.
  • Silver as Leading Indicator: Silver tends to bottom before Bitcoin. Currently, silver is still falling, so caution is warranted. Bitcoin could find its bottom within the next 7 weeks.
  • On-Chain Metrics: The On-Chain Risk Score is at 0.149, close to the historical low below 0.1. The Balance Price (~$37,300) is a possible but not guaranteed lower bound.
  • Strategy: The speaker recommends a DCA approach during the second half of the midterm year and advises against panic selling. Historically, a capitulation leads to a quick recovery.

Conclusion:

What is happening is 'par for the course' for Bitcoin. While it feels bad now, it is actually positive for 2027. Investors should have a plan and avoid selling at the bottom.

How Base Wants to Bring Tokenized Stocks With Voting Rights to the US
2 min read

How Base Wants to Bring Tokenized Stocks With Voting Rights to the US

Background and Regulatory Progress
  • The speaker praises the SEC and CFTC for their forward-thinking and collaborative approach, which now allows for innovation in U.S. capital markets.
  • An innovation exemption from the SEC enables the offering of tokenized securities within the U.S.
Tokenized Stocks: Features and Benefits
  • These tokenized stocks are fully backed, direct equity (not debt), and include voting rights and dividends.
  • Benefits: global, 24/7 access at low cost, transparent shareholder records, and new capabilities like stock bonuses.
Coinbase's Strategy and Market Approach
  • The first version of tokenized stocks was launched outside the U.S. Coinbase now plans to bring it to the American market.
  • The B20 version (permissionless DeFi variant) already lists over 250 assets and offers deep liquidity.
  • Discussions with potential issuers are ongoing; Coinbase emphasizes education and transparency to overcome initial hesitancy.
Coinbase's Philosophy
  • A measured, steady pace to build trust.
  • Let the technology speak for itself through honest advice and clear communication.
  • Ultimately, leveraging blockchain's power to democratize access to capital markets globally.

Note: This conversation is an excerpt from the podcast 'Unchained' hosted by Laura Shin.

How North Korea Is Really Using AI to Hack
2 min read

How North Korea Is Really Using AI to Hack

North Korea's Use of AI in Cyber Operations: Between Mundanity and Future Danger

In this episode, the speakers discuss how North Korea (DPRK) is using Artificial Intelligence (AI) for its cyber operations. The findings are surprisingly understated.

Current Use: Pragmatic but Unspectacular

  • Social Engineering: AI is primarily used to refine social engineering scripts – making phishing and scam messages more credible.
  • Code Optimization: Previously, hackers had to program their frontends, backends, and command-and-control servers (C2 servers) themselves. Now, AI helps improve these workflows.
  • Deepfake Videos: The quality of these videos is worse than older, manually created versions – no breakthrough so far.

Disappointment vs. Expectations

  • The speakers are disappointed because AI does not autonomously and spectacularly hack, as often feared.
  • Instead, it is about „tiny micro-aggressions“ – small but effective optimizations of existing methods.
  • This could cause observers to underestimate the danger, thinking: „Everyone is just exaggerating.“

Future Evolution: The „AI-First“ Approach

  • So far, North Korean hackers only optimize existing campaigns with AI.
  • The decisive shift will occur when a new generation of hackers (or another country) starts from scratch.
  • They might build an „AI-first“ campaign – an attack fundamentally based on AI technologies, rather than just integrating them later.
  • This shift is 100% certain, especially for groups that genuinely need AI to increase their efficiency.

Key Message: North Korea currently uses AI in a pragmatic and unspectacular manner – but this underestimates the future danger. If cybercriminals develop from the ground up with AI, the threat landscape could change dramatically.

New Episode: Every Wednesday at 4:00 PM Eastern Time.