The $7.5 Trillion Question: Credit Markets Are Already Pricing AI's First Bankruptcy

The credit default swap market is a brutal machine. It strips away narrative, ignores the keynote speeches, and reduces a company's entire future to a single number that traders can price in real time. On Monday, that number flashed a warning that equity markets have refused to acknowledge for eighteen months. Nvidia's five-year CDS hit 82 basis points. A record. Fourteen basis points of that came in a single day, the largest jump since the contract began trading in November 2025. Let me translate that for you. The company with the world's most advanced AI chips, the one with eye-watering gross margins and a product backlog that would make most manufacturers weep, just saw the insurance premium on its debt spike harder than at any point in its short credit derivative history. And the market barely blinked. I have spent my career hunting for the story that defines the next cycle. I sat through the 2021 NFT mania with a cryptography background, watching sentiment decouple from intrinsic value in real time. I published the whitepaper deconstructing TerraUSD's incentive misalignment within 48 hours of the collapse, and I watched credit signals—not equity prices—tell the true story two weeks before public markets capitulated. The pattern is always the same. Equity markets trade on stories. Credit markets trade on math. When those two diverge, credit is the one to trust. Right now, they are diverging at a magnitude that should worry every institutional allocator in this space. What we are witnessing is not merely a repricing of AI companies. It is the first genuine balance sheet war in the history of the technology sector. The AI buildout—the data centers, the GPU purchases, the power infrastructure, the cooling systems—has been financed through a combination of direct debt, lease commitments, and off-balance-sheet guarantees that now total more than one and a half trillion dollars. And the credit market is starting to ask the question that equity investors have been too intoxicated by the narrative to raise: who is going to pay this back? THE CREDIT SIGNAL: WHAT THE DATA ACTUALLY SHOWS Let me be precise about what we are seeing in the derivatives market. A credit default swap is, at its core, an insurance contract. The buyer pays a premium, and in exchange, the seller agrees to cover the buyer's losses if the underlying company defaults on its debt. The premium—quoted in basis points—represents the market's collective assessment of default risk. At 82 basis points, Nvidia's five-year CDS implies roughly a 4 percent probability of default over the next half decade. That is low in absolute terms. But the trajectory is the story. Fourteen basis points in a single session, the largest move since the contract launched, tells us something profound about how market participants are reconceiving Nvidia's risk profile. The market is no longer treating Nvidia as a pure semiconductor company. It is beginning to price Nvidia as a lender. A guarantor. An institution whose fortunes are inextricably linked to the creditworthiness of its customers. And that is not a position the market is comfortable with. Consider the Bloomberg reporting from the same week: Nvidia is preparing artificial intelligence commitments that could exceed 750 billion dollars. These are not purchase orders. They are structural commitments—partnerships, financing arrangements, and guarantee agreements that fundamentally alter the company's risk architecture. The SK Group partnership alone accounts for more than 500 billion dollars of that figure. And then there is the OpenAI arrangement: negotiations for Nvidia to guarantee up to 250 billion dollars of compute lease obligations for the leading AI lab. Let me repeat that number because it deserves emphasis. Two hundred and fifty billion dollars. That is not a line item. That is a balance sheet transformation. Think about what this means structurally. Nvidia is the world's leading GPU manufacturer. The AI industry built its entire computational foundation on Nvidia silicon. Now the company is being asked to become the financial guarantor of its own demand. It is the most elegant example of self-reinforcing market power I have seen in two decades of observing this industry. Nvidia guarantees the leases, OpenAI buys the compute, OpenAI builds the models, and the models require more Nvidia GPUs. Demand is secured. Revenue is secured. But the risk—the ultimate risk of whether those models generate enough economic value to justify the cost of the compute—has been transferred from the customer onto Nvidia's balance sheet. Or more precisely, onto Nvidia's off-balance-sheet commitments. This is where the CDS market comes into focus. The credit market is not naive. It sees the 750-billion-dollar commitment and immediately understands what it means: Nvidia's creditworthiness is now a composite function of chip demand plus customer credit risk. If OpenAI's revenue fails to materialize—if the models don't get deployed, if the adoption curve flattens, if the price of inference compute collapses—Nvidia is on the hook. The credit market has started pricing that risk. The equity market has not. I find myself thinking back to the 2022 Terra collapse. There, too, the signal appeared early. The credit market began pricing UST's default risk at levels that implied systemic failure three weeks before the actual death spiral. But nobody wanted to hear it because the narrative was too good. The algorithmic stablecoin was going to revolutionize finance. The same blindness is now visible in AI equities. The narrative is so seductive—artificial superintelligence, autonomous agents, the fourth industrial revolution—that investors are comfortable ignoring the structural fragility of the very companies building the infrastructure. THE ORACLE WARNING AND THE INVESTMENT-GRADE CLIFF Take Oracle. Please. I say that with respect because Oracle's pivot to AI cloud infrastructure under Larry Ellison's leadership has been nothing short of aggressive. But the credit market has delivered its verdict. Oracle's five-year CDS has risen to over 215 basis points, up from roughly 145 basis points at the end of last year. That is a 48 percent increase in the market's assessment of Oracle's default risk in a matter of months. When I worked with legal experts in Singapore and Vancouver on the 2025 regulatory compliance initiative, we spent considerable time analyzing how credit ratings interact with institutional investor mandates. That experience taught me something crucial: a downgrade to BBB- is not merely an opinion. It is a structural event that triggers forced selling across a vast ecosystem of pension funds, insurance companies, and fixed-income mutual funds that are not permitted to hold junk-rated debt. Oracle's 30-year bond due in 2035 is now trading at a spread of 263 basis points over comparable Treasuries. For context, that is the kind of spread you normally see in a company with one foot in the distressed bucket. The S&P downgrade to BBB- places Oracle exactly one notch above junk. If the rating agencies deliver one more downgrade—if the debt-to-EBITDA ratio deteriorates further, if the AI cloud segment fails to achieve the margin expansion Oracle has projected—Oracle will cross the line into high-yield territory. The name will be removed from billions of dollars of investment-grade indices. The forced selling will be measured in tens of billions. And the company's cost of capital, already elevated due to its aggressive acquisition strategy, will jump another 150 to 200 basis points almost overnight. I am not predicting that outcome. I am saying the credit market is assigning it a probability that the equity market has not yet internalized. When I speak with institutional investors, they are still debating Oracle's AI cloud revenue growth. The equity analysts are still modeling 15 to 20 percent forward growth. But the CDS market has moved on. It is already trading Oracle like a company at the edge of the investment-grade cliff, a company whose bond spreads are widening faster than any of its peers. That divergence—between the revenue story and the credit story—is precisely the kind of signal that I have learned to respect. During my work modeling institutional inflow scenarios ahead of the 2024 Bitcoin ETF approvals, I collaborated with a team of four analysts to project liquidity impacts. We concluded that ETF approvals would trigger volatility compression rather than parabolic growth, a forecast that ran against the consensus at the time. The Bloomberg Terminal data feeds cited our report, and the prediction proved accurate. The lesson I took from that experience was that institutional narratives are driven by regulatory clarity and liquidity mechanics, not just technological innovation. The same principle applies here. Oracle's AI cloud story matters, but what matters more is the regulatory framework that forces institutions to sell when a rating gets cut. That is mechanical. That is structural. That is the kind of force that moves markets in ways that fundamental analysis cannot fully capture. COREWEAVE AND THE FIFTY PERCENT QUESTION Now we arrive at the epicenter of the AI credit storm. CoreWeave. The AI cloud provider that went public at the peak of the AI infrastructure frenzy. The company that raised billions of dollars of debt to finance GPU purchases, signing multi-year contracts with hyperscalers and AI labs alike. CoreWeave's five-year CDS is trading above 855 basis points. Standard CDS pricing models translate that number into an approximately 50 percent probability of default over the next five years. Let me pause here. This is a company whose equity was trading at a multi-billion-dollar valuation just months ago. A company with revenue growth that any CFO would envy. A company that is so deeply embedded in the AI compute ecosystem that its collapse would send shockwaves through every layer of the infrastructure stack. The credit market is telling us this company has a coin-flip chance of surviving the next half-decade. I have seen this pattern before. In the crypto lending sector, we watched companies with fantastically growing revenue—companies like Celsius and BlockFi—maintain triple-digit revenue growth quarter after quarter right up to the moment they collapsed. The revenue was real. The problem was that the revenue was dependent on an asset base whose value was highly correlated with a single factor. For the crypto lenders, that factor was the price of Bitcoin and Ethereum. For CoreWeave, that factor is the utilization rate of AI compute and the willingness of OpenAI, Microsoft, and other customers to continue paying premium prices for GPU rental. The key structural weakness is leverage. CoreWeave is not a software company with high margins and minimal capital intensity. It is a capital-intensive leasing business masquerading as a technology company. Every GPU it owns has been financed, either through term loans, sale-leaseback arrangements, or in some cases, securitization vehicles that bundle compute contracts into tradable debt instruments. The economics work beautifully as long as the utilization rate stays high and the rental prices stay firm. The moment those two variables deteriorate—the moment an unutilized rack sits idle or a customer renegotiates its contract—the asset base starts to erode, and with it, the collateral backing billions of dollars of debt. This is what the CDS market is pricing. Not a specific event, not a particular quarter, but a structural feature of the business model. CoreWeave is a leveraged bet on the continued exponential growth of AI compute demand. If that demand materializes, the company thrives. If it grows at a merely linear pace, the debt service burden becomes intolerable. The credit market is assigning roughly equal probabilities to both outcomes. That is the definition of a coin flip. And remember: 2025 was supposed to be the year when the capital markets finally understood the AI infrastructure opportunity. Instead, they are treating CoreWeave's survival as a matter of probability. THE MOODY'S SIX AND THE 1.66 TRILLION DOLLAR OBLIGATION The systemic nature of this shift becomes clearer when you examine Moody's recent warning. The rating agency took the unusual step of publicly flagging six companies whose credit quality it believes is directly threatened by unprecedented AI spending: Microsoft, Amazon, Alphabet, Meta, Oracle, and CoreWeave. Not a single one of these companies escapes the warning. Not the hyperscalers with their trillion-dollar valuations. Not the diversified technology conglomerates with decades of cash-flow history. Not the AI cloud upstarts. The entire AI ecosystem, from the largest incumbent to the newest entrant. The numbers behind this warning are staggering. These six companies hold approximately 460 billion dollars in direct debt. But that direct debt is only the beginning. When you add in the lease commitments—for land, for office space, for data centers, for power infrastructure—the total obligation balloons to 1.2 trillion dollars. Combined, they represent 1.66 trillion dollars of liabilities, almost all of it tied to the AI infrastructure buildout. This is not a speculative number. These are contractual obligations. Someone is going to have to pay this money back. The phrase that keeps coming through my mind is "balance sheet war." AI competition is no longer purely technological. It has become a contest of financial endurance, a test of which companies can absorb the highest debt service burden while simultaneously funding the most aggressive capital expenditure program. The hyperscalers have an advantage because they can cross-subsidize. Microsoft can fund AI data centers from Azure profits and Office 365 subscription revenue. Amazon can draw on AWS's operational cash flow. Alphabet can redirect advertising dollars. But there is a limit to how much cross-subsidization works, and Alphabet just demonstrated that limit in the most concrete terms possible. Alphabet reported its first quarter of negative free cash flow since its 2004 IPO. Not since the dot-com era. Not since the financial crisis. Since 2004. The company that generates more cash from online advertising than any entity in history is now spending more money than it generates. That is a transformation of the company's financial character. It is a signal that the AI buildout is consuming cash at rates that even the most profitable companies in history cannot sustain without tapping reserves. Alphabet's CDS has risen to 67 basis points. That is not alarming in absolute terms. But the direction of travel matters. When the most financially stable technology company in the world starts seeing its credit spreads widen, when the company with a cash hoard of more than 100 billion dollars starts spending down its free cash flow, the credit market reads it as a structural change. Alphabet is no longer a cash cow. It is now a capital-intensive AI contender with a cash-flow problem. The bond market has noticed. THE 600 PERCENT SURGE IN CREDIT HEDGING Perhaps the most telling data point in the entire story is the volume surge in the credit derivatives market. According to JPMorgan and DTCC data, CDS trading activity related to AI companies and technology stocks reached approximately 650 million dollars in notional volume in the second quarter of last year. That represents a year-over-year increase of nearly 600 percent. Six hundred percent. The market is not watching the AI credit story from the sidelines. It is piling in with unprecedented urgency. I want to parse this number carefully because it tells us two distinct stories. The first story is about hedging. Institutions that hold AI equities or AI-related debt want to protect themselves against downside scenarios. They buy CDS protection as insurance. The second story is about speculation. A growing cohort of traders is buying CDS protection not because they hold the underlying assets, but because they believe the underlying credit will deteriorate. Both motives are legitimate, but they have different implications for pricing. When hedging demand dominates, CDS spreads can widen even when fundamentals are stable, simply because there is a shortage of sellers. When speculative demand dominates, CDS spreads become a directional bet that can overshoot in either direction. The 600 percent volume increase tells me that both dynamics are at play. There is genuine hedging demand from institutions that own AI exposure and want downside protection. And there is a growing cohort of speculators—Michael Burry fans, credit market contrarians, distressed debt investors—who are building positions that will pay off handsomely if the AI credit cycle turns. Michael Burry's entry into this narrative deserves attention. The investor who shorted the subprime mortgage market in 2008 has accused AI companies of "round-tripping"—the practice of companies purchasing each other's services to inflate reported revenues for the purpose of securing financing. If this accusation is accurate, it would mean that some AI companies are not merely facing a credit cycle, but a fundamental revenue quality problem. The reports would not reflect genuine end-user demand. They would reflect an elaborate circular procurement system designed to manufacture financial credibility. Now, let me be clear about what I think of Burry's accusation. He is a brilliant investor and a contrarian thinker whose track record deserves respect. But the "round-tripping" claim, as stated, lacks specific evidence. He has not named the companies. He has not provided transaction details. He has not quantified the scale of the problem. What he has done is introduce a narrative that casts doubt on the quality of AI revenue across the sector. And in a market environment where credit spreads are widening and the equity market is showing fatigue, that narrative gains traction regardless of its evidentiary basis. This is the kind of situation where my pre-mortem analytical orientation kicks in. The pre-mortem framework, adapted from cognitive psychology and organizational behavior, asks us to assume that a failure has occurred and then work backward to identify the causes. Applied to the current AI credit situation, the pre-mortem approach yields something like this: it is early 2028, and CoreWeave has defaulted on its debt obligations. The default triggered cross-default clauses in related credit agreements, causing a cascade of margin calls across the AI cloud sector. Oracle has been downgraded to junk status, triggering forced selling from investment-grade-only mandates. Nvidia has been forced to honor 250 billion dollars in guarantees on OpenAI compute leases after OpenAI's restructuring. The equity markets have corrected 30 to 40 percent from peak. The question is: what warning signs existed in 2026 that we should have paid attention to? The answer is this CDS data. The widening spreads. The record volume. The one specific investment signal that appeared months—perhaps years—before the equity market embedded it into prices. THE CONTRARIAN CASE: WHY THE CREDIT MARKET MIGHT BE WRONG Let me now make the case against my own thesis. Credit markets are not omniscient. They are susceptible to fads, to liquidity distortions, to crowded trades, and to the same herding instincts that afflict equity markets. In fact, several specific factors suggest that the current CDS pricing may be overstated. First, liquidity. The CDS market for these AI names is relatively thin. Nvidia only began trading credit derivatives in November 2025. Oracle, CoreWeave, and the others have actively traded CDS for years, but at volumes far below the major banks and industrial conglomerates. In a thin market, a relatively small volume of directional trades can move spreads significantly. The 14-basis-point jump in Nvidia's CDS might reflect a handful of large institutional trades rather than a broad-based reassessment. If that is the case, the spread could mean-revert quickly once the marginal buyer or seller is satisfied. Second, the purpose of the hedging flow. A substantial portion of the CDS buying could be coming from equity investors who are using credit protection as a portfolio hedge rather than making a direct judgment on the issuer's creditworthiness. A large hedge fund holding Nvidia equity may buy CDS protection to neutralize idiosyncratic risk while maintaining its equity exposure for earnings upside. This creates an artificial bid for protection that widens spreads without any fundamental deterioration. The 600 percent volume increase includes a meaningful amount of this motive. Third, the regulatory and structural context. Credit markets are heavily influenced by the regulatory framework, and the current macro environment could tighten or loosen credit conditions in ways that are not yet reflected in the data. If the Federal Reserve cuts rates more aggressively than the market currently expects, refinancing becomes cheaper, maturities become easier to roll, and default probabilities decline. The CDS market is pricing a particular path for interest rates. If that path is wrong, the spreads will adjust accordingly. Fourth, the cash and liquidity buffers that the AI companies maintain. The Moody's warning focuses on debt and lease obligations, but it does not fully account for the cash holdings of these companies. Microsoft, Alphabet, and Amazon collectively hold several hundred billion dollars in cash and marketable securities. Even in a severe downturn, they have significant capacity to absorb losses before they would face genuine solvency concerns. CoreWeave and Oracle are more exposed, but even there, the companies have undrawn revolving credit facilities and relationships with banks that could provide emergency liquidity. And fifth, the simple possibility that the AI buildout succeeds. If OpenAI, Anthropic, and Google DeepMind deliver on their product roadmaps, if AI adoption reaches the levels that the most optimistic forecasters predict, then the compute demand will materialize, the data centers will be fully utilized, the revenue will flow, and the debt will be serviced. In that scenario, the current CDS pricing becomes a historical artifact—a moment when the market overreacted to temporary uncertainty. The 50 percent default probability assigned to CoreWeave could collapse toward single digits. The honest position is that both scenarios are possible. The CDS market is a probability-weighted assessment of outcomes, not a prediction of a specific future. It is a warning signal, not a prophecy. The mistake would be to treat it as either one. The correct response is to take the signal seriously and to structure portfolios accordingly. THE STRUCTURAL INTERSECTION OF CREDIT AND COMPUTE I want to return to a theme that sits at the center of my analysis: the entanglement of the AI compute cycle with the credit cycle. This is not a temporary phenomenon. It is a structural feature of how modern AI infrastructure is being built. The 460 billion dollars in direct debt and the 1.2 trillion dollars in lease and contractual commitments represent a fundamental shift in how technology infrastructure is financed. Earlier waves of technology investment—the dot-com era, the mobile era, the cloud era—were primarily equity-financed. Companies raised venture capital or issued public equity to fund growth. The AI era is different. It is being financed with debt. This has profound implications. Debt creates obligation. Equity creates optionality. A company financed with equity can pivot, can accept lower returns, can sacrifice profitability for growth without facing an existential crisis. A company financed with debt has mandatory payments that must be made on schedule. If a debt-financed company's business model underperforms, the failure mode is not reduced growth but insolvency. This is a different risk profile, and the credit market is the first to recognize it. I was in this position in 2022. I had built a reputation in that post-mortem period after the Terra collapse, and I watched the entire crypto lending sector unwind because of the same structural feature: leverage. The companies that failed—Celsius, BlockFi, Voyager—were not failing because their products were worthless. Celsius had real lending products. BlockFi had real customer demand. The problem was that their capital structures could not withstand a drawdown in the underlying asset price. They had borrowed short-term and lent long-term, and when the volatility hit, they could not meet their obligations. The AI infrastructure sector is running a similar play. The companies have made enormous fixed commitments. They have signed multi-year contracts that obligate them to spend billions of dollars regardless of output. They have borrowed through instruments that cannot be easily renegotiated. If the demand trajectory falters, the debt service becomes a death sentence. The comparison to the Terra collapse is not perfect. Terra's failure involved a fundamentally broken mechanism—an algorithmic stablecoin whose economic incentives were mathematically incoherent. The AI infrastructure buildout, by contrast, is building real assets with real utility. Data centers produce compute. GPUs execute models. Models generate outputs that have genuine economic value. The question is whether that value will be sufficient to service the debt that was used to build the infrastructure. It is a question of magnitude and timing, not of intrinsic worth. WHERE THE OPPORTUNITY SITS Now let me pivot from risk to opportunity, because a complete analysis requires acknowledging both sides of the ledger. In my 2024 Institutional Squeeze report, I argued that verified institutional flows would create volatility compression rather than immediate price appreciation. The market consensus was for a parabolic post-ETF move, and my analysis flagged the counter-intuitive outcome. I learned something valuable from that experience: the easiest returns come from being on the right side of a structural repricing, not from following the emotional narrative. The structural repricing happening now is the credit market's reassessment of AI risk. For investors who understand this dynamic, several opportunities emerge. The first is the opportunity to identify quality assets that have been marked down due to sector-wide concerns. The differentiation between AI credit stories is stark. Microsoft's CDS spread, while elevated, remains far below CoreWeave's. Alphabet's 67 basis points is a far cry from Oracle's 215. If the market eventually distinguishes between companies with genuine cash generation capacity and companies with leveraged balance sheets, the high-quality names will regain their relative attractiveness. In the cryptocurrency ecosystem, we watch this play out all the time. Bitcoin trades differently than an unknown altcoin. Institutions go to quality in times of stress, and the same dynamic will play out in AI equities and credit. The second opportunity is in direct credit instruments. The bond market is offering yields that have not been available in technology debt for a decade. Oracle's 30-year bond at 263 basis points over Treasuries provides an income stream that would have been unthinkable at the beginning of this cycle. Investors who believe the credit market is overstating the risk—who believe that Oracle will not default on its obligations—can now collect a significant premium for taking on a risk that might be substantially lower than the market price implies. This is the classic distressed-credit opportunity set, and it is now available in the technology sector. The third is the potential for sector consolidation. When debt markets tighten, the weakest players face existential pressure. CoreWeave, if its cost of capital remains elevated, may be forced to sell assets or merge with a stronger partner. Data center providers with leverage may find themselves acquired by entities with stronger balance sheets. This will not destroy value across the AI ecosystem; it will redistribute it. The companies with cash and credit capacity will acquire compute assets at depressed valuations, and the consolidation will ultimately produce a healthier sector. In the crypto space, we saw the same dynamic after the 2022 credit crisis. The stronger exchanges and platforms acquired the weaker ones, and the sector emerged with a more concentrated but more stable structure. The fourth opportunity is the convergence of crypto and AI infrastructure. Decentralized compute protocols—Render, Akash, and others—are building markets where idle GPU capacity can be shared and monetized. The financial distress among centralized AI cloud providers could accelerate the adoption of these decentralized alternatives. If a company like CoreWeave is forced to raise prices or ration capacity due to its debt burden, AI startups might migrate toward decentralized compute markets where the marginal cost per teraflop is lower. This is not a direct credit play, but it is an AI infrastructure play that responds to the same macro forces. THE REGULATORY DIMENSION I cannot conclude this analysis without addressing the regulatory dimension, because the enforcement actions and policy decisions of the next few years will substantially influence the outcome of this credit cycle. The regulatory moat concept—the idea that every economic actor must contend with legal and compliance barriers—applies to the AI sector in ways that few analysts have fully recognized. The "round-tripping" accusation, if it attracts regulatory attention, could change the information environment for AI credit risk. If the SEC begins investigating revenue recognition practices among AI infrastructure companies, the disclosures could reveal that some reported revenue is less robust than it appears. That would trigger a repricing of both credit and equity, likely to the downside. My experience with the 2025 Regulatory Compliance Initiative taught me that the SEC and its counterparts in other jurisdictions are actively examining the private financing structures that have emerged around AI and Web3. The off-balance-sheet entities, the special purpose vehicles, the complex leasing arrangements—these are not outside the regulatory perimeter. They are squarely within it. Alternatively, regulation could provide a backstop. If policymakers decide that AI compute is a strategic infrastructure asset, they might support the sector through favorable lending programs, loan guarantees, or antitrust exemptions that enable cooperative cost-sharing. The CHIPS Act of 2022 set a precedent for government support of semiconductor manufacturing. A similar approach for AI compute infrastructure would not eliminate credit risk, but it would reduce the probability of systemic failure. The outcome will depend on the political direction of the jurisdictions involved. Washington, Brussels, Beijing, and Singapore all have different approaches to AI policy. The credit market is pricing a world in which policy does not intervene. If it does, the current CDS spreads will have been an overreaction. If it does not, the spreads will have been a leading indicator. THE NARRATIVE THAT FOLLOWS THE CREDIT CYCLE The story of AI has been told for two years now as an unambiguous growth story. The narrative is familiar: AI models are advancing at exponential rates, the demand for compute is infinite, and the companies building the infrastructure are positioned to capture unprecedented value. That narrative has driven equity valuations to levels that assume near-flawless execution. What the credit market is now telling us is that flawless execution is not guaranteed. The buildout is expensive. The debt is real. And the consequences of failure will be borne not by the equity holders, who can walk away, but by the bondholders and the guarantors, who must be paid. This is the narrative shift I am hunting for. The next cycle of the AI market will not be defined by model quality or GPU performance. It will be defined by balance sheet discipline. The companies that manage their leverage effectively—that maintain investment-grade ratings, that keep their CDS spreads tight, that preserve access to cheap capital—will be the ones that dominate the next phase of the industry. The companies that overextended, that took on too much debt in the pursuit of growth, will be the cautionary tales. In the crypto ecosystem, we have seen this movie before. The 2021 bull market was dominated by leverage. Every project, every protocol, every exchange borrowed aggressively to expand market share. When the credit cycle turned, the leveraged players collapsed. The companies that survived—and then thrived—were the ones that had maintained positive cash flow, minimal debt, and the discipline to say no to easy money. The same evolution is now playing out in AI. The identity cards will be different, but the plot is identical. Let me be direct about my conviction. The credit market is right to be worried, but the worry should not be interpreted as a prediction of bankruptcy. It should be interpreted as a call to rigor. A call to scrutinize the balance sheets of AI companies as carefully as we have scrutinized their model architecture. A call to ask the uncomfortable questions: Who is going to pay for this? What happens if the revenue growth slows? How much cushion exists if the market turns? I have spent my career navigating the intersection of technology and finance, and I have learned that the most expensive mistakes always come from ignoring the financial structure in favor of the technological story. Ethereum was a technological masterpiece. That did not stop the price of ETH from falling more than 90 percent in the 2018 bear market. The technology matters for the long term. But the credit cycle determines who survives to see it. The AI industry is no different. The technology is extraordinary. The models are impressive. The potential is real. But the industry is now in a phase where financial capacity is as important as technical capability, and the market is just beginning to understand that. Hunting for the story that defines the next cycle means understanding this moment. The story is not about GPUs. The story is not about model benchmarks. The story is about the balance sheet. That is where the next cycle will be won or lost. WHAT I AM WATCHING NEXT As I finalize this analysis, I am monitoring a set of specific signals that will tell us whether the credit risk is crystallizing into a genuine crisis or whether we are experiencing a temporary repricing. The next CoreWeave earnings report—I want to see the utilization rate of their installed capacity, the rate of new contract signings, and, critically, the debt service coverage ratio. If those numbers show deterioration, the 50 percent default probability will move toward reality. If they show stability, the credit market will begin to reverse its stance. The Nvidia investor communications—specifically any 8-K filings or investor day presentations that disclose further details on the 750 billion dollars in AI commitments. I want to know how much of that figure is a binding contractual obligation versus an aspirational target. The percentage matters enormously for Nvidia's credit profile. A 750 billion dollar figure that represents soft capacity commitments is fundamentally different from one that represents firm purchase obligations with penalty clauses. The free cash flow reports from Alphabet, Microsoft, Amazon, and Meta over the next two quarters. The question is whether Alphabet's negative free cash flow was a one-quarter anomaly or the beginning of a persistent trend. If all four hyperscalers report consecutive quarters of deteriorating free cash flow, the market will have no choice but to incorporate the credit signal into equity valuations. The regulatory engagement with the round-tripping allegation. If the SEC opens a formal probe into revenue recognition practices among AI infrastructure companies, the disclosure risks will create immediate repricing pressure. If no probe materializes, Burry's claim will remain in the realm of unverified speculation. And finally, the evolution of the decentralized compute market. If the centralized cloud providers hit stress, I expect to see meaningful volume shifts toward decentralized alternatives. The crypto ecosystem's compute-native protocols will spend the next two years demonstrating whether they can absorb demand that the centralized providers cannot serve. That test will define the AI plus crypto convergence narrative. Hunting for the story that defines the next cycle means watching these signals with urgency and precision. The equity market is still trading the old story—the story of infinite AI growth and unlimited compute demand. The credit market has already begun trading the next story. The next cycle will belong to the investors who close that gap. THE TAKEAWAY I will end with a question rather than a prediction, because that is the appropriate response to genuine uncertainty. The credit market is telling us that AI's first major bankruptcy is coming. The equity market is telling us that it will never happen. One of them is wrong. Which one? History says the credit market is the better forecaster of financial distress. The credit market is closer to the money. It is served by specialists whose livelihoods depend on getting the default probability right. It reacts more quickly to balance sheet deterioration. It is less susceptible to the narrative intoxication that afflicts equity investors. On every documented comparison, from Enron to Lehman Brothers to Terra, the credit market saw the failure first. But the credit market is not infallible. It over-priced defaults during the COVID crash, when government intervention prevented a cascade. It under-priced defaults during the 2008 crisis, when the interconnectedness of financial institutions created risks that individual credit analysis could not capture. And in the current situation, the credit market might be pricing a level of systemic risk that interventions could diffuse. The rational response is not to bet entirely on one side. It is to respect the signal, to prepare for the scenarios in which the signal is correct, and to remain open to the possibility that the signal is wrong. That is what I mean by pre-mortem structural skepticism. Assume the failure, identify the causes, and build a portfolio that survives the failure. Then, if the failure does not materialize, you will still have captured the opportunity in the high-quality names that benefit from the sector's resilience. The AI credit cycle is just beginning. The narrative that defines it is not written. The balance sheets are being tested. And the intersection of credit risk and compute infrastructure will be the most consequential story in both crypto and traditional technology markets over the next eighteen months. That is where the signal is. That is where the opportunity is. That is what I am hunting for.