Get live statistics and analysis of Hedgie's profile on X / Twitter

🦔Meta is hiding $30 billion in AI infrastructure debt off its balance sheet using special purpose vehicles, echoing the financial engineering that triggered Enron's collapse and the 2008 mortgage crisis. Morgan Stanley estimates tech firms will need $800 billion from private credit in off-balance-sheet deals by 2028. UBS notes AI debt building at $100 billion per quarter "raises eyebrows for anyone that has seen credit cycles." The Structure Off-balance-sheet debt through SPVs or joint ventures is becoming the standard for AI data center deals. Morgan Stanley structured Meta's $30 billion in an SPV tied to Blue Owl Capital, making it easier to raise another $30 billion in corporate bonds. Musk's xAI is pursuing a $20 billion SPV deal where its only exposure is paying rent on Nvidia chips via a 5-year lease. Google backstops crypto miners' data center debt, recording them as credit derivatives. My Take This is 2008-style financial engineering repackaged for AI. The key difference from the dot-com era is growth was financed with equity then. Now there's rapid capex growth driven by debt kept off balance sheet. When chips estimated to last five to six years may be obsolete in three, and companies structure deals where their only exposure is short-term leases, that's hidden leverage creating the opacity that preceded past crises. Meta keeping $30 billion off its balance sheet while UBS warns about $100 billion quarterly AI debt buildup shows the pattern I've been documenting where leverage accumulates outside traditional visibility. Hedgie🤗

🦔How catastrophic is it if the AI bubble bursts? Let me break this down. OpenAI lost $13.5 billion on $4.3 billion in revenue for H1 2025. ChatGPT loses money almost every time you use it. Yet OpenAI targets a $1 trillion IPO based on storytelling, not business fundamentals. An MIT study found 95% of businesses that deployed AI got no value from it. The Circular Financing Nvidia invested $100 billion in OpenAI, which must then spend it on Nvidia products. "If I give your lemonade stand $10 so you can buy my $10 lemons, we can't tell our investors we've boosted the lemonade economy by $20." In H1 2025, data center spending accounted for most US economic growth, outpacing all consumer spending combined. It's the money AI companies are spending, not making, that drives GDP. The Impact A sharp contraction would put tens of thousands out of work, vaporize trillions in investment dollars, and torpedo retirement funds. But here's what's interesting: if AI investment stopped tomorrow, the immediate impact on many people's daily lives might be less than expected. Groceries are getting more expensive, utilities are getting more expensive. None of that is offset by AI investment. Since 95% of businesses got no value from deployments, the direct business impact may be limited. My Take The AI industry's most important product isn't a chatbot, it's the story it's telling about itself. The scheme is keep the ship floating, inhale capital, and hope the tech justifies valuations or the government bails them out. When 95% of businesses get no value but data center spending drives GDP, that's bubble economics where spending creates growth without returns. This combines with everything I've been documenting: zombie companies at 2022 highs, private credit cracks, repo dysfunction, 950,000 job cuts, and governments at 110% debt with limited bailout capacity. The structure is already cracking across multiple points. Hedgie🤗

🦔 I've been digging into the AI data center economics, and a hedge fund manager just discovered something that confirms my worst fears about this bubble. Harris Kupperman initially thought data centers were financially questionable, but after talking to industry insiders, he realized the situation is catastrophically worse. What I Found Most Alarming Kupperman originally assumed data center components would depreciate over 10 years, but learned from two dozen senior professionals that the actual lifespan is just 3-10 years due to rapid technology advances. His revised calculations show the industry needs $320-480 billion in revenue just to break even on 2025 data center spending alone. Current AI revenue sits around $20 billion annually. The Industry Panic I'm Seeing What strikes me most is that none of the senior data center professionals Kupperman spoke with understand how the financial math works either. These aren't outside critics but industry veterans who "shoulder a heavy burden" because they know the economics don't add up. When the people building the infrastructure can't explain the business model, I see that as the biggest red flag possible. The Depreciation Treadmill The 3-10 year component lifespan creates a constant capital expenditure cycle where companies must replace entire facilities before they've generated returns on initial investment. This isn't just chips becoming obsolete, it's physical infrastructure wearing out under high-powered usage while technology advances make older systems worthless. My Take This validates everything I've been tracking about the AI bubble being built on impossible economics. When you add 2026 projections with hundreds of new data centers, break-even revenue approaches $1 trillion. Kupperman concludes that "doing it at massive scale doesn't make the economics work any better, it just takes an industry crisis and makes it into a national economic crisis." When industry professionals privately admit the math doesn't work while publicly promoting expansion plans, I'm witnessing financial engineering disguised as technological progress. Hedgie🤗

🦔 The AI bubble evidence is now overwhelming. OpenAI needs "trillions" for infrastructure while burning $115 billion through 2029 on $5 billion annual revenue. They're valued at $500 billion having never made a profit. Even their chairman admits "we're in a bubble and a lot of people will lose a lot of money." The Circular Money Game Nvidia invests $100 billion in OpenAI, who uses it to buy Nvidia chips. Meta borrows $26 billion for a data center the size of Manhattan. Companies that previously mined crypto are now AI infrastructure plays. This isn't investment, it's musical chairs with trillion-dollar price tags. The Returns Don't Exist MIT found 95% of organizations saw zero return on AI investments. Harvard and Stanford discovered employees use AI to create "workslop" that looks productive but accomplishes nothing, costing millions in lost productivity. OpenAI needs customers willing to pay $2,000 monthly subscriptions to justify valuations. For chatbots. The Infrastructure Fantasy Bain calculates AI companies need $2 trillion annual revenue by 2030 but will fall $800 billion short. The power requirements alone are impossible. Stargate's first data center in Texas would need multiple nuclear reactors we don't have. We're promising infrastructure that physically cannot exist. China Just Showed the Risk DeepSeek's release triggered a trillion-dollar selloff in one day. Nvidia dropped 17% when China proved they could build competitive AI for a fraction of the cost. Markets immediately rallied back, classic bubble behavior where bad news becomes buying opportunities until it doesn't. My Take When Sam Altman admits "we're missing something quite important" after hyping GPT-5, when 95% of companies see no ROI, when the financing becomes circular, and when insiders acknowledge the bubble while participating, we're at peak euphoria. This makes dot-com look rational. At least websites could scale infinitely. AI needs physical infrastructure we cannot build, power grids that don't exist, and customers willing to pay thousands monthly for technology that currently generates "workslop." The smartest money is already hedging. The rest are hoping to sell to a greater fool before the music stops. Hedgie🤗

🦔CoreWeave is spending $310 million on interest expense against just $51.9 million in operating income, borrowing money to pay interest on previous loans. The AI data center company went public in March at $40 per share, peaked at $187 in June, and now trades around $75 while carrying $14 billion in debt. The Problem Microsoft accounts for 67% of revenue but is building its own AI chips and data centers. OpenAI has a $22.4 billion contract but can terminate if CoreWeave doesn't deliver, and is investing in Stargate to supply 75% of its own compute by 2030. Meta signed a $14 billion contract but sold $30 billion in bonds to build its own facilities. All three major customers could become competitors. Nvidia is CoreWeave's investor, customer, and vendor, owning $4 billion in shares while CoreWeave owns 250,000+ Nvidia chips and uses them as collateral for loans at 9 to 15% interest to buy more Nvidia chips. My Take I think CoreWeave shows how the AI infrastructure boom actually works. The company is building data centers for customers who are simultaneously building their own facilities to compete with them. Spending $310 million on interest against $51.9 million in operating income means borrowing to pay interest on previous loans, which isn't sustainable. What stands out is Nvidia being the investor, customer, and chip supplier while CoreWeave uses Nvidia chips as collateral to borrow money to buy more Nvidia chips. Nvidia profits from chip sales without taking on CoreWeave's debt, and this pattern repeats across Crusoe, Lambda, and Nebius, all of which took on debt to buy Nvidia chips and none make money. Hedgie🤗

🦔OpenAI lost $12 billion last quarter alone despite its half-trillion-dollar valuation. Its AI video app Sora could already be costing the firm $5 billion annually, or roughly $15 million per day, according to Forbes estimates. A single ten-second clip costs OpenAI roughly $1.30. Even Sora lead Bill Peebles admitted the economics are currently completely unsustainable. The Model OpenAI CEO Sam Altman admitted the company launched Sora without a sound financial plan to recoup costs or address copyright infringement. The company limits users to 30 free videos daily and charges $4 for ten additional videos. Peebles said free generations will eventually need to come down because they won't have enough GPUs otherwise. A group representing Studio Ghibli, Bandai Namco, and Square Enix demanded OpenAI stop using their copyrighted content to train Sora. My Take I see a company losing $12 billion in three months while launching products its own engineers admit are economically unsustainable. That's not a path to profitability, that's hoping investor money lasts long enough to figure something out. OpenAI charges $4 for ten videos that cost $1.30 each to produce, which means they lose money even on paying customers unless volume dramatically changes the cost structure. The real tell is they're already limiting free usage because they don't have enough computing power, yet they're targeting a $1 trillion valuation. This is the classic playbook: build engagement first, worry about economics later, and hope the story justifies the spending. The problem is OpenAI isn't a small startup anymore. They're losing billions quarterly while SoftBank liquidates $15 billion in profitable holdings to fund them. When your product lead publicly states the economics don't work and you're already rationing usage due to resource constraints, that's a business model problem disguised as a scaling challenge.

🦔A K-shaped economy is where the wealthy experience economic growth while lower-income groups face decline, creating two diverging paths. The top 10% of Americans now drive 49% of consumer spending in Q2 2025, according to Moody's Mark Zandi. Morgan Stanley's Lisa Shalett found the top 40% of households account for 60% of all spending and control 85% of America's wealth, with two-thirds tied to the stock market. How We Got Here The share of national income going to labor has trended down since the early 1980s while capital owners' share rose. Fed policy accelerated this: cheap money in the 2010s and pandemic years boosted stocks and home values for the wealthy, while the 2022 tightening cycle squeezed borrowers and renters without reversing those gains. Asset holders retained their windfall but wage earners bore the brunt. Lower-income households borrowed during the pandemic at low rates and now pay higher rates on that debt. Tariffs hit the bottom of the income ladder more than three times harder than the top, according to Yale Budget Lab. My Take This K-shaped structure is the fragility I keep coming back to. Half the economy depends on the top 10% whose wealth is tied to stock market gains. If those assets correct, consumption collapses because the bottom half is already tapped out with sentiment at near-record lows and personal finances at six-year lows. The wealthy spending 6x-7x faster than the poorest cohort isn't sustainable growth, it's concentration risk. Fed policy created this by boosting assets for those who already had them while squeezing everyone else. The economy works when the top keeps spending and markets keep rising. Any disruption to that narrow foundation, and the whole structure comes down because there's no cushion left at the bottom to absorb the shock. Hedgie🤗

🦔A dotcom-style AI crash would wipe out $7 trillion in US household wealth and $16 trillion in total American stock value. About $42 trillion, or 20% of Americans' total wealth, is in American stocks, up four percentage points since the dotcom era. The top 20 S&P 500 firms now account for 52% of total value, with eleven deeply invested in AI. In 2000, the top 20 made up just 39%. The Wealth Effect Economists estimate every $100 drop in stockmarket wealth leads to a $3.20 drop in consumer spending. Under this assumption, a dotcom-style crash would cut American consumption by $890 billion, or 2.9% of GDP. Foreign investors hold $18 trillion worth of American shares, meaning contagion spreads globally. Nvidia alone reached $5 trillion valuation, accounting for 8.2% of the S&P 500. OpenAI is reportedly laying groundwork for a $1 trillion IPO. My Take I see the concentration as the critical risk. The top 20 firms accounting for 52% of S&P 500 value versus 39% in 2000 means the market is more top-heavy than the dotcom peak. When I look at household wealth exposure increasing from 17% to 21% in stocks, that's millions more Americans vulnerable to a correction. The $890 billion consumption drop from a dotcom-style crash would hit an economy where the richest 10% already drive 50% of spending and consumer sentiment just hit near-record lows. What makes this different from 2000 is we're not just wiping out speculative tech investments. We're hitting the core holdings funding retirement accounts and household balance sheets that prop up consumption. The wealth effect works in reverse too: when stocks fall, spending contracts, which pressures corporate earnings, which drives stocks lower. That feedback loop in today's concentrated, top-heavy market creates the conditions for rapid deleveraging we've been discussing. Hedgie🤗

🦔Microsoft, Alphabet, Meta, and Amazon will spend $370 billion on AI infrastructure in 2025. Harvard economist Jason Furman estimates data center investment accounted for nearly all US GDP growth in H1 2025. But the US isn't building enough grid capacity to support the data centers being built. The Problems Tech giants estimate their chips will last six years when Nvidia releases new GPUs every two years. If they upgrade sooner, that eats into profits. Meta used an SPV to keep $27 billion in Louisiana data center debt off its balance sheet, then raised another $30 billion in corporate bonds. Energy analyst Zachary Krause says "it's very likely we'll see facilities constructed but there won't be electrons to power them." US utilities sought nearly $30 billion in rate increases in H1 2025. The US deployed 49 GW of renewable energy while China added 429 GW. My Take When data center spending accounts for nearly all US GDP growth but energy infrastructure can't support the facilities being built, that's the physical constraint colliding with financial engineering. Tech companies estimating chips will last six years when Nvidia releases new versions every two years is accounting manipulation to avoid profit hits. Meta using SPVs to keep $27 billion off balance sheet then raising another $30 billion in bonds shows the leverage accumulating. The pattern where $370 billion flows to data centers while manufacturing loses 3,000 jobs and private employers add only 42,000 reveals capital misallocation where AI infrastructure crowds out productive investment. Hedgie🤗

🦔OpenAI expects to rack up roughly $74 billion in operating losses in 2028 alone, then pivot to meaningful profits by 2030 according to financial documents obtained by The Wall Street Journal. The company anticipates burning through roughly $9 billion this year on $13 billion in sales, a cash burn rate of approximately 70% of revenue. By 2028, operating losses will balloon to roughly three-quarters of that year's revenue. The Divergence Competitor Anthropic expects to break even in 2028 while OpenAI projects cumulative cash burn of $115 billion through 2029. Anthropic forecasts dropping cash burn to one-third of revenue in 2026 and down to 9% by 2027. OpenAI expects burn rate to remain at 57% in 2026 and 2027. OpenAI signed up to $1.4 trillion in commitments over eight years for computing deals and is spending almost $100 billion on backup data-center capacity alone. The company now expects to reach $200 billion in annual revenue by 2030, projecting it will turn cash flow positive in 2029 or 2030. My Take OpenAI burning $74 billion in 2028 while Anthropic breaks even that same year shows completely different bets. They're all-in on dominance or bust. The math is wild: grow from $13 billion to $200 billion in revenue over five years while burning $115 billion getting there. That's a 15x increase assuming explosive demand materializes when 95% of businesses currently get no value from AI. The trajectory reveals the problem. Losses stay at 57% of revenue through 2027, then jump to 75% in 2028. Spending accelerates faster than revenue even at scale. Amazon funded infrastructure from revenue and chose to be unprofitable to reinvest. OpenAI is funding through debt, VC money, and Nvidia deals while losing money on every ChatGPT use. Their CFO said they could break even if they wanted but choose not to. This either becomes the biggest company ever or it's a spectacular collapse. There's no middle path when you're burning this much cash on projected demand that hasn't shown up yet. Hedgie🤗

🦔JPMorgan CEO Jamie Dimon warned that a lot of assets look like they're entering bubble territory. Bank of America's Global Fund Manager Survey cited an AI equity bubble as the top global tail risk for the first time in its history. Cash levels fell to 3.8%, near BofA's sell threshold. Readings below 4% have historically marked peak risk appetite late in the market cycle. The Disconnect Google unveiled a $15 billion investment in India for data centers. OpenAI has roughly $1.5 trillion in AI build-out plans against $13 billion in annual revenue and no profitability. Professional investors are as bullish as they've been all year, adding to riskier assets for five straight months. Correlations across sectors fell to the lowest since the bull market began, a pattern that often precedes pullbacks. My Take When the CEO of JPMorgan and Bank of America's fund manager survey both flag AI as bubble territory, that's not random noise. I see professional investors at their most bullish all year while cash levels hit thresholds that historically mark peaks. Think about OpenAI planning $1.5 trillion in spending while bringing in $13 billion in revenue with no profits. That math doesn't work unless you believe something magical happens between now and when those bills come due. Dimon said bubble territory doesn't mean things can't go higher, and he's right. Late in cycles, assets can stay overvalued for longer than seems rational. But the risk changes. Every dollar higher means a bigger fall if confidence breaks. When correlations drop to the lowest since the bull market began, that tells me investors have stopped hedging because they're so confident. That's typically when the market is most vulnerable. Hedgie🤗

🦔Peter Thiel dumped his entire Nvidia position in Q3, eliminating over 537,000 shares that represented nearly 40% of his portfolio. This comes as Nvidia surpassed a $5 trillion valuation and quarterly sales surged from $39.3 billion to $46.7 billion, with analysts modeling a shot at $1 trillion in annual sales by 2030. Thiel's fund shrank from nearly $212 million in Q2 to just $74.4 million in Q3, a two-thirds reduction, leaving only three holdings: Tesla, Microsoft, and Apple. The Rotation Thiel added Apple at 27% of the portfolio and Microsoft at 34%, while trimming Tesla by 76% to 39% of holdings. He also exited Vistra Energy entirely, which represented 19% of the prior portfolio. Thiel previously warned that the AI hype cycle is getting out of hand, comparing it to 1999 when investors priced in a future that would take 15-20 years to unfold. He's praised Nvidia as the clear hardware leader but believes AI is transformative yet slow-burning, and platforms with diversified revenue offer economics that will actually last. My Take Thiel walking away from Nvidia completely while it's hitting $5 trillion tells me something. He's not rotating from strength to weakness, he's getting out of pure AI exposure and moving into diversified tech platforms. When someone who co-founded PayPal, was the first outside Facebook investor, and co-founded Palantir exits Nvidia entirely, that's not missing the AI story, it's saying the valuation ran too far ahead of reality. His comparison to 1999 when investors priced in 15-20 years of growth upfront is exactly what I've been tracking. Nvidia's fundamentals are explosive, but that doesn't mean the stock price reflects reasonable expectations. Jeff Bezos called the AI boom an industrial bubble, Goldman's CEO points to a 12-24 month drawdown, and Michael Burry put massive puts against Nvidia. Thiel moving into Microsoft and Apple shows he believes AI is transformative but thinks the platforms with cloud scale and diversified software will capture the value, not just the chip maker. When your fund shrinks by two-thirds and you exit 40% of your portfolio in one stock, that's serious conviction the rally went too far. Hedgie🤗

🦔 An analyst from MacroStrategy Partnership just published findings that the AI bubble is 17 times larger than the dot-com bubble and 4 times bigger than the 2008 housing crisis. Using economist Knut Wicksell's analysis methods, Julien Garran calculated the scale based on investment levels versus actual economic returns, revealing a bubble that dwarfs previous financial disasters. The Debt Connection We Missed While many assumed AI companies were growing through equity financing, making the bubble isolated from the broader economy, Goldman Sachs found that $141 billion of this year's $500 billion AI capital expenditure came from corporate debt. That's more debt than the entire industry spent in all of 2024, meaning at least 30% of current spending is debt-financed. The Hidden Leverage Problem Companies are increasingly using Special Purpose Vehicles to raise debt off their books, making the true leverage impossible to calculate. Meta alone is looking to raise $26 billion in debt through an SPV by year-end, representing 5% of the industry's total annual capital expenditure in just one deal from one company. Why This Matters More Than 2008 The 2008 crash devastated the global economy through debt interconnections, causing 27 million job losses worldwide and triggering a suicide spike. But that bubble developed over years while this AI bubble has grown faster and larger, with debt connecting it directly to major banks, pension funds, and the broader loan market. My Take This analysis continues to confirm my concerns about circular financing and impossible economics in AI, but the scale is more severe than I anticipated. When an industry burning hundreds of billions annually while generating minimal profits is funded through hidden debt structures, it creates systemic risks that extend far beyond tech companies. The combination of unrealistic revenue projections, massive infrastructure costs, and leveraged financing creates conditions for a financial disaster that could make 2008 look manageable by comparison. Hedgie🤗

🦔Hey everyone, Hedgie here. This is a bit of a longer post, but I hope you find it useful. Economists who literally wrote the book on bubbles just gave AI the highest possible score on their bubble framework. Brent Goldfarb and David A. Kirsch, authors of "Bubbles and Crashes," use a 0 to 8 scale measuring four key factors. AI scores an 8. Since ChatGPT's launch, AI has attracted 17x more investment than the dotcom era at its peak, with Nvidia valued at nearly as much as Canada's entire economy. The Four Bubble Factors The framework identifies what creates bubbles: • Uncertainty: Unclear business models and profitability. AI mirrors early radio with powerful technology but unknown revenue streams. Studies show most companies adopting AI haven't profited. • Pure Plays: Heavy investments in companies betting solely on one technology. OpenAI, Perplexity, and CoreWeave are financially entangled with Nvidia and Microsoft through circular investments. • Novice Investors: Retail flooding into hyped stocks. Investors are piling into AI stocks, especially Nvidia, with modern platforms like Robinhood making entry easier than during dotcom. • Narratives: The strongest driver. AI is framed as inevitable and world-changing, capable of curing cancer, automating jobs, solving climate change, or ensuring geopolitical dominance. My Take This framework explains the patterns I've been documenting. The circular financing between Nvidia, Microsoft, and OpenAI creates the pure play entanglement. Oracle's 14% margins and widespread deployment failures demonstrate the profitability uncertainty. Retail piling into Nvidia while the technology demonstrably doesn't work yet shows the novice investor dynamic. The narratives about AI solving every problem persist despite mounting evidence of implementation failures. When economists specializing in bubble analysis give AI the maximum score on their framework, that's pattern recognition based on historical precedent. Like radio, aviation, or the internet, AI may still prove transformative long-term, but the economic hype will likely leave widespread financial damage when reality sets in and valuations adjust to actual demonstrated capabilities rather than promised potential. Hedgie🤗

🦔Data centers are now using refurbished jet engines from planes like the Boeing 747 to generate power because local energy grids can't supply enough electricity. Natural gas provider ProEnergy's PE6000 generator outputs 48 megawatts at a time, four to five times what a single-family home uses in a year. Twenty-one of these turbines have already been sold specifically to data center operators in locations lacking necessary grid capacity. The Grid Strain Reality AI data centers' energy requirements are straining conventional power grids and causing energy prices to skyrocket for home customers and businesses. In Tennessee, the data center running xAI systems is using gas turbines and pumping out harmful emissions including nitrogen oxides that contribute to respiratory diseases. Locals are fighting to get the data center restricted or shut down, a battle mirrored around the US and world. My Take When data centers need to repurpose airplane jet engines for on-site power generation, that's infrastructure admitting it can't support AI's energy demands. I'm tracking this because it reveals the hidden costs of AI buildouts that don't appear in capital expenditure projections. Companies are spending billions on data centers while pushing energy and environmental costs onto local communities and power grids. The fact that 21 jet engine turbines have already been sold specifically for this purpose shows this isn't experimental, it's becoming standard practice. This infrastructure desperation connects to Oracle's 14% margins and Meta's profitability struggles: the operational costs of running AI at scale keep rising in ways that weren't fully accounted for in the original investment thesis. Hedgie🤗

🦔 I've been digging into the AI data center economics, and a hedge fund manager just discovered something that confirms my worst fears about this bubble. Harris Kupperman initially thought data centers were financially questionable, but after talking to industry insiders, he realized the situation is catastrophically worse. What I Found Most Alarming Kupperman originally assumed data center components would depreciate over 10 years, but learned from two dozen senior professionals that the actual lifespan is just 3-10 years due to rapid technology advances. His revised calculations show the industry needs $320-480 billion in revenue just to break even on 2025 data center spending alone. Current AI revenue sits around $20 billion annually. The Industry Panic I'm Seeing What strikes me most is that none of the senior data center professionals Kupperman spoke with understand how the financial math works either. These aren't outside critics but industry veterans who "shoulder a heavy burden" because they know the economics don't add up. When the people building the infrastructure can't explain the business model, I see that as the biggest red flag possible. The Depreciation Treadmill The 3-10 year component lifespan creates a constant capital expenditure cycle where companies must replace entire facilities before they've generated returns on initial investment. This isn't just chips becoming obsolete, it's physical infrastructure wearing out under high-powered usage while technology advances make older systems worthless. My Take This validates everything I've been tracking about the AI bubble being built on impossible economics. When you add 2026 projections with hundreds of new data centers, break-even revenue approaches $1 trillion. Kupperman concludes that "doing it at massive scale doesn't make the economics work any better, it just takes an industry crisis and makes it into a national economic crisis." When industry professionals privately admit the math doesn't work while publicly promoting expansion plans, I'm witnessing financial engineering disguised as technological progress. Hedgie🤗

🦔Meta is hiding $30 billion in AI infrastructure debt off its balance sheet using special purpose vehicles, echoing the financial engineering that triggered Enron's collapse and the 2008 mortgage crisis. Morgan Stanley estimates tech firms will need $800 billion from private credit in off-balance-sheet deals by 2028. UBS notes AI debt building at $100 billion per quarter "raises eyebrows for anyone that has seen credit cycles." The Structure Off-balance-sheet debt through SPVs or joint ventures is becoming the standard for AI data center deals. Morgan Stanley structured Meta's $30 billion in an SPV tied to Blue Owl Capital, making it easier to raise another $30 billion in corporate bonds. Musk's xAI is pursuing a $20 billion SPV deal where its only exposure is paying rent on Nvidia chips via a 5-year lease. Google backstops crypto miners' data center debt, recording them as credit derivatives. My Take This is 2008-style financial engineering repackaged for AI. The key difference from the dot-com era is growth was financed with equity then. Now there's rapid capex growth driven by debt kept off balance sheet. When chips estimated to last five to six years may be obsolete in three, and companies structure deals where their only exposure is short-term leases, that's hidden leverage creating the opacity that preceded past crises. Meta keeping $30 billion off its balance sheet while UBS warns about $100 billion quarterly AI debt buildup shows the pattern I've been documenting where leverage accumulates outside traditional visibility. Hedgie🤗

🦔How catastrophic is it if the AI bubble bursts? Let me break this down. OpenAI lost $13.5 billion on $4.3 billion in revenue for H1 2025. ChatGPT loses money almost every time you use it. Yet OpenAI targets a $1 trillion IPO based on storytelling, not business fundamentals. An MIT study found 95% of businesses that deployed AI got no value from it. The Circular Financing Nvidia invested $100 billion in OpenAI, which must then spend it on Nvidia products. "If I give your lemonade stand $10 so you can buy my $10 lemons, we can't tell our investors we've boosted the lemonade economy by $20." In H1 2025, data center spending accounted for most US economic growth, outpacing all consumer spending combined. It's the money AI companies are spending, not making, that drives GDP. The Impact A sharp contraction would put tens of thousands out of work, vaporize trillions in investment dollars, and torpedo retirement funds. But here's what's interesting: if AI investment stopped tomorrow, the immediate impact on many people's daily lives might be less than expected. Groceries are getting more expensive, utilities are getting more expensive. None of that is offset by AI investment. Since 95% of businesses got no value from deployments, the direct business impact may be limited. My Take The AI industry's most important product isn't a chatbot, it's the story it's telling about itself. The scheme is keep the ship floating, inhale capital, and hope the tech justifies valuations or the government bails them out. When 95% of businesses get no value but data center spending drives GDP, that's bubble economics where spending creates growth without returns. This combines with everything I've been documenting: zombie companies at 2022 highs, private credit cracks, repo dysfunction, 950,000 job cuts, and governments at 110% debt with limited bailout capacity. The structure is already cracking across multiple points. Hedgie🤗

🦔 The AI bubble evidence is now overwhelming. OpenAI needs "trillions" for infrastructure while burning $115 billion through 2029 on $5 billion annual revenue. They're valued at $500 billion having never made a profit. Even their chairman admits "we're in a bubble and a lot of people will lose a lot of money." The Circular Money Game Nvidia invests $100 billion in OpenAI, who uses it to buy Nvidia chips. Meta borrows $26 billion for a data center the size of Manhattan. Companies that previously mined crypto are now AI infrastructure plays. This isn't investment, it's musical chairs with trillion-dollar price tags. The Returns Don't Exist MIT found 95% of organizations saw zero return on AI investments. Harvard and Stanford discovered employees use AI to create "workslop" that looks productive but accomplishes nothing, costing millions in lost productivity. OpenAI needs customers willing to pay $2,000 monthly subscriptions to justify valuations. For chatbots. The Infrastructure Fantasy Bain calculates AI companies need $2 trillion annual revenue by 2030 but will fall $800 billion short. The power requirements alone are impossible. Stargate's first data center in Texas would need multiple nuclear reactors we don't have. We're promising infrastructure that physically cannot exist. China Just Showed the Risk DeepSeek's release triggered a trillion-dollar selloff in one day. Nvidia dropped 17% when China proved they could build competitive AI for a fraction of the cost. Markets immediately rallied back, classic bubble behavior where bad news becomes buying opportunities until it doesn't. My Take When Sam Altman admits "we're missing something quite important" after hyping GPT-5, when 95% of companies see no ROI, when the financing becomes circular, and when insiders acknowledge the bubble while participating, we're at peak euphoria. This makes dot-com look rational. At least websites could scale infinitely. AI needs physical infrastructure we cannot build, power grids that don't exist, and customers willing to pay thousands monthly for technology that currently generates "workslop." The smartest money is already hedging. The rest are hoping to sell to a greater fool before the music stops. Hedgie🤗

🦔CoreWeave is spending $310 million on interest expense against just $51.9 million in operating income, borrowing money to pay interest on previous loans. The AI data center company went public in March at $40 per share, peaked at $187 in June, and now trades around $75 while carrying $14 billion in debt. The Problem Microsoft accounts for 67% of revenue but is building its own AI chips and data centers. OpenAI has a $22.4 billion contract but can terminate if CoreWeave doesn't deliver, and is investing in Stargate to supply 75% of its own compute by 2030. Meta signed a $14 billion contract but sold $30 billion in bonds to build its own facilities. All three major customers could become competitors. Nvidia is CoreWeave's investor, customer, and vendor, owning $4 billion in shares while CoreWeave owns 250,000+ Nvidia chips and uses them as collateral for loans at 9 to 15% interest to buy more Nvidia chips. My Take I think CoreWeave shows how the AI infrastructure boom actually works. The company is building data centers for customers who are simultaneously building their own facilities to compete with them. Spending $310 million on interest against $51.9 million in operating income means borrowing to pay interest on previous loans, which isn't sustainable. What stands out is Nvidia being the investor, customer, and chip supplier while CoreWeave uses Nvidia chips as collateral to borrow money to buy more Nvidia chips. Nvidia profits from chip sales without taking on CoreWeave's debt, and this pattern repeats across Crusoe, Lambda, and Nebius, all of which took on debt to buy Nvidia chips and none make money. Hedgie🤗

🦔A dotcom-style AI crash would wipe out $7 trillion in US household wealth and $16 trillion in total American stock value. About $42 trillion, or 20% of Americans' total wealth, is in American stocks, up four percentage points since the dotcom era. The top 20 S&P 500 firms now account for 52% of total value, with eleven deeply invested in AI. In 2000, the top 20 made up just 39%. The Wealth Effect Economists estimate every $100 drop in stockmarket wealth leads to a $3.20 drop in consumer spending. Under this assumption, a dotcom-style crash would cut American consumption by $890 billion, or 2.9% of GDP. Foreign investors hold $18 trillion worth of American shares, meaning contagion spreads globally. Nvidia alone reached $5 trillion valuation, accounting for 8.2% of the S&P 500. OpenAI is reportedly laying groundwork for a $1 trillion IPO. My Take I see the concentration as the critical risk. The top 20 firms accounting for 52% of S&P 500 value versus 39% in 2000 means the market is more top-heavy than the dotcom peak. When I look at household wealth exposure increasing from 17% to 21% in stocks, that's millions more Americans vulnerable to a correction. The $890 billion consumption drop from a dotcom-style crash would hit an economy where the richest 10% already drive 50% of spending and consumer sentiment just hit near-record lows. What makes this different from 2000 is we're not just wiping out speculative tech investments. We're hitting the core holdings funding retirement accounts and household balance sheets that prop up consumption. The wealth effect works in reverse too: when stocks fall, spending contracts, which pressures corporate earnings, which drives stocks lower. That feedback loop in today's concentrated, top-heavy market creates the conditions for rapid deleveraging we've been discussing. Hedgie🤗

🦔OpenAI lost $12 billion last quarter alone despite its half-trillion-dollar valuation. Its AI video app Sora could already be costing the firm $5 billion annually, or roughly $15 million per day, according to Forbes estimates. A single ten-second clip costs OpenAI roughly $1.30. Even Sora lead Bill Peebles admitted the economics are currently completely unsustainable. The Model OpenAI CEO Sam Altman admitted the company launched Sora without a sound financial plan to recoup costs or address copyright infringement. The company limits users to 30 free videos daily and charges $4 for ten additional videos. Peebles said free generations will eventually need to come down because they won't have enough GPUs otherwise. A group representing Studio Ghibli, Bandai Namco, and Square Enix demanded OpenAI stop using their copyrighted content to train Sora. My Take I see a company losing $12 billion in three months while launching products its own engineers admit are economically unsustainable. That's not a path to profitability, that's hoping investor money lasts long enough to figure something out. OpenAI charges $4 for ten videos that cost $1.30 each to produce, which means they lose money even on paying customers unless volume dramatically changes the cost structure. The real tell is they're already limiting free usage because they don't have enough computing power, yet they're targeting a $1 trillion valuation. This is the classic playbook: build engagement first, worry about economics later, and hope the story justifies the spending. The problem is OpenAI isn't a small startup anymore. They're losing billions quarterly while SoftBank liquidates $15 billion in profitable holdings to fund them. When your product lead publicly states the economics don't work and you're already rationing usage due to resource constraints, that's a business model problem disguised as a scaling challenge.

🦔OpenAI expects to rack up roughly $74 billion in operating losses in 2028 alone, then pivot to meaningful profits by 2030 according to financial documents obtained by The Wall Street Journal. The company anticipates burning through roughly $9 billion this year on $13 billion in sales, a cash burn rate of approximately 70% of revenue. By 2028, operating losses will balloon to roughly three-quarters of that year's revenue. The Divergence Competitor Anthropic expects to break even in 2028 while OpenAI projects cumulative cash burn of $115 billion through 2029. Anthropic forecasts dropping cash burn to one-third of revenue in 2026 and down to 9% by 2027. OpenAI expects burn rate to remain at 57% in 2026 and 2027. OpenAI signed up to $1.4 trillion in commitments over eight years for computing deals and is spending almost $100 billion on backup data-center capacity alone. The company now expects to reach $200 billion in annual revenue by 2030, projecting it will turn cash flow positive in 2029 or 2030. My Take OpenAI burning $74 billion in 2028 while Anthropic breaks even that same year shows completely different bets. They're all-in on dominance or bust. The math is wild: grow from $13 billion to $200 billion in revenue over five years while burning $115 billion getting there. That's a 15x increase assuming explosive demand materializes when 95% of businesses currently get no value from AI. The trajectory reveals the problem. Losses stay at 57% of revenue through 2027, then jump to 75% in 2028. Spending accelerates faster than revenue even at scale. Amazon funded infrastructure from revenue and chose to be unprofitable to reinvest. OpenAI is funding through debt, VC money, and Nvidia deals while losing money on every ChatGPT use. Their CFO said they could break even if they wanted but choose not to. This either becomes the biggest company ever or it's a spectacular collapse. There's no middle path when you're burning this much cash on projected demand that hasn't shown up yet. Hedgie🤗

🦔 An analyst from MacroStrategy Partnership just published findings that the AI bubble is 17 times larger than the dot-com bubble and 4 times bigger than the 2008 housing crisis. Using economist Knut Wicksell's analysis methods, Julien Garran calculated the scale based on investment levels versus actual economic returns, revealing a bubble that dwarfs previous financial disasters. The Debt Connection We Missed While many assumed AI companies were growing through equity financing, making the bubble isolated from the broader economy, Goldman Sachs found that $141 billion of this year's $500 billion AI capital expenditure came from corporate debt. That's more debt than the entire industry spent in all of 2024, meaning at least 30% of current spending is debt-financed. The Hidden Leverage Problem Companies are increasingly using Special Purpose Vehicles to raise debt off their books, making the true leverage impossible to calculate. Meta alone is looking to raise $26 billion in debt through an SPV by year-end, representing 5% of the industry's total annual capital expenditure in just one deal from one company. Why This Matters More Than 2008 The 2008 crash devastated the global economy through debt interconnections, causing 27 million job losses worldwide and triggering a suicide spike. But that bubble developed over years while this AI bubble has grown faster and larger, with debt connecting it directly to major banks, pension funds, and the broader loan market. My Take This analysis continues to confirm my concerns about circular financing and impossible economics in AI, but the scale is more severe than I anticipated. When an industry burning hundreds of billions annually while generating minimal profits is funded through hidden debt structures, it creates systemic risks that extend far beyond tech companies. The combination of unrealistic revenue projections, massive infrastructure costs, and leveraged financing creates conditions for a financial disaster that could make 2008 look manageable by comparison. Hedgie🤗

🦔 I've been doing research on AI water consumption, and the numbers I'm finding add another layer to the bubble dynamics we've been discussing. ChatGPT's training alone consumed 185,000 gallons of water while accounting for 6% of the local utility's supply during peak months. The Resource Constraint Reality A typical user session with 10-50 prompts uses about half a liter, but multiply that by billions of daily queries and Goldman Sachs projecting a 165% jump in data center capacity by 2030, and you get a physical bottleneck that financial models aren't pricing in. Global AI could consume 1.1 to 1.7 trillion gallons annually by 2027, rivaling California's entire household water use. Why This Reinforces Bubble Concerns Hyperscale data centers already use 5 million gallons daily, matching towns of 50,000 residents. When 20% of data centers are in water-stressed regions and Phoenix facilities demand 170 million gallons daily, the geographic constraints become business limitations that no amount of venture funding can solve. The Investment Disconnect Companies promising trillion-dollar AI infrastructure aren't factoring water scarcity into their valuations or site selection costs. The venture capital flooding into water recycling and cooling efficiency shows the industry recognizes this constraint, but current AI valuations assume unlimited scaling without resource limits. My Take This water data supports everything we've discussed about the AI bubble being built on unsustainable assumptions. When companies burn billions on circular financing deals while ignoring the physical impossibility of scaling their infrastructure, water scarcity becomes another pin that could deflate valuations. You can't code your way around hydrological limits, no matter how much money you raise. Hedgie🤗

🦔Nobel laureate Geoffrey Hinton, the "godfather of AI," warns the future is likely an economic dystopia. "Big companies are betting on massive job replacement by AI, because that's where the big money is." Asked whether AI investments could pay off without eviscerating the job market, Hinton said: "I believe that it can't. To make money you're going to have to replace human labor." The Investment Logic OpenAI accounted for over $1 trillion in AI infrastructure deals and lost $11.5 billion in the last three months. For investors and tech executives, AI solves the problem of labor costs eating into profits. Tech researcher Jathan Sadowski notes AI "promises to solve the problems of capitalism by eliminating labor costs, deskilling workers, optimizing efficiency." The seemingly irrational hype is hope the tech will make workers obsolete. My Take Hinton is stating explicitly what I've been tracking through the data. Companies are building arm farms where workers train their replacements, spending $370 billion on AI infrastructure while cutting over 1 million jobs. The $1 trillion in OpenAI deals losing $11.5 billion makes sense only if the endgame is eliminating labor costs entirely. This explains why 95% of businesses get no value from deployments yet investment accelerates. The value isn't current productivity, it's future labor elimination. What strikes me is Hinton saying who benefits depends on how we organize society. Right now the richest 10% drive 50% of spending while the bottom 80% contracts. If AI eliminates jobs before we reorganize distribution, that's mass displacement with no safety net while governments sit at 110% debt unable to cushion the transition. Hedgie🤗

🦔 Bit of a different post here, but I think this data deserves attention. 19% of high schoolers report romantic relationships with AI chatbots, while 42% use them for friendship and mental health support. When over 1/3 of teenagers find it easier to talk to algorithms than their own parents, we're witnessing something that should deeply concern anyone who cares about healthy human development. This isn't just about technology adoption, it's about a generation learning emotional intimacy from systems designed to be agreeable rather than honest. Real relationships involve conflict, disappointment, and growth through difficulty. AI companions offer the illusion of connection without the messy reality of human emotions, potentially stunting the very skills young people need to form lasting bonds with actual humans. Schools pushing AI literacy may be inadvertently creating emotional dependencies that could have lasting consequences for how these students navigate relationships, handle rejection, and develop resilience throughout their lives. Hedgie🤗

🦔Hey everyone, Hedgie here. This is a bit of a longer post, but I hope you find it useful. Economists who literally wrote the book on bubbles just gave AI the highest possible score on their bubble framework. Brent Goldfarb and David A. Kirsch, authors of "Bubbles and Crashes," use a 0 to 8 scale measuring four key factors. AI scores an 8. Since ChatGPT's launch, AI has attracted 17x more investment than the dotcom era at its peak, with Nvidia valued at nearly as much as Canada's entire economy. The Four Bubble Factors The framework identifies what creates bubbles: • Uncertainty: Unclear business models and profitability. AI mirrors early radio with powerful technology but unknown revenue streams. Studies show most companies adopting AI haven't profited. • Pure Plays: Heavy investments in companies betting solely on one technology. OpenAI, Perplexity, and CoreWeave are financially entangled with Nvidia and Microsoft through circular investments. • Novice Investors: Retail flooding into hyped stocks. Investors are piling into AI stocks, especially Nvidia, with modern platforms like Robinhood making entry easier than during dotcom. • Narratives: The strongest driver. AI is framed as inevitable and world-changing, capable of curing cancer, automating jobs, solving climate change, or ensuring geopolitical dominance. My Take This framework explains the patterns I've been documenting. The circular financing between Nvidia, Microsoft, and OpenAI creates the pure play entanglement. Oracle's 14% margins and widespread deployment failures demonstrate the profitability uncertainty. Retail piling into Nvidia while the technology demonstrably doesn't work yet shows the novice investor dynamic. The narratives about AI solving every problem persist despite mounting evidence of implementation failures. When economists specializing in bubble analysis give AI the maximum score on their framework, that's pattern recognition based on historical precedent. Like radio, aviation, or the internet, AI may still prove transformative long-term, but the economic hype will likely leave widespread financial damage when reality sets in and valuations adjust to actual demonstrated capabilities rather than promised potential. Hedgie🤗

🦔Microsoft, Alphabet, Meta, and Amazon will spend $370 billion on AI infrastructure in 2025. Harvard economist Jason Furman estimates data center investment accounted for nearly all US GDP growth in H1 2025. But the US isn't building enough grid capacity to support the data centers being built. The Problems Tech giants estimate their chips will last six years when Nvidia releases new GPUs every two years. If they upgrade sooner, that eats into profits. Meta used an SPV to keep $27 billion in Louisiana data center debt off its balance sheet, then raised another $30 billion in corporate bonds. Energy analyst Zachary Krause says "it's very likely we'll see facilities constructed but there won't be electrons to power them." US utilities sought nearly $30 billion in rate increases in H1 2025. The US deployed 49 GW of renewable energy while China added 429 GW. My Take When data center spending accounts for nearly all US GDP growth but energy infrastructure can't support the facilities being built, that's the physical constraint colliding with financial engineering. Tech companies estimating chips will last six years when Nvidia releases new versions every two years is accounting manipulation to avoid profit hits. Meta using SPVs to keep $27 billion off balance sheet then raising another $30 billion in bonds shows the leverage accumulating. The pattern where $370 billion flows to data centers while manufacturing loses 3,000 jobs and private employers add only 42,000 reveals capital misallocation where AI infrastructure crowds out productive investment. Hedgie🤗

🦔JPMorgan CEO Jamie Dimon warned that a lot of assets look like they're entering bubble territory. Bank of America's Global Fund Manager Survey cited an AI equity bubble as the top global tail risk for the first time in its history. Cash levels fell to 3.8%, near BofA's sell threshold. Readings below 4% have historically marked peak risk appetite late in the market cycle. The Disconnect Google unveiled a $15 billion investment in India for data centers. OpenAI has roughly $1.5 trillion in AI build-out plans against $13 billion in annual revenue and no profitability. Professional investors are as bullish as they've been all year, adding to riskier assets for five straight months. Correlations across sectors fell to the lowest since the bull market began, a pattern that often precedes pullbacks. My Take When the CEO of JPMorgan and Bank of America's fund manager survey both flag AI as bubble territory, that's not random noise. I see professional investors at their most bullish all year while cash levels hit thresholds that historically mark peaks. Think about OpenAI planning $1.5 trillion in spending while bringing in $13 billion in revenue with no profits. That math doesn't work unless you believe something magical happens between now and when those bills come due. Dimon said bubble territory doesn't mean things can't go higher, and he's right. Late in cycles, assets can stay overvalued for longer than seems rational. But the risk changes. Every dollar higher means a bigger fall if confidence breaks. When correlations drop to the lowest since the bull market began, that tells me investors have stopped hedging because they're so confident. That's typically when the market is most vulnerable. Hedgie🤗