$40 trillion in U.S. debt, $500 billion in AI financing—who will global long-term capital save first? Jensen Huang is about to start competing with Trump for money. Trump is selling U.S. debt in Washington, while Huang is offering installment plans for GPU customers in Silicon Valley. Whoever secures more long-term capital in the end may push the other's interest rates even higher. Just yesterday, Huang called together six financial giants including BlackRock, Blackstone, Goldman Sachs, and KKR to prepare to leverage over $500 billion in long-term capital from global capital markets for AI. Simply put, Nvidia’s customers are about to be unable to pay full price for his GPUs. So Huang is helping them find money, then letting Wall Street turn that money into loans and leases so customers can keep buying and using Nvidia’s chips. Meanwhile, Trump is waiting for the same group to put up money. U.S. debt is approaching $40 trillion, and he’s constantly urging the Federal Reserve to cut interest rates, hoping to make government borrowing cheaper. Why is Huang stepping in to find money right now? Because AI is no longer just short on chips—it’s also short on the money to buy those chips. Nvidia’s customers aren’t poor, but the AI bill has grown so large even they are struggling to pay it. How extreme is it? Research from the Dallas Fed estimates that since 2023, tech giants like Google, Microsoft, Meta, and Amazon have invested $500 billion to $600 billion in AI, much of it from their own profits. But that’s just the beginning. In the next 3 to 5 years, global AI data centers will consume another $3 trillion to $5 trillion. The $500 to $600 billion already spent only fills a small part of this huge gap. So even the richest tech giants can’t keep funding AI solely from their profits. Large cloud computing companies have started issuing long-term bonds, and many AI startups can’t even afford to buy GPUs outright, relying instead on loans or equipment leases. With money tight, Huang can’t just wait, so he brought in six financial giants. These six institutions plan to establish independent AI compute financing platforms, raising money from pensions, insurance funds, sovereign wealth funds, and other long-term investors, aiming to leverage over $500 billion in the future to build data centers, power infrastructure, and purchase GPUs. Of course, this $500 billion isn’t coming from Nvidia alone, nor is it being raised all at once. It’s currently just a cooperation framework; specific projects will be implemented one by one. Though the money isn’t fully in place yet, the tasks are clear. Previously, Nvidia only sold GPUs; now it also has to help customers secure loans. In other words, the biggest AI bottleneck used to be chip shortages. Now chip production capacity is expanding, but the money to buy chips is already insufficient. And this plan isn’t just theoretical—Nvidia has already tried it once. In January this year, Apollo arranged a $5.4 billion compute deal for XAI, with $3.5 billion provided by Apollo. The approach was straightforward: an independent company first bought compute equipment including Nvidia’s GB20, then leased it long-term to XAI. This way, XAI didn’t have to pay all cash upfront but could immediately access GPUs. Nvidia successfully sold chips, and the funding institutions received long-term lease payments. More importantly, Nvidia didn’t just stand by; it also participated as an investor in financing this compute infrastructure company. With this experience, Huang is now preparing to scale one project into a $500 billion financing system. At this stage, AI company growth no longer depends solely on securing GPUs—they also have to answer who is willing to pay upfront for them. However, Huang isn’t the first to sell equipment this way. In business, it’s called supplier financing—in plain terms, the seller helps the buyer borrow money first, then the buyer comes back to buy the seller’s goods. In the late 1990s, Cisco did the same. Telecom companies couldn’t afford routers and switches, so Cisco’s financial arm provided loans and leases. Customers got equipment, Cisco got orders, banks and investors got interest, and everyone seemed to profit. The business grew bigger and bigger. By 2001, Cisco had about $1.9 billion in loan commitments provided but not yet used by customers. Then the internet bubble burst, telecom companies suddenly lacked funds to expand, customers cut orders, inventory piled up at Cisco, and loan risks increased. Cisco’s products were useful, and the internet did change the world, but customers hadn’t earned enough yet, and equipment and debt had already piled up prematurely. Today, Nvidia faces the same risk: customers haven’t earned enough from AI yet but are already burdened with equipment leases and interest. Current GPUs have real demand, but a GPU’s depreciation usually lasts only five to six years, while loans and leases may extend longer. Years later, when chip performance lags, rent and interest still have to be paid. As long as AI customers earn enough to cover these costs, Huang’s financial system can continue accelerating AI development. But if AI revenue growth slows and old GPU rents decline, lenders will reassess the value of these devices. When financing tightens, customers will reduce purchasing pressure, eventually returning to Nvidia’s order book. Because of this, the Bank for International Settlements has warned of this risk: chip companies and cloud providers investing in AI customers create a mutually supporting financial chain through chip and compute purchase commitments. Everyone is ultimately betting on one thing: whether AI customers can quickly generate revenue. If any link’s income falls short, everyone will realize they were relying on the same future profits. Even if customers hold up, Huang faces another competitor: U.S. Treasury debt. Just in 2026, the Dallas Fed estimates AI-related companies may issue about $300 billion in investment-grade bonds. If different maturities are converted to 10-year equivalents, the long-term interest rate pressure is about one-eighth that of simultaneous U.S. Treasury issuance. When this $300 billion hits the market, AI companies will compete with the U.S. government for insurance funds, pensions, and global institutional investment. To get AI debt

Vitalik has laid out a brand-new grand vision for Ethereum, focusing next on three major technologies: Quantum-resistant cryptography, deep AI integration, and top-tier advanced privacy protection. 1 Quantum resistance: Prevent future quantum computers from cracking on-chain cryptography 2 Deep AI integration: Enable smart contracts to run AI models, improving on-chain efficiency 3 Top-level privacy protection: Build native private transfers, ending full-network exposure 90% of ETH traders don’t know these 5 key points hidden in the new roadmap: 1. Verkle tree development has taken years but is inherently incompatible with post-quantum encryption schemes. Instead of costly rework and reconstruction later, it’s better to replace it now, saving years of upgrade costs and proactively avoiding a long-term security pitfall. 2. Currently, ZEC, Monero, and Solana’s privacy ecosystems keep attracting capital, and enterprises hesitate to use Ethereum’s fully public chain for transfers. With built-in privacy at the base layer, funds and institutions won’t need to convert to privacy coins for large transfers, directly retaining massive liquidity. 3. In the future, smart contract security will rely entirely on AI automated audits, making it easier for ordinary retail developers to pass reviews. However, AI will filter out high-risk innovative code, making it harder for niche experimental DApps to launch. 4. This round of reconstruction is called Ethereum’s third major iteration. The first was the genesis, the second the merge, implemented slowly over 7 hard forks, each changing only a small part to prevent network paralysis. It’s unlikely to immediately boost ETH’s price in the short term; the benefit is a long-term fundamental improvement. 5. Major banks and asset managers have clearly stated they won’t allocate ETH at scale unless quantum risks are resolved. Prioritizing quantum resistance is essentially clearing the way for trillion-dollar traditional capital inflows. In summary: This is a fundamental overhaul of Ethereum’s base layer to meet security, institutional capital, and privacy needs for the next decade. It’s a long-term positive but with a very long implementation timeline, so it won’t significantly impact the market in the short term.

Calm before the storm? The well-established major exchange Crypto.com has been exposed for having many hidden risks in its accounts and operations, causing the market to worry whether it will become the next time bomb. Their core weapon to attract retail investors is the crypto debit card, which requires users to lock up their platform token CRO to receive high cashback and VIP benefits. The fatal weakness of this model is: once the market turns bearish or subsidies decline, users will dump CRO and withdraw funds, causing the platform's balance sheet to shrink instantly. If the exchange itself uses CRO as a major part of its reserves, it is very easy to fall into a death spiral similar to FTX and FTT. Additionally, they are known in the industry for being extremely willing to spend big on marketing. For example, they spent $700 million to buy the 20-year naming rights of the Los Angeles Staples Center and sponsored the World Cup and F1. However, in the history of the crypto industry, platforms that splurged on sports naming rights before bear markets or at bull market peaks, such as FTX sponsoring the Miami Heat arena and Miami racing, later faced severe cash flow backlash due to extremely high fixed costs. Even more critically, although Crypto.com is a compliant major US-listed company, its main operational entities are actually complexly structured and have long been dispersed across Singapore, Malta, and various offshore tax havens. While this structure avoided strict early-stage regulation, as global (especially US and European) crypto regulatory laws are gradually implemented, the compliance costs and fines required to obtain proper licenses are aggressively eroding the already modest net profits of these second-tier exchanges. Summary: 1. Remember to diversify the assets you hold there. 2. If it collapses, the market trend will inevitably be no less severe than the collapse of FTX.

This kind of play, doing a hundred-billion-level employee buyback before going public, is unprecedented in the history of technology. OpenAI paid $7 billion out of its own pocket to buy back the stocks held by its employees. Previously, they would find large external institutions like SoftBank and Thrive Capital to take over employee stocks, but this time they directly used their own cash reserves for the buyback. This transaction keeps OpenAI's valuation at $852 billion. They want to prove that the company has plenty of money on the books, and also to clean up the equity structure before officially listing on the US stock market, so no new external institutions come in to take a share. Why not bring in new external investors? The biggest fear before an IPO is a mixed equity structure. If new external institutions are recruited, not only would more seats need to be opened in listing negotiations, but it would also dilute the influence of existing old shareholders (like Microsoft, SoftBank, etc.) and management. Internal digestion by OpenAI protects the valuation and maintains strong absolute control. Behind the abundance of money is also the absurd wealth creation tug-of-war with competitor Anthropic. OpenAI has already created hundreds of cash millionaires with net worths of tens of millions of dollars before even going public. The competitor Anthropic just allowed employees to sell stocks to cash out through financing, but due to employees' reluctance to sell, institutions didn't buy enough. OpenAI must let employees cash out every six months to a year, allowing them to get tens of millions of dollars in cash, otherwise top talent would jump ship the next day. Summary: This hardcore "clearing + paying" strategy is clearly the final obstacle clearing for the upcoming largest IPO in history. Once Ultraman finishes going public, the big crash will come faster, and market liquidity will become even smaller

Last week, Bitmine bought another 7,391 ETH, already hoarding 4.8% of the total Ethereum supply, close to their 5% target. They staked 87% of their ETH holdings, earning $194 million in interest over a year. They also repurchased 3 million shares of their own stock, BMNR. They are even smarter than MicroStrategy, turning capital appreciation into interest-bearing assets; this interest is pure profit. In traditional finance, a single institution holding more than 5% of shares signals a takeover and absolute control. 5% of Ethereum’s total supply effectively locks up a huge amount of market liquidity. Moreover, staking 87% means less Ethereum is available for spot trading in the market. Once a major bull market arrives, this extreme liquidity tightening will trigger a strong squeeze and explosive rally effect. Why do they buy coins while also repurchasing their own stock BMNR? Because in the US stock market, many institutional investors are restricted by compliance from directly buying ETH spot; they can only buy proxy shares like BMNR. When market sentiment is poor, BMNR’s market value may be lower than the actual value of the ETH it holds (a discount). Tom Lee is extremely skilled at this kind of Wall Street capital operation, aggressively repurchasing his own stock when the price is undervalued, raising the ETH amount per share. When the market catches on, both the stock and the coin will experience a double Davis double tap.

Saylor is once again selling Bitcoin to save his preferred shares STRC. Does this trick still work? He just sold 1,690 BTC, using the entire $108.6 million to buy back STRC. The average selling price was $64,000, far below his holding cost of $75,000, selling at a pure loss. STRC should have a $100 face value, but now it's at $95, and even a 12% annual dividend can't pull it back. The awkward part is he just sold a batch last week, net selling for two consecutive weeks. However, cash reserves have reached $4.65 billion. His current essence is relying on selling Bitcoin at a loss to fill the preferred shares' gap. It seems well-funded, but the average cost of the 840,000 BTC held is $75,385. Selling Bitcoin below cost to support the stock can be understood as using loss-making assets to subsidize high interest/buybacks. If Bitcoin prices remain sluggish long-term, this strategy of selling BTC to maintain dividends will continuously erode his holding strength. The more he sells, the more he loses; the fewer Bitcoins he holds, the more STRC will fall, a double blow. Wall Street wolves are waiting to feast on this fat piece of meat.

Who controls Bitcoin? Most Bitcoin Core developers are technical purists, who believe Bitcoin should only serve as electronic cash or digital gold, and that filling the chain with NFTs and inscriptions pollutes the ledger. But for miners who invest heavily in mining machines and pay electricity bills, as block rewards decrease after Bitcoin halving, high transaction fees become the lifeline for miners to stay profitable. Miners cannot sacrifice themselves for the developers' idealism. Therefore, the Bitcoin BIP-110 proposal got stuck. Why can't developers force upgrades like Ethereum? This precisely demonstrates Bitcoin's most terrifying decentralized charm. Ethereum can change its roadmap whenever Vitalik and the Foundation decide. So, some in the community joke that if Vitalik is taken care of, ETH might be finished, and institutions are advised to hire bodyguards for him. In the Bitcoin world, developers only write code; whether the code takes effect depends entirely on miners' (hash power) votes. During the famous 2017 Hong Kong consensus and block size battle, miners and developers clashed fiercely, ultimately proving that upgrade proposals (BIPs) without miner support are basically just empty talk. It's said that wherever there is profit, undercurrents flow; miners privately engage in MEV accelerators. Even if the official client (Core) tries to filter out transactions packaging inscriptions at the node level, many large mining pools have long started using custom clients or directly accept users' private packaging requests. No matter how strict the on-chain rules are, as long as someone is willing to pay miners privately, miners will include that data in blocks. The consequence is that Bitcoin will shift from an open and fair rule-based society to a shadow society where whoever has more money can take shortcuts. Big money privately paying miners can get priority inclusion, while ordinary people cannot get in no matter how much normal fees they pay. Miners, to earn under-the-table money, gain the privilege to package whoever they want, tearing apart Bitcoin's anti-censorship and decentralization foundations. Moreover, only large mining pools with resources can take such private jobs to earn huge profits; retail and small miners can't compete, leading to increasing concentration of hash power. Thus, Bitcoin's decentralization ceases to exist.

The skills on your body all grew during times of economic hardship. The flaws on your body all developed during times of comfort and plenty. Skills grow through suffering, flaws multiply during ease. If you don't want to repeat the cycle, you must reverse it. Only those who have truly experienced it can understand this saying.

Japan's four major life insurers' losses on US Treasury bonds have surged to $96 billion, with quarterly losses expanding by 7%. Now the Bank of Japan is caught in a dilemma: raising interest rates can save the yen and curb inflation, but it will continue to enlarge the book losses on domestic bonds; if it doesn't raise rates, the yen will keep depreciating. The awkward situation is that Japan holds $1.14 trillion in US Treasuries, and if forced to sell, it would push up US Treasury yields, severely damaging US financing. Japan has become the largest overseas creditor of the US. Previously, when the US issued debt recklessly, Japan would continuously help pay for it, but now Japan is struggling to manage itself and can no longer blindly backstop US debt like before. When the world's largest buyer starts to withdraw, the US has no choice but to have the Federal Reserve print money to buy its own debt. Key point regarding the relationship with crypto: Short term is bearish, long term is the ultimate bullish. Short term: Yen liquidity tightening will cause a brief bleed and decline in Bitcoin. Long term: It thoroughly exposes the unsustainability of the traditional fiat credit system (which relies on printing money and borrowing to survive). When investors find that even risk-free assets like government bonds start to lose value and depreciate, assets like Bitcoin, which have a fixed total supply and cannot be arbitrarily printed or confiscated by any government, will become a global hedge asset for large funds.

Why is it so hard for children of poor families to turn their lives around? Let me use myself as an example. Only in recent years have I somewhat understood the objective society. The previous thirty or twenty years were basically wasted. When I was young, my mind was filled with the perceptions of them (the older generation), and this is why it's so hard for poor people to make progress. It’s possible that you spend most of your life just to overcome the wrong perceptions your original family gave you. When you finally overcome them, you realize you are no longer young. Because many things are tools, tools that don’t let you leave the zoo. Suddenly, one day, you come out of the fish tank. You see, this moment of realization is what truly means changing your fate. Damn, it took me over 40 years. It’s not about how much money you make today to change your fate, It’s that your thoughts and mindset are completely different. This is what they call enlightenment. You begin to understand the rules of society and the nature of humanity, And when you can see your future self, that’s enlightenment. Being a person who has awakened is the most painful. What is this like? It’s a bit like The Truman Show, a bit like The Shawshank Redemption. Only when you suddenly figure out what you should do, You look at those people you once cared about, they are still living in Truman’s world, they are still there. At this point, making money is a result and proof. But you also need to know, When you truly reach the income level you imagined, you will be lonely. This loneliness is real, not acted. So you have to choose, what to choose? To gain something, you must give something up. Do you know what you have to give up? Let me use myself as an example: I was originally an ordinary person, with no qualification to be myself, which is quite difficult. Looking at history, very few people truly have the qualification to be themselves, because the cost is too high and it also depends on fate. But when you soberly realize you don’t have the qualification to be yourself, You can do business, you can make money. Do you understand? Because you are not coming out as yourself, From the bottom of your bones, you truly accept that coming out means selling yourself. Why do some people want both? Because they haven’t completely sold themselves yet. They want to prove who they are: I’m Zhang San, I’m a certain director, I’m someone’s son, I’m superior. Actually, you still haven’t let go of yourself, You still think you are important, you haven’t fully entered the world of making money. Many people put on airs and pretend, So naturally, there are those who earn less than 100,000 a year but think 1 million is small money. So there are too many people and things that are handled superficially. Naturally, there are those who think they are awesome, and others just can’t stand them, Especially those from the bottom who manage to get money, people say they’re just lucky or shameless. Isn’t that so? You must think clearly, what do we really want when we come out to hustle? Back to the beginning, poor people, because they can’t accept that they have come out as themselves, fully letting go of the self, There are such people, but very few, and those who can do it succeed. To make money, you have to forget your original self, let go of moral scruples, and face your former self head-on.