I just finished watching the latest episode of All In Podcast (280), where several experts talked about SpaceX's IPO samples, OpenAI/Anthropic's IPO progress, the real returns of AI Token burning money, and the crazy data of Trump's account going online. There is a lot of information, so let's organize it SpaceX's IPO has set a template for future giants to go public SpaceX raised $75 billion at a valuation of $1.75 trillion, and after going public, it once soared to $200 per share. Currently, it is stable at around $150, corresponding to a market value of approximately $2 trillion and forward revenue of $35 billion, ranking as the seventh largest publicly traded company in the world. The consensus among everyone is that SpaceX's phased unlocking and early inclusion in the index is equivalent to setting a textbook level model for the subsequent listing of these super unicorns. Anthropic has secretly submitted its listing application, and there are reports that its annualized revenue is sprinting towards $100 billion. If it really goes public, its market value may reach $3 trillion. OpenAI's revenue has also rebounded to approximately $70 billion, and it is said that GPT-6 will be released within 30 days. However, due to the complex restructuring of the company and the fast burning of money before, the listing progress may be slower than Anthropic's. Brad mentioned that as a large institution, their company would still subscribe to the IPOs of these two companies on a large scale even at a valuation of 3 trillion yuan, because their compound annual growth rate of revenue in the coming years can still exceed 30%. The consumption of AI tokens is exploding, but the true returns on the enterprise side are being questioned This discussion is quite sharp. Chamath mentioned that his company's CTO reported that token spending doubles every 45 days, but the downstream productivity improvement it brings is only about 5%. The more heart wrenching data is that only 0% to 2% of the earnings per share growth of the S&P 500 (excluding Nvidia) is truly driven by AI. If companies cannot prove that token consumption truly brings investment returns, this consumption frenzy may face a correction now. However, several companies have also provided relatively clever response samples. Uber already has 99% of its engineers using AI to assist in coding, and has specifically established an "intelligent agent team" to delve into HR, legal, and marketing departments, refining cost and efficiency calculations. The CTO of DoorDash has developed a "model routing" strategy, which assigns simple and low-level tasks to cheap open-source models, while only the most difficult core tasks are handled by Anthropic's top-level models, successfully reducing costs. The battle between closed source big models and open-source/sovereign AI is becoming increasingly fierce Brad insists that the advantages of closed source top-level models still exist, and the economic value is still highly concentrated in OpenAI and Anthropic, the two top laboratories. For enterprises, using a slightly more expensive top-level model to replace a senior engineer or consultant with an hourly wage of $200 is a rational choice to spend more than a dozen dollars. But after Chamath joined the United Nations AI Committee, he observed another line: countries such as the United Arab Emirates, Saudi Arabia, and Japan are unwilling to completely hand over their technological lifeline to the closed source model of the United States, and are desperately building independent and controllable open-source "sovereign AI" technology stacks. At the same time, Zuckerberg is not idle on his side. Meta's newly released Muse Spark 1.1 model, which focuses on the same quality but only costs 1%, has directly launched a price war, attempting to use open source and low price to break through the high price barrier of closed source models. There is news from China that top-level models may be restricted from going global It is reported that Chinese regulators are holding closed meetings with Tongyi Qianwen, ByteDance, and Zhipu to consider limiting overseas access to top-level open source and closed source models, in order to prevent technology disclosure and use by the United States. Sacks' interpretation of this is that it is actually a typical strategy of "switching ecosystems through open source during the catch-up period, and immediately turning to closed source once catching up or approaching the forefront level", which is somewhat similar to OpenAI's own path transformation back then. The current core consensus on the US side (including the White House and the Treasury Department) is to "maintain AI leadership in China at all costs". Due to the discovery that some Chinese models have distilled a large amount of output from American models, the US side is also simultaneously launching anti distillation countermeasures. Trump account officially launched, data frenzy is a bit outrageous According to the US Investment Act, the Trump account was officially launched on July 4th, the 250th anniversary of the founding of the country, and the official app directly topped the US App Store download chart. The core mechanism is that every child born in the United States will automatically receive a lifetime tax-free investment account opened by the government, with an initial injection of $1000, and the funds will be directly invested in the S&P 500 index without any management fees for life. I opened 1.5 million accounts and deposited over 1 billion US dollars 24 hours before going live. The tax incentives are also designed very aggressively, with relatives and friends being able to deposit up to $5000 per year for their children and enjoy 18 years of tax-free compound interest; Even more ruthless is that employers can inject up to $2500 annually into employees' children tax-free, which is directly deducted from the taxable income of both employers and employees, and is even more powerful than traditional 401 (k) and IRA. After the child reaches the age of 18, this money can be seamlessly deposited into a Roth IRA, which is commonly referred to as a "historic wealth inheritance and tax avoidance tool" in the CPA community. If you start saving money every year from the age of 0, you can easily reach millions of dollars by the age of 18 to 28 with compound interest. Charitable donations are also exaggerated: Brad himself donated $100 million, the Dell couple anchored a donation of $6 billion to support the children of 25 million low - and middle-income families, the SpaceX CEO donated $350 million worth of SpaceX stock (equivalent to allowing each low-income child to directly hold SpaceX equity), and Micron also donated $250 million to pair employee children. It is widely believed that this mechanism completely bypasses the traditional NGO model of intermediaries taking multiple layers of commission and low efficiency, and funds are directly transferred to personal accounts. It is expected that over the next 10 years, more than 100 million accounts will be opened and assets of 2 to 4 trillion US dollars will be injected, which can increase the holding rate of the US stock market from about 50% to over 75%, to some extent turning all citizens into direct beneficiaries of the capital market.