Post

500亿到账了，但亚马逊可能拿不回一分钱兄弟姐妹们，说个事。 7月31日，亚马逊在10-Q里低调地披露了一行字：已完成对OpenAI总计500亿美元的全额投资。 500亿。不是5亿，不是50亿。500亿美元。什么概念？够买下两个Coinbase还有剩。但这个故事没那么简单。钱怎么给的？一季度先投150亿，二季度再投137亿，报告期后补上213亿。三笔钱，分三个阶段，全部到位。而且注意——原定的触发条件根本没达成。原本说好的“OpenAI完成IPO”或“实现重大AI突破”，一个都没发生。亚马逊还是把钱打过去了。为什么？因为今年4月，OpenAI和微软重新谈了云合同。微软之前几乎垄断OpenAI的云服务，甚至考虑过起诉亚马逊。合同一重谈，AWS终于能进场了。 500亿到账，换来的是AWS给OpenAI做云服务的机会。亚马逊拿的是C轮优先股，持股约5%。优先股听起来高级是吧？要等OpenAI完成IPO或其他流动性事件，才能转换成普通股。 IPO完了还有惯例锁定期，不能马上卖。也就是说，这500亿现在就是一张“欠条”。 OpenAI一天不上市，亚马逊一天拿不到流动性。问题来了：OpenAI什么时候上市？原本说今年年底。6月已经秘密递交了IPO申请。但最新消息是——可能要推迟到明年了。知情人士说，“上市时机尚早”，理由是“不少战略目标，作为非上市公司推进更加便利”。翻译一下：我们现在上市可能会露馅。露什么馅？ OpenAI 2026年第一季度营收57亿美元。听起来还行是吧？同期现金消耗37亿美元。非GAAP亏损69亿美元。每赚1美元，亏1.22美元。去年全年净亏损385亿美元，是2024年的8倍。而OpenAI承诺未来八年要在AWS花1000亿美元。投出去500亿，换回1000亿的云订单。账面看赚了。但问题是——一个每赚1美元亏1.22美元的公司，拿什么来付这1000亿？更扎心的是整个AI圈的现状。四大巨头——微软、谷歌、亚马逊、Meta——2026年合计资本开支逼近7500亿美元。比去年猛增超过60%。而整个生成式AI行业的直接营收呢？根本填不上这个窟窿。有分析说，这种“投资—消费闭环”就是在放大AI繁荣叙事。你投我，我租你的云，你记收入，我记支出——钱转了一圈，又回到了起点。那亚马逊这笔500亿，到底是押注还是泡沫？押注派说：八年1000亿云订单，Trainium芯片锁定2吉瓦产能，AWS第二季营收年增36.7%——值！泡沫派说：OpenAI还在巨亏，IPO遥遥无期，优先股变不了现，500亿就是一张“远期彩票”。我的看法？答案不在投了多少，在1000亿能不能兑现成真金白银的收入。如果OpenAI的算力需求是真实的、可持续的，这500亿就是历史上最聪明的战略投资之一。如果AI泡沫破裂，OpenAI的云订单缩水—— 那这500亿，就是史上最贵的“云会员费”。 $OPENAI $AMZN $XAMZN #亚马逊向OpenAI投500亿美元：押注还是泡沫

50 billion spent, exchanged for a 100 billion order—Is Amazon investing or just buying itself a big client? Amazon has poured 50 billion dollars, all to OpenAI. This is not a rumor, not an intention, it's real money transferred. 15 billion in Q1, 13.7 billion in Q2, and another 21.3 billion before July 31. Three payments, each faster than the last. But the slickest move is here— OpenAI promises: over the next eight years, it will spend 100 billion dollars on AWS. Invest 50 billion, get back 100 billion in cloud orders. Mathematically doubled, but logically? Picture this: Amazon paid 50 billion to OpenAI and got about 5% equity. By this calculation, OpenAI is valued at 852 billion dollars. Then OpenAI turns around and says: "Okay, I accept this 50 billion, and over the next eight years, I will spend 100 billion on AWS." Isn't this equivalent to Amazon placing a 100 billion procurement order to itself? Money moves from left hand to right hand, circles around and comes back—only with OpenAI acting as a "channel" in between. Some say: this is a strategic bet, a must-have position in the AI track. Let me tell you— If it were truly a strategic bet, why take preferred shares instead of common shares? What do preferred shares mean? OpenAI hasn't IPOed, no liquidity event—this 50 billion is just a "debt," not equity. You invest 50 billion, company valued at 852 billion, sounds glamorous. But since they don't go public, your preferred shares never convert to common shares, and you can never exit. Even more cunning—the originally planned 35 billion final payment was supposed to be made only after "OpenAI completes IPO or achieves a major AI technology breakthrough." But neither condition was met, and Amazon transferred all the money early. What does it mean to be in a hurry? This is being in a hurry. The most painful part? Amazon didn't just bet on OpenAI alone. It is also a core investor in Anthropic—already invested 18 billion, with a total commitment up to 33 billion. Left hand OpenAI, right hand Anthropic. Betting on both sides, whichever wins is fine. But this also means—neither is a true "bet." This is a hedging strategy. AWS's Bedrock platform now hosts GPT-5.6, Claude Opus 5, Gemma 4, and Grok 4.3 simultaneously. Who wins doesn't matter, AWS winning is what matters. Another data point: Amazon raised its 2026 capital expenditure guidance to 220 billion dollars. Q2 AWS revenue was 42.2 billion, up 37%. AI and chip businesses both have annualized revenues exceeding 25 billion. Everything looks great on paper. But Citibank's report clearly states: the 100 billion cloud order promised by OpenAI is expected to consume 2 gigawatts of Trainium 3 and 4 generation capacity. What does 2 gigawatts mean? Equivalent to the power of two nuclear power units, all used to run OpenAI's models. If OpenAI's "1 billion users" don't turn into "1 billion in revenue"—this 100 billion order is just an empty check. Finally, to be blunt: This deal, in essence, is— Amazon using 50 billion to buy itself a 100 billion "big client." But the money from this "big client" is given by Amazon itself. Ordering from itself, paying itself, counting revenue from itself. In accounting, this is called a "related party transaction." In business, this is called "fighting with oneself." In the AI track, this is called—"No matter how big the bubble, I’ll take a spot first." OpenAI valued at 852 billion, 50 billion invested for only 5%. Do you think this is a bet or a bubble? $AMZN $OPENAI $BTC #亚马逊向OpenAI投500亿美元：押注还是泡沫