Post

50 billion has arrived, but Amazon may not get a penny back Brothers and sisters, let me tell you something. On July 31, Amazon quietly disclosed a line in its 10-Q: it has completed a total investment of 50 billion USD in OpenAI. 50 billion. Not 5 billion, not 50 billion. 50 billion USD. What does that mean? Enough to buy two Coinbase and still have some left. But the story is not that simple. How was the money given? 15 billion was invested in Q1, 13.7 billion in Q2, and 21.3 billion after the reporting period. Three payments, three stages, all in place. And note— the originally set trigger conditions were not met at all. The originally agreed "OpenAI completes IPO" or "achieves major AI breakthrough" did not happen. Amazon still transferred the money. Why? Because in April this year, OpenAI and Microsoft renegotiated the cloud contract. Microsoft previously almost monopolized OpenAI's cloud services and even considered suing Amazon. With the contract renegotiated, AWS finally got a chance to enter. 50 billion arrived, in exchange for AWS getting the opportunity to provide cloud services to OpenAI. Amazon holds Series C preferred shares, about 5% ownership. Preferred shares sound fancy, right? They can only be converted into common stock after OpenAI completes an IPO or other liquidity events. After the IPO, there is also a customary lock-up period, so shares cannot be sold immediately. In other words, this 50 billion is currently just an "IOU." As long as OpenAI doesn't go public, Amazon won't get liquidity. The question is: when will OpenAI go public? Originally said by the end of this year. The IPO application was secretly submitted in June. But the latest news is—it may be postponed to next year. Insiders say, "The timing for going public is premature," reasoning that "many strategic goals are easier to advance as a private company." Translation: Going public now might expose some secrets. Expose what secrets? OpenAI's revenue in Q1 2026 is 5.7 billion USD. Sounds okay, right? Cash burn in the same period is 3.7 billion USD. Non-GAAP loss is 6.9 billion USD. For every 1 dollar earned, 1.22 dollars are lost. Last year's net loss was 38.5 billion USD, eight times that of 2024. And OpenAI promises to spend 100 billion USD on AWS over the next eight years. Invested 50 billion, in return for 100 billion cloud orders. On paper, it looks like a profit. But the problem is—what does a company that loses 1.22 dollars for every 1 dollar earned use to pay that 100 billion? Even more painful is the current state of the entire AI circle. The four giants—Microsoft, Google, Amazon, Meta—will have combined capital expenditures approaching 750 billion USD in 2026. An increase of over 60% compared to last year. And the direct revenue of the entire generative AI industry? It simply cannot fill this gap. Some analyses say this kind of "investment-consumption loop" is amplifying the AI boom narrative. You invest in me, I rent your cloud, you record revenue, I record expenses—the money goes in a circle and returns to the starting point. So is Amazon's 50 billion a bet or a bubble? The bet side says: 100 billion cloud orders over eight years, Trainium chips locking 2 gigawatts of capacity, AWS Q2 revenue up 36.7% year-over-year—worth it! The bubble side says: OpenAI is still losing heavily, IPO is far away, preferred shares cannot be cashed out, 50 billion is just a "long-term lottery ticket." My view? The answer is not how much was invested, but whether the 100 billion can be realized as real cash income. If OpenAI's computing power demand is real and sustainable, this 50 billion is one of the smartest strategic investments in history. If the AI bubble bursts and OpenAI's cloud orders shrink— then this 50 billion is the most expensive "cloud membership fee" in history. $OPENAI $AMZN $XAMZN #亚马逊向OpenAI投500亿美元：押注还是泡沫