What are the benefits of the three tech giants "transferring" 110 billion to OpenAI?

Author: Su Yang

On February 27th, OpenAI announced the completion of a new $110 billion funding round, valuing the company at $730 billion pre-money.

This is OpenAI's second record-breaking fundraising round, following its $41 billion funding round last year, and is the largest private technology company funding round to date.

This round of funding was led by strategic investors, with Amazon investing $50 billion, and SoftBank and Nvidia each investing $30 billion.

Following this round of financing, OpenAI's cash reserves have increased to approximately $150 billion. OpenAI stated that the funds will primarily be used to expand its computing infrastructure to meet rapidly growing user demand. Amazon to Invest $50 Billion in Phases As the most generous investor in this round of financing, Amazon's committed $50 billion investment will be implemented in two phases: the first phase of $15 billion has been confirmed, and the remaining $35 billion will be available after OpenAI completes its IPO or achieves AGI. Therefore, Altman quickly stated that he is open to going public at the appropriate time. Regarding AGI, according to the previous contract between OpenAI and Microsoft, AGI refers to "a highly autonomous system capable of surpassing humans in the most economically valuable work." OpenAI's charter includes a special clause: once AGI is achieved, Microsoft will lose access to its technology. But Amazon isn't just providing financial support; the two companies have also signed a comprehensive technology cooperation agreement—Amazon will exchange its own chips and cloud services for OpenAI's models and technology. The core clause of this cooperation is the purchase of computing power. OpenAI will pay AWS an additional $100 billion in computing power over the next eight years, on top of its existing $38 billion contract. A significant portion of this $100 billion will be used on Amazon's self-developed Trainium chips. OpenAI has committed to purchasing approximately 2 gigawatts of Trainium computing power, including the upcoming Trainium 3 and the Trainium 4, launching in 2027. Amazon CEO Andy Garcetti specifically mentioned that Trainium chips offer 30% to 40% better performance than comparable GPUs, and "now both of our leading AI labs are using them." Another company is Anthropic, which has recently been embroiled in a dispute with the US government. OpenAI's other collaboration with Amazon is at the product level; the two companies will jointly develop a "Stateful Runtime Environment" to run on Amazon Bedrock. This environment allows developers to maintain context, remember previous work, and collaborate across tools when calling OpenAI models. AWS has also secured exclusive third-party distribution rights for OpenAI's enterprise platform, Frontier. There's also a more direct layer of collaboration: Amazon will use OpenAI models to optimize its consumer applications. The two companies will jointly develop custom models for use by Amazon's internal teams, covering product lines from e-commerce to smart hardware. Nvidia's $30 Billion Investment: Shareholder + Supplier The collaboration between Nvidia and OpenAI was once rumored to have fallen through, especially after Jensen Huang's statement that investing $100 billion in OpenAI was never a promise. However, the investment was ultimately finalized, albeit at a significantly lower amount than the rumored $100 billion, with only $30 billion to be paid in three installments. According to the announcement, the two companies finalized a new computing power partnership: Nvidia will provide OpenAI with 3 gigawatts of dedicated inference computing power and 2 gigawatts of training computing power for the Vera Rubin system—on top of the Hopper and Blackwell systems already running on Microsoft, Oracle Cloud, and CoreWeave. If we convert this to computing power, 1GW corresponds to $50 billion. Nvidia's support for OpenAI is worth far more than the rumored $100 billion. From this perspective, Nvidia is not only a shareholder of OpenAI and its largest chip supplier, but it's also using its own money to help OpenAI buy its own chips. Most of the $30 billion in funding will ultimately return to Nvidia in the form of purchasing computing power. In short, it's still the same "circular financing" narrative. SoftBank's $30 billion investment: IPO is more important than anything else. Like Nvidia, SoftBank's $30 billion investment commitment will also be disbursed in three installments. Masayoshi Son's thinking is relatively simple: OpenAI is currently the most likely candidate for an IPO in the AI ​​sector, and SoftBank needs an IPO to boost the performance of its Vision Fund. Sources say this round of financing is seen as a crucial step before OpenAI's IPO, which could begin as early as the end of this year. Compared to the other two investors, SoftBank's other role is that of a matchmaker. According to sources, OpenAI is expected to secure approximately $10 billion in primary equity financing, with commitments finalized within the next month. Investors include sovereign wealth funds and investment firms. It is speculated that these investors may have entered through SoftBank. Microsoft: No Funding This Round, But Position Remains Unchanged. In response to Amazon's entry into the market, Microsoft and OpenAI issued a joint statement today, emphasizing that their partnership remains unaffected. The statement says that Microsoft remains the exclusive cloud provider for OpenAI's stateless API for its models. All API calls to OpenAI models that are stateless (without retaining context) are hosted on Azure. This is currently the most common way to call OpenAI models. Meanwhile, OpenAI's own products (including the newly released Frontier) will continue to be hosted on Azure. Frontier is OpenAI's newly launched enterprise platform, enabling enterprises to deploy their own AI agents. The intellectual property licensing and revenue-sharing arrangements between Microsoft and OpenAI remain unchanged. However, OpenAI has the right to purchase computing power from elsewhere, including large-scale infrastructure projects like Stargate. The “Recurring Financing” Narrative: A Copy-Paste Approach Looking at the investments and returns of the aforementioned investors together, as mentioned earlier, it still follows the “recurring financing” narrative: Amazon invested $50 billion in exchange for OpenAI’s model technology licensing and a $100 billion cloud service order. Nvidia invested $30 billion in exchange for OpenAI’s chip procurement commitment. SoftBank invested $30 billion, betting on an IPO exit opportunity. Microsoft didn’t invest in this round, but retained its API exclusivity; Microsoft will receive a share of every API call OpenAI sells in the future. Following this round of financing, OpenAI has an additional $110 billion in its accounts and holds the computing power resources of three tech giants. The **cost** is that it has pre-locked future technology licensing and cloud service procurement to its shareholders. Moreover, Amazon and SoftBank have made an OpenAI IPO a key condition for fulfilling their investment commitments. Now, OpenAI's shareholder list includes Microsoft, Amazon, Nvidia, and SoftBank—four giants with overlapping and conflicting businesses. Balancing the interests of these shareholders will be a major challenge for Sam Altman. ChatGPT has over 900 million weekly active users. Returning to the business level, OpenAI certainly needs this money right now. OpenAI already had approximately $40 billion in cash on hand, and with this round of financing, its available funds have increased to approximately $150 billion. The company expects to achieve positive free cash flow for the first time by 2030. Until then, most of the funds will be invested in data center operations, chip procurement, and cloud service expenditures. Financial data cited by the Financial Times, based on sources familiar with the matter, is telling: Revenue is projected to be approximately $13 billion in 2025, reaching $30 billion this year, with a target of over $60 billion in 2027 and over $280 billion in 2030. The growth curve is steep, but costs are increasing even faster. Regarding user data: According to official data released by OpenAI, ChatGPT currently has over 900 million weekly active users, over 50 million consumer subscribers, and over 9 million paying business users. Codex, a programming tool launched this year, has seen its weekly active users triple since the beginning of the year, reaching 1.6 million. While this data growth is impressive, competitors are not idle. Google's Gemini is making steady progress in the consumer market, while Anthropic has already taken the lead in the enterprise market. OpenAI executives stated that by the end of this year, half of their revenue will come from enterprise customers, up from the current 40%. This means that in the coming quarters, OpenAI must directly compete with Anthropic to achieve a 10% increase. Another detail worth noting is the change in the equity structure. Following this round of financing, the value of the for-profit business shares held by the OpenAI Foundation, a non-profit organization spun off from OpenAI, will rise to over $180 billion by 2025. Sources familiar with the matter say that if investor demand is strong, the foundation may sell up to $10 billion in stock for charitable expenditures and recruitment. In other words, after this round of 110 billion yuan in financing, there may still be another wave of OpenAI stocks available for purchase in the private equity secondary market.