SpaceX와 기타 빅테크 기업들이 AI 칩 구매에 수십억 달러를 쓰고 있다고, xAI의 Colossus와 관련된 보도들이 전했다.

Major players in the artificial intelligence (AI) field are collectively seeking tens of billions of dollars in funding to acquire computing hardware, amid a race to build AI data centers that is driving up capital demand.

SpaceX is raising $40 billion to acquire Nvidia chips. Billionaire Elon Musk's aerospace company plans to raise approximately $10 billion through bank loans and $30 billion through investment-grade bonds, Reuters reported on October 7. The deal is still under negotiation and has not been officially announced.

Apollo Global Management is reportedly set to lead the deal and assist SpaceX in distributing the debt to multiple investors, according to a source who confirmed this to the Financial Times. SpaceX has also been in discussions with PIMCO, one of the world's largest bond managers, about financing options.

According to CNBC, SpaceX will likely use Nvidia GPUs as collateral for the loan. This is a noteworthy move as financial institutions increasingly seek to turn AI computing infrastructure into an asset class that can raise capital.

However, using GPUs as collateral remains a subject of debate. Nvidia argues that its computing systems can retain their value over the long term, but lenders are more cautious and typically apply a depreciation period of around three to four years for this type of asset, due to concerns about the technology quickly becoming obsolete.

SpaceX's plan comes as demand for AI chips surges. Morgan Stanley estimates that global AI infrastructure could require up to $1.5 trillion in external funding by 2028.

The $40 billion funding will help SpaceX significantly expand its computing power to support Musk's AI ambitions. Last month, Musk said that xAI's Colossus 2 data center could more than double the number of Nvidia chips it uses by December. SpaceX also plans to use Nvidia hardware as the foundation for its data centers.

SpaceX has now been able to raise a significant amount of capital following its IPO in June, a deal described by Reuters as the largest IPO in US history at approximately $86 billion. However, continued reliance on debt shows that the cost of expanding AI infrastructure is becoming a major financial challenge even for companies with strong fundraising capabilities.

News of SpaceX's plan to raise $40 billion caused its stock to fall about 2% during the October 7 trading session, while Nvidia's stock also dropped 0.7% at the close.

**The AI race requires enormous capital investment.**

SpaceX's borrowing to purchase GPUs also reflects a broader trend. In August, Nvidia partnered with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to build financing platforms aimed at raising over $500 billion in external capital for AI infrastructure projects. Nvidia wants to transform computing power and the entire AI infrastructure into an asset class that can attract long-term capital flows.

Over the past few weeks, Broadcom has also sought to secure more than $50 billion in funding for OpenAI to purchase custom AI chips jointly developed by the two companies, according to people familiar with the negotiations. Apollo and Blackstone are two of the lenders with whom Broadcom has been in discussions about the possibility of participating in the deal. Negotiations are still in their early stages, and the size of the financing is subject to change.

In a separate deal, Oracle is in negotiations with Apollo and Goldman Sachs to arrange funding for a large-scale chip purchase plan.

This wave of fundraising demonstrates how rapidly the cost of building AI infrastructure is increasing. Previously, cloud service providers such as Amazon Web Services and Oracle typically used their own cash flow to invest in computing hardware.

However, to expand their AI infrastructure, businesses have issued hundreds of billions of dollars in bonds, pushing the public debt market to its limits. Now, some buyers are turning to Wall Street finance firms to raise tens of billions of dollars for individual acquisitions, according to the WSJ.

A new group of customers buying the new chips is also emerging, including OpenAI and Anthropic. Leading AI labs that previously relied heavily on cloud computing power from providers are now seeking to own more of their own infrastructure to reduce costs and reliance on partners.

The new funding that Broadcom is arranging for OpenAI could include multi-gigawatt chip operating capacity, according to a source. The deal is expected to be finalized before the end of the year.

OpenAI's chip development program is internally codenamed Nexus. The program's customized chip generations are named after types of chili peppers, with the first and second generations named Jalapeño and Serrano, respectively.

This deal comes about a year after OpenAI and Broadcom announced a collaboration to develop a 10-gigawatt custom chip system, incorporating Broadcom's networking technology. The two companies stated at the time that the systems would be deployed from the second half of 2026 through the end of 2029.

**Oracle wants to reduce debt pressure as it expands its AI capabilities.**

Oracle is also pushing to finalize the deal this year and is still in negotiations with several potential financing partners. The funding aims to address the cash flow gap between when Oracle has to pay for hardware and when revenue from its cloud computing services begins to be recognized.

It is unclear how many chips Oracle intends to raise. However, the amount of Nvidia chips needed for a 1-gigawatt data center alone could cost tens of billions of dollars.

The proposed solution involves investors injecting capital into a separate company that would purchase chips and then lease them back to Oracle for a specified period. This approach would allow Oracle to avoid borrowing excessively directly, thereby controlling debt costs while competing with more financially powerful technology corporations.

These deals demonstrate that the AI race is expanding from a competition for chips and computing power to a race to raise capital. As the cost of building data centers intensifies, tech companies are seeking new financing structures to continue investing without burdening their balance sheets.