OpenAI's Frenzied Orders for Broadcom and Oracle: Is the Cash Enough?

TradingKey - On Tuesday, cloud computing giant Oracle reported its earnings. Despite falling short of revenue expectations, a substantial cloud infrastructure order from OpenAI sent the stock soaring nearly 30% in after-hours trading, marking its best single-day performance since the dot-com bubble.

Similarly, Broadcom previously experienced a more than 4% jump after securing a multi-billion-dollar custom AI chip order from OpenAI, prompting it to significantly raise future AI revenue forecasts.

OpenAI's business partners are almost omnipresent in Silicon Valley. This year, Google Cloud partnered with OpenAI to provide computational support for its model training. Earlier, OpenAI signed a multi-billion-dollar compute agreement with cloud service provider CoreWeave. This aligns with OpenAI CEO Sam Altman's assertion that OpenAI is likely the most capital-intensive startup ever.

The backbone of OpenAI's robust financial backing is its aggressive capital expenditure plans. Recently, OpenAI disclosed its latest financial projections, anticipating cumulative cash outflows of up to $115 billion from this year through 2029 — an increase of about $80 billion from projections made six months ago.

Where's the Money Going? AI is the Main Focus

Altman stated that in the "not-too-distant future," OpenAI plans to invest trillions into AI infrastructure, including chips, data centers, and power systems that support AI operations, as AI is deemed the most crucial element in the long term. He acknowledges that the current AI investment frenzy is reminiscent of the late 1990s internet bubble, but he remains confident in AI's genuine and profound long-term value.

According to Altman, funds will primarily be allocated to AI training, inferencing, building server facilities, and talent acquisition. The bottomless pit of expenditure is AI model training, with OpenAI expecting to spend over $9 billion on it this year, $2 billion more than originally predicted, and around $19 billion next year. Training costs will continue to rise steadily through 2030 and potentially beyond, due to the inherently uncertain nature of AI training, where any failure and retry can directly inflate costs.

Another reason for OpenAI's substantial outlays is its status as one of the world's largest cloud server tenants. To manage these costs, OpenAI is developing its own data center server chips and associated facilities, with projected expenditures potentially exceeding $8 billion this year.

Where’s the Money Coming From? ChatGPT Remains the Cash Cow

OpenAI's data reveals that revenue growth driven by ChatGPT is outpacing expectations from six months ago. The company now anticipates generating nearly $10 billion in revenue through ChatGPT this year, contributing to a total revenue of $13 billion, demonstrating accelerated commercialization. By 2030, OpenAI has raised its revenue forecast by 15% to $200 billion, with ChatGPT revenues projected to reach $90 billion.

Beyond paid users, OpenAI is also exploring the commercial potential of a vast free user base, potentially monetizing through affiliate fees or advertising, with gross margins for such products expected to rival social media platforms like Facebook, at an impressive 80%-85%. OpenAI expects to generate approximately $110 billion in revenue from this user segment between 2026 and 2030.

Some analysts question the sustainability of OpenAI's cost-intensive, aggressive investment strategy. However, this approach is essentially a reflection of the current "arms race" in AI, where other tech giants are also investing heavily to secure a leading position in AI development.

Faced with massive investment plans, Altman has expressed interest in designing a new financial instrument to facilitate funding and acquisition of computational resources. He hinted that the scale of financing would surpass the $500 billion "Stargate" announced by the White House earlier this year, although no concrete implementation has been made yet.

In contrast, some companies have already implemented financing measures. To maintain a healthy balance sheet while investing heavily in AI infrastructure, certain firms are employing complex financial instruments, such as Meta's joint venture to transfer data center projects externally, offloading risk to outside parties.

UBS has warned that these methods, resulting in a surge of private credit, may elevate market overheating risk; moreover, the high concentration among a few tech giants raises concerns about excessive credit concentration risk.