오라클의 3000억 달러 오픈AI 거래는 재무 구조와 실행 리스크 측면에서 심층 검토를 받고 있다.

Oracle (NYSE: ORCL) could emerge as a vulnerability in the artificial intelligence boom if OpenAI encounters difficulties, according to Council on Foreign Relations senior fellow Sebastian Mallaby. Speaking on CFR’s *The Spillover* podcast, Mallaby warned, “If OpenAI runs into trouble, I think we can say that there’s a good chance that Oracle may also run into trouble,” citing the company’s elevated debt burden and heavy reliance on OpenAI.

This assessment arrives as OpenAI CEO Sam Altman has begun flagging signs of “unsustainable silliness” across the broader compute infrastructure buildout. Mallaby highlighted that while competitors such as Alphabet, Amazon, Microsoft, and Meta entered the AI race with robust balance sheets and ample internal funding capacity, Oracle has depended significantly more on borrowed capital to finance its data-center expansion.

Oracle’s financial positioning has drawn scrutiny from ratings agencies. On July 9, S&P Global downgraded Oracle’s credit rating to BBB-, leaving it one notch above junk status. The agency estimated that approximately half of Oracle’s $638 billion in contracted future revenue is linked to OpenAI. Additionally, S&P projects Oracle will require between $90 billion and $95 billion in capital expenditures for fiscal 2027. To help fund this expansion, Oracle intends to raise an additional $20 billion through an equity offering later this year.

Despite these leverage concerns, underlying demand remains strong. Customers are currently prepaying or providing hardware against $75 billion of Oracle’s largest AI contracts, and the company’s cloud infrastructure revenue surged 77% in fiscal 2026.

At the center of the risk profile is OpenAI’s agreement to purchase roughly $300 billion of Oracle computing capacity over a five-year period beginning in 2027, averaging approximately $60 billion annually. That figure exceeds OpenAI’s current reported annualized revenue of around $40 billion. Mallaby emphasized that the $40 billion represents revenue rather than liquid cash, meaning OpenAI must sustain rapid growth and continuously secure external financing to meet such substantial long-term commitments.

While Altman maintains confidence in OpenAI’s own infrastructure strategy—expecting the purchased capacity to be deployed profitably—he remains wary of industry-wide excesses. He pointed to emerging neocloud providers promising massive scale without sufficient revenue streams or committed customers, characterizing segments of the sector as “unsustainable silliness.” When asked whether a broader market downturn could impact OpenAI, Altman replied, “If the whole economy blows up, yes,” noting that systemic stress could impair the company’s ability to “confidently pay for the compute” it has already secured.

Should OpenAI pursue an initial public offering, it would gain access to fresh capital precisely as its multi-billion-dollar Oracle obligations begin scaling in 2027. Market data reflects shifting expectations around the timing: Kalshi traders currently assign an 85% probability that OpenAI will not announce an IPO by December 31, but place the likelihood of an announcement by March 31, 2027, at 48%. This suggests early 2027 as the most probable window for OpenAI to tap public markets as its financial commitments accelerate. Kalshi and Benzinga operate under an existing data collaboration agreement.