Big RPO, Bigger Risk: Why Oracle’s OpenAI Deal Isn’t the Slam Dunk It Seems

TradingKey - A wave of massive orders from clients like OpenAI has pushed Oracle’s Remaining Performance Obligations (RPO) to $455 billion, fueling a 36% stock surge — yet the rally quickly reversed. The pullback reflects investor skepticism about this “pie in the sky.” Wall Street is raising red flags: Can Oracle actually deliver? And can OpenAI afford to pay?

On Thursday, September 11, one day after soaring 35.95% — briefly making Oracle CEO Larry Ellison the world’s richest person — Oracle’s stock fell 6.23%, even as all three major U.S. indices hit record highs.

In its Q1 FY2026 earnings report, Oracle revealed that its RPO — future revenue from signed but unfulfilled contracts — surged 359% to $455 billion. Of that, $317 billion came from new agreements with OpenAI, xAI, Meta, and Nvidia.

The most eye-catching deal is a five-year compute agreement with OpenAI, under which Oracle will provide $300 billion worth of computing capacity starting in 2027 — meaning OpenAI must pay Oracle an average of $60 billion per year.

However, analysts warn that both the supply side (Oracle) and demand side (OpenAI) face significant risks:

• Can Oracle, already burdened with high debt, afford to spend billions expanding data center infrastructure?
• How will contract fulfillment timelines impact short-term revenue?
• Will long-term margins be pressured?
• Does customer concentration make growth more vulnerable?
• Can OpenAI remain profitable and meet these massive payment obligations?

Oracle May Be "Too Big to Deliver"

Compared to the explosive growth in backlog, Oracle’s near-term financials and guidance were underwhelming.

• Q1 Revenue: $14.93 billion, up 11% YoY, below analyst expectations
• EPS: $1.47, also below consensus
• Q2 Guidance: Revenue growth of 14–16%, EPS of $1.61–$1.65, within expected range

Long-term contracts mean revenue recognition lags, potentially clouding Oracle’s short-term growth. Both Morgan Stanley and JPMorgan expect only about 10% of the $455 billion RPO to convert into actual revenue over the next 12 months due to slow fulfillment timelines.

JPMorgan noted that this makes it difficult for Oracle’s near-term contracted revenue to show strong monthly growth. Most of the RPO offers little near-term financial benefit.

With short-term cash flow constrained, Oracle faces mounting pressure to fund capital expenditures for AI infrastructure. To fulfill such large-scale orders, Oracle has already raised its FY2026 capex guidance from $25 billion to $35 billion.

Yet, with current revenue growth lagging, Oracle’s free cash flow has already burned out. In Q1, capex reached $8.5 billion, dragging free cash flow to -$362 million, far below the expected $1.27 billion. Over the past 12 months (LTM), capital spending hit $27.4 billion, while free cash flow stood at -$10.9 billion.

Morgan Stanley pointed out that LTM free cash flow and return on assets are now significantly below historical levels, weakening Oracle’s ability to leverage revenue growth.

Moreover, JPMorgan noted that most new orders relate to lower-margin AI model training, which could drag down long-term operating profit growth.

As a result, the firm expects the market may raise Oracle’s revenue forecasts in coming years, but EPS may not follow, and free cash flow expectations could be further cut.

Even more concerning: Oracle ended Q1 with over $80 billion in net debt, a leverage ratio above 4x — approaching or exceeding thresholds that rating agencies might use to downgrade its credit rating.

Morgan Stanley predicts Oracle may need to launch a new public bond offering or seek innovative financing before year-end to manage liquidity pressure.

OpenAI: Not the Perfect Anchor

The $300 billion OpenAI contract is the crown jewel of Oracle’s RPO — but relying so heavily on a single client is risky. If OpenAI builds its own data centers or switches providers, Oracle’s revenue pipeline could dry up.

More critically, OpenAI itself faces immense profitability pressure. While ChatGPT became the fastest-growing consumer app in history — amassing over 700 million users in under three years — it has yet to turn a profit, losing billions annually.

Last year, OpenAI CEO Sam Altman revealed he expects the company to lose $44 billion by 2029 — the year he projects it will finally break even.

OpenAI is projected to generate $13 billion in revenue in 2025, far below the $60 billion annual payment it would owe Oracle. This means Oracle’s ability to realize this promised revenue hinges almost entirely on OpenAI sustaining explosive growth.

The Wall Street Journal noted:

“The fate of OpenAI’s massive commercial deals relies on what is increasingly looking like an uncertain prospect: the need for hundreds of millions of people to pay a lot more money for its tools and services in the near future.”

Estimates suggest global AI spending could reach $3 trillion by 2028, but some academics and industry advisors argue that converting users into paying customers may take far longer than markets assume.

A Menlo Ventures survey found that only 3% of consumers currently pay for AI services — totaling around $12 billion. Moreover, people use AI chatbots and models for scattered tasks, suggesting widespread, daily adoption is still a long way off.