Backlog Isn’t Bank Balance: A Reality Check on Oracle’s AI Story

Reading Time: 7 minutes

Prefer watching instead of reading? Watch the video here. Prefer reading instead? Scroll down for the full text. Prefer listening instead? Scroll up for the audio player.

P.S. The video and audio are in sync, so you can switch between them or control playback as needed. Enjoy Greyhound Standpoint insights in the format that suits you best. Join the conversation on social media using #GreyhoundStandpoint.

---

When Oracle announced a reported $300 billion, five-year compute contract with OpenAI, markets responded with their usual mix of exuberance and hysteria. The company’s shares surged nearly 40 percent in a single day — its sharpest rise since 1992. Larry Ellison briefly overtook Elon Musk as the world’s richest man. The optics were immaculate and the choreography flawless, a champagne moment for investors.

But theatre is not delivery. What Oracle served was less a coronation than a carefully staged performance: a heady cocktail of ambition, backlog, and speculation. At Greyhound Research, we argue that such moments call not for applause but for scrutiny. The right instinct is not to toast, but to check the bill.

$300 Billion on Paper: What’s Actually Been Promised

According to the Wall Street Journal, OpenAI has signed a contract worth $300 billion, beginning in 2027 and spread across five years. Oracle itself reported that its Remaining Performance Obligations (RPO) had soared to $455 billion, up 359 percent year-on-year at the close of Q1 FY26.

These numbers are impressive, but they are not the same as cash. RPO is an accounting device — a forward obligation, often conditional, and invariably subject to renegotiation. In enterprise software, multi-year SaaS contracts can be cancelled or trimmed at renewal. In infrastructure, commitments typically involve staged ramps and thresholds that allow for significant adjustment. The $300 billion is not an iron-clad cheque but a ceiling of possibility. It is, at best, a promissory note.

We at Greyhound Research believe this distinction matters. Between now and 2027, Oracle must invest billions in new facilities, power, and chips. It is being celebrated today for revenues that may or may not materialise years into the future. In less polite terms, Oracle has been paid in applause for a banquet that has yet to be cooked.

Oracle’s Wallet vs. Its Wish List

The financial arithmetic is stark. Oracle has pledged $35 billion in capital expenditure for FY2026, a dramatic increase intended to fund 37 new data centres. Yet its annual operating cash flow remains at $20–22 billion. Rivals such as Amazon and Microsoft generate five times that.

This imbalance means Oracle will almost certainly need to borrow. It already carries significant debt — more than $80 billion — a legacy of past acquisitions and payouts. Bond investors have tolerated this thanks to steady software margins, but the pivot to capex-intensive infrastructure may test their patience. Rising interest rates and ratings agency scrutiny could drive up the cost of capital, putting further pressure on margins.

Meanwhile, equity markets have priced Oracle as if none of this matters. The stock trades at a forward price-to-earnings ratio of roughly 45–50×, its richest valuation since the dot-com bubble.

We at Greyhound Research believe such multiples imply a future of flawless execution. And flawless execution is the one thing infrastructure rarely delivers. Data centres, unlike software releases, cannot be patched overnight.

The Buyer That Might Not Have the Bank Balance

The other side of this story is equally uncomfortable. OpenAI is projected to earn ~$13 billion in revenue in 2025. The Oracle contract, as reported, implies spending far beyond that amount. OpenAI is not expected to achieve profitability until 2029. That leaves an uncomfortable gap, one that can only be bridged if capital markets continue to subsidise the company on an unprecedented scale.

OpenAI has been sustained to date by injections of capital from Microsoft, venture funds, and sovereign investors. Each of these pools has its limits. Microsoft may tire of bankrolling a rival to its own Azure cloud. VCs may reconsider underwriting losses that stretch into the tens of billions. Sovereign investors may prefer to support domestic champions rather than subsidise US-based ventures. Should those taps close, OpenAI’s ability to pay Oracle would quickly falter.

Even if the funding flows, the contract is far from watertight. The $300 billion figure is a reported maximum, not an upfront payment. Key details — GPU classes, rate cards, usage thresholds, and renegotiation gates — have not been disclosed. Chip efficiency gains could reduce compute requirements sharply. And Microsoft’s right-of-first-refusal on OpenAI workloads adds further uncertainty.

We at Greyhound Research believe, Oracle has tied its fortunes to a client whose obligations are conditional and whose ability to pay rests on investor generosity.

From Cloud Dreams to Kilowatts and Cables

Execution may prove the most formidable obstacle. Oracle’s Stargate programme promises 4.5–5 gigawatts of new capacity. To contextualise, Ireland’s entire data-centre fleet consumes roughly 2 GW today. Oracle is pledging to build more than twice that in just a few years.

The scale of the challenge is immense. Grid operators in the US and Europe already warn of interconnect queues that stretch for years. The International Energy Agency projects global data-centre electricity demand will double by 2030, with AI workloads the primary driver. Securing power on this scale will require long-term contracts, co-located generation, and significant political support.

At the same time, NVIDIA’s most advanced GPUs remain in chronic short supply. Allocation is dictated as much by geopolitics as by purchase orders. Oracle must therefore compete not just with rivals but with entire nations for access to silicon. The result is an execution environment where even minor setbacks could cascade into delays and cost overruns.

Greyhound CIO Pulse 2025 reveals that nearly half of CIOs — 47 percent — already view energy cost and availability as their top barrier to AI scale. For these executives, the problem is not algorithms but amperage.

Four Ways This Can Come Undone

The fragility of Oracle’s story rests on four interwoven risks. The first is concentration: by placing so much of its future in the hands of OpenAI, Oracle has exposed itself to the fortunes of a single, volatile partner. Should OpenAI stumble, Oracle’s backlog will unravel with alarming speed.

The second is contract fragility. The $300 billion figure, however dazzling, is a ceiling not a floor. It is contingent on usage ramps, subject to renegotiation, and vulnerable to disruption by chip efficiencies and alternative architectures. What looks monumental today may appear much smaller once the clauses are tested.

The third risk is execution. Building 5 GW of capacity, securing power, and sourcing GPUs in a constrained global market is a task of extraordinary difficulty. Grid queues, chip scarcity, and permitting hurdles all threaten to slow progress. Oracle is promising to bend physics and politics to its will. History suggests neither bends easily.

Finally, there is valuation. Oracle’s shares soared nearly 40 percent on the announcement, then fell 6–7 percent the very next day when analysts and investors began to question how much of the backlog was truly bankable. Markets are notoriously fickle. When narrative races ahead of fundamentals, corrections are rarely gentle.

We at Greyhound Research are of the firm belief that Individually, each of these risks would warrant caution. Together, they form a cocktail strong enough to give any CIO a hangover.

What CIOs Must Nail Down Before Signing Anything

For enterprises, the implications are immediate. Oracle’s backlog headlines must be treated as theatre, not as guarantees. Contracts should be anchored in milestone-based schedules tied to actual interconnect approvals and energy sourcing strategies. Anything less is an act of faith.

Pricing transparency must be non-negotiable. Rate cards, GPU classes, indexed pricing, and conversion rights must all be spelled out in detail. Without these protections, Oracle will be tempted to pass its capital risk directly onto customers.

Exit and portability provisions must be hardwired into contracts. With OpenAI absorbing so much of Oracle’s projected capacity, sudden allocation shocks are inevitable. Enterprises must negotiate multi-cloud portability and financial remedies for non-delivery.

Energy scrutiny is no less critical. Contracts should include energy SLAs covering sourcing mix, curtailment protocols, and co-location strategies. According to Greyhound CIO Pulse 2025, 47 percent of CIOs cite energy cost as their top barrier, while 62 percent dismiss mega-announcements as more optics than delivery. Enterprises would be wise to let this scepticism guide their procurement.

Procurement leaders must also consider the wider context. Oracle’s backlog is dominated by one client, and its capex binge depends on continued investor generosity. Should either falter, customers risk being collateral damage. Guardrails in contracts are not a luxury. They are essential.

Greyhound Standpoint: Ambition Tempered by Reality

Oracle has told a strong story, casting itself as OpenAI’s partner of choice and winning a round of applause from markets that briefly treated backlog as though it were cash in the bank. But stories are not balance sheets. Behind the noise sits a business leaning heavily on a client, preparing for an infrastructure build of unusual scale, and trading at valuations that assume everything will go right.

For CIOs and CFOs, the advice is uncomplicated. Take Oracle’s ambition seriously, but don’t base strategy on headlines. Contracts should be tied to visible capacity, with clear safeguards and flexibility to shift if delivery falls short. Diversification remains the surest defence.

In the end, the organisations that will benefit most from AI are not those chasing the loudest announcements, but those choosing partners who can deliver consistently and at sensible cost. Oracle may well grow into that role. The proof, however, will lie in execution, not in press releases.

Copyright Policy. All content contained on the Greyhound Research website is protected by copyright law and may not be reproduced, distributed, transmitted, displayed, published, or broadcast without the prior written permission of Greyhound Research or, in the case of third-party materials, the prior written consent of the copyright owner of that content. You may not alter, delete, obscure, or conceal any trademark, copyright, or other notice appearing in any Greyhound Research content. We request our readers not to copy Greyhound Research content and not republish or redistribute them (in whole or partially) via emails or republishing them in any media, including websites, newsletters, or intranets. We understand that you may want to share this content with others, so we’ve added tools under each content piece that allow you to share the content. If you have any questions, please get in touch with our Community Relations Team at connect@thofgr.com.