Oracle’s $664B Backlog Masks a $300B Concentration Risk

Oracle just posted 121% year-over-year cloud infrastructure growth and a $664 billion remaining performance obligation backlog, yet the stock trades at a meaningful discount to its three-year average earnings multiple. That gap between operational momentum and market valuation is where the investment question lives.

The disconnect is not accidental. Buried inside the headline numbers sit a $300 billion dependency on a single private counterparty, OpenAI, a $20 billion equity issuance that will dilute existing shareholders by roughly 5.5% to 6%, and a balance sheet carrying close to $130 billion in total debt.

These are not peripheral concerns. They are the structural features that decide whether the bull case survives contact with reality. Here is the framework for stress-testing both sides of the thesis at the $152 entry point: what the numbers need to go right, what would break them, and which variables to watch before committing capital.

What Oracle’s Q1 FY2027 numbers actually tell you

Oracle reported fiscal Q1 FY2027 results on 10 September 2026, and on the surface they read as an unambiguous acceleration. Total revenue reached $19.35 billion, up roughly 30% year-over-year and ahead of the roughly $19.14 billion consensus. Cloud infrastructure revenue, the engine of the whole story, surged 121% to $7.4 billion.

Total cloud, combining infrastructure and applications, hit $11.6 billion, up 62%. Non-GAAP diluted earnings per share came in at $1.92, up 30% and comfortably past the $1.73 to $1.74 analysts expected. GAAP diluted EPS rose 55% to $1.56.

Oracle’s SEC EDGAR filings, including the 8-K current report submitted on 10 September 2026, are the primary regulatory source for the revenue, EPS, capital expenditure, and RPO figures cited throughout this analysis.

Then there is the number that complicates the clean read.

Operating cash flow ran around $23 billion for the quarter. But capital expenditures consumed roughly $28 to $28.5 billion, pushing free cash flow to somewhere between negative $5 billion and negative $5.4 billion. Oracle is not generating cash right now. It is burning it.

That capex figure is not a financial accident. It is a deliberate strategic bet, the cost of building the infrastructure the backlog demands.

Remaining performance obligations reached approximately $664 billion at quarter-end, the scale of contracted demand Oracle is now racing to service.

Here is what the combination tells you. Triple-digit infrastructure growth alongside deeply negative free cash flow means Oracle is spending ahead of revenue to build capacity it has already sold. Whether that trade is rational depends entirely on how much of that backlog actually converts, and any investor anchoring only on revenue and EPS is reading half the picture. Full-year FY2027 guidance calls for non-GAAP EPS of roughly $8.10 on at least $90 billion in revenue.

The capital structure behind the growth: debt, dilution, and what it costs you

The cash burn does not fund itself. To understand the dilution figure that follows, you have to walk through the financing decisions that made it inevitable.

Start with the debt. Oracle carries total debt of roughly $129.5 billion, though some sources put the figure as high as $134.6 billion to $135 billion following recent senior note issuance. Net debt sits around $97.6 billion. Net debt to EBITDA runs an elevated 3.96x to 4.08x, and the gearing ratio (total debt to equity) sits near 3.01x.

That is a heavily levered profile, and it explains why Oracle could not simply borrow its way through the buildout. When leverage is already stretched, equity becomes part of the funding mix whether management prefers it or not.

Rating agencies have reportedly cut Oracle to BBB-, one notch above speculative grade, citing extreme customer concentration.

For FY2027, Oracle plans to raise roughly $40 billion in new capital through a blend of debt and equity. FY2026 capex had already overshot, landing at $55 to $56 billion against a $50 billion target, so management has now exceeded its own capex guidance at least once.

AI cloud margins in Oracle’s infrastructure segment currently run at 14-16%, far below the 30-40% management target, which means the company’s fastest-growing revenue line is also its least profitable at precisely the moment capex is running at its highest level in the company’s history.

How the dilution layers stack for existing shareholders

The equity portion is where existing holders pay. Included in the plan is a $20 billion at-the-market issuance, meaning Oracle sells new shares gradually into the open market. At a reference price around $126, that translates to roughly 158 million new shares, or 5.5% to 6% dilution of the outstanding count.

At-the-market equity issuance mechanics differ from registered public offerings in a critical way: NYSE rules that would trigger a shareholder vote on a large discounted private placement do not apply to ATM programmes, meaning Oracle can sell the full $20 billion in new shares without a shareholder vote regardless of the dilution magnitude.

But the ATM issuance is not the only dilution vector. Stock-based compensation adds more.

• ATM equity issuance: roughly 5.5% to 6% dilution
• Base stock-based compensation: approximately 1% annualised
• Maximum potential dilution from all outstanding awards: up to 3.5%

Combine these and the picture widens. A shareholder buying today owns a claim that shrinks on multiple fronts if all dilution materialises. Even if Oracle executes flawlessly on revenue, existing holders capture less of that upside per share than the headline EPS growth implies.

There is one interaction worth noting. If the share price appreciates meaningfully, the ATM issuance requires fewer shares to raise the same $20 billion, softening that layer. The stock-based compensation ceiling, however, does not move with the price. Understanding the full stack, ATM plus SBC plus ongoing leverage cost, is the minimum due diligence for anyone underwriting the bull case at $152. These are structural features of the Oracle-as-AI-infrastructure trade, not temporary balance sheet conditions.

The OpenAI dependency: concentration risk hiding inside a record backlog

A $664 billion backlog sounds like the ultimate margin of safety. It is not, and the reason is who sits inside it.

OpenAI signed a cloud and computing contract with Oracle valued at approximately $300 billion over roughly five years, starting in 2027. That works out to around $30 billion per year in revenue. To service it, Oracle is developing 4.5 gigawatts of additional Stargate data-center capacity in the US, bringing total Stargate capacity under development past 5 GW.

OpenAI is estimated to represent approximately 50% of Oracle’s $664 billion RPO backlog.

A large backlog is not the same as a diversified one. Roughly half of Oracle’s contracted future revenue rests on a single, private, cash-burning counterparty. That is the definition of concentration risk, and it is the reason a BBB- credit rating now hangs over the balance sheet.

Here is the arithmetic that should give any buyer pause. For OpenAI to honour its commitments, analysts estimate its revenue must scale from roughly $10 billion currently to over $60 billion by 2027. That is a private company with no obligation to tell you whether it is on track. The RPO number is not a safety net unless OpenAI’s business roughly six-times in about 18 months.

The failure pathways are specific:

• Revenue shortfall: OpenAI misses its scaling targets and cannot absorb the contracted capacity
• Renegotiation: OpenAI reworks terms, reducing the value Oracle recognises
• Self-build: OpenAI internalises its compute, walking away from third-party capacity

Any one of these leaves Oracle holding stranded capacity it built on spec.

What happens if xAI and Meta become competitors, not customers

The concentration problem has a second edge. Oracle’s other marquee AI contracts, with xAI and Meta, come with the same structural irony: both have signalled intentions to build their own infrastructure and potentially offer cloud capacity to third parties.

The tenants most capable of paying Oracle’s bills are also the ones with the capital and technical depth to internalise their compute entirely. Today’s customer becomes tomorrow’s competitor, and Oracle has limited leverage to prevent it.

There is a broader risk sitting underneath all of this. If open-source AI models close the performance gap with frontier proprietary models, the economic case for massive centralised superclusters weakens. That would reduce demand from every hyperscale tenant at once, not just one, which is precisely when Oracle’s capacity flexibility matters least.

Open-source commoditisation pressure on frontier AI models has repeatedly closed benchmark gaps within 6-12 months of a new release, structurally compressing the pricing window that underpins OpenAI’s ability to sustain the API charges needed to scale revenue to $60 billion per year and honour its Oracle obligations.

The bull case in numbers: what re-rating to fair value requires

The bull case is best treated as a model to stress-test, not a conclusion to adopt. Its foundation is straightforward. At $152, Oracle trades at roughly 24.7x earnings, well below its three-year average of 36.1x, according to Trefis. On that basis alone, the stock looks cheap relative to its own history.

The upside targets, though, each rest on specific assumptions.

FY2028 consensus revenue sits near $130.74 billion. The bull models assume operating margins expand into the 25% to 30% range on that higher base, with long-term infrastructure economics supporting gross margins of 35% or more on large multi-year AI deals.

AI infrastructure demand forecasts underpin the entire Oracle bull case: Citi has more than tripled its 2031 global data center IT load projection to 370 GW at a 25% compound annual growth rate, a figure that, if realised, would validate Oracle’s capacity buildout at scale, but which also implies that any demand shortfall from the current trajectory lands across the entire sector simultaneously.

For the bull case to clear, three conditions have to hold at the same time:

1. OpenAI executes, scaling its revenue to absorb the contracted capacity
2. Margins expand, with operating margins reaching the modelled 25% to 30%
3. The multiple re-rates, with the market awarding Oracle a valuation closer to its historical average

The gap between $152 and Trefis’s $237, the most conservative of the fair-value targets, implies roughly 56% upside. But that return is fully contingent on a private company’s revenue trajectory and Oracle’s ability to hold margins while absorbing $28 billion per quarter in capex. The $152 entry point is only cheap if every one of those inputs holds. Read the targets as optimism priced with conviction, and decide for yourself which of the three assumptions you actually believe.

Where Oracle stands if the model breaks: margin of safety at $152

Turn the bull model over and the bear case is simply the same inputs failing. If OpenAI pulls back or renegotiates, Oracle is left carrying 4.5 GW of stranded Stargate capacity and a multi-year free cash flow squeeze, all on top of $97.6 billion in net debt.

Bulls counter that Oracle can reassign AI computing capacity between tenants, so a single customer falling through need not be fatal. That flexibility is real, but it erodes fast in a broader oversupply scenario where multiple hyperscalers cut third-party cloud spending at once, or where open-source model convergence softens demand for centralised superclusters across the board.

The 24/7 Wall St bear scenario lands at $176.22, implying the downside from $152 is narrower than the bull-case upside, an asymmetric risk profile worth weighing explicitly.

That $176 figure is actually above the current $152 share price, meaning the bear scenario, which assumes only partial execution, still implies meaningful upside from here. The catch is that it assumes partial execution, not failure. A full OpenAI default scenario has no established floor in the analyst models reviewed here, and management has already shown it will overshoot capex guidance, spending $55 to $56 billion in FY2026 against a $50 billion target.

So the analysis converts into a watchlist rather than a verdict. Three observable signals will tell you whether the thesis is working or breaking:

1. RPO growth rate: is the backlog still expanding, or has it plateaued?
2. IaaS share of total cloud: does infrastructure keep outpacing SaaS, confirming the AI mix shift?
3. OpenAI-related disclosures: revenue milestones, infrastructure announcements, or any sign of self-build acceleration

A decision at $152 is not bull versus bear in the abstract. It is a question of whether you are willing to accept concentrated counterparty risk, structural dilution, and elevated leverage in exchange for exposure to one of the largest AI infrastructure buildouts on record.

Making an informed call on Oracle at a crossroads moment

Oracle’s infrastructure thesis is credible on the numbers available today. The problem is that its most important variables, OpenAI’s revenue scaling, hyperscaler commitment durability, and the broader AI demand trajectory, are not observable in real time. You are underwriting a private company’s growth curve without the disclosures a public one would give you.

The entry point at $152 makes sense for a specific profile: an investor with a three-year horizon, tolerance for leverage and dilution, and genuine conviction that the AI buildout is durable and that OpenAI specifically will grow into its obligations. For anyone without all three, the concentration risk is difficult to size responsibly.

You do not have to decide today. Watch the RPO growth rate, the IaaS share of cloud revenue, and OpenAI’s public milestones through Q2 FY2027 and beyond. Those three signals will tell you whether the model is holding before you commit.

This article is for informational purposes only and should not be considered financial advice. Investors should conduct their own research and consult with financial professionals before making investment decisions. Past performance does not guarantee future results, and financial projections are subject to market conditions and various risk factors.