Oracle with No Plan B: A High-Stakes Computing Power Gamble Challenging AWS

After the US stock market closed on June 10, Oracle released its Q4 FY2026 financial report, which exceeded expectations in almost all core metrics — revenue of $19.18 billion, up 21% year-on-year, surpassing Wall Street's consensus estimate of $19.1 billion; Non-GAAP EPS of $2.11, nearly 8% higher than the consensus of $1.96.

However, after-hours stock price plunged over 10%. The Nasdaq itself fell nearly 2% that day, and CPI data hit a three-year high, putting pressure on the broader market. But Oracle’s decline far exceeded the index, meaning the market saw more than just macro noise.

The numbers are impressive, but investors aren't just reading numbers

First, let’s look at the growth side. IaaS (cloud infrastructure) revenue was $5.79 billion, up 93% year-on-year. Here’s the growth sequence over the past eight quarters: 45%, 52%, 49%, 52%, 55%, 68%, 84%, 93% — still accelerating as the base expands, which is extremely rare among major tech companies.

RPO (Remaining Performance Obligations) rose from $553 billion last quarter to $638 billion, up $85 billion in a single quarter. Management stated on the earnings call that the Q3 and Q4 RPO increases "were almost entirely from large-scale AI contracts."

Put together, these numbers tell a tale of massively surging demand. But on the night, the market wasn’t concerned about how strong demand is, but how much it’s costing to meet it.

FY2026 full-year capital expenditure was $55.7 billion, far exceeding the previous guidance of $50 billion. Q4 CapEx alone burned $15.9 billion. At the same time, management announced a refinancing plan for FY2027 of another $40 billion, with $20 billion from ATM (at-the-market offerings), and the rest from debt. The $20 billion offering means direct dilution for existing shareholders, and was one trigger for the after-hours sell-off.

Customers buy their own GPUs, Oracle just operates

One of the biggest highlights on the earnings call was the progress of the BYOH (Bring Your Own Hardware) model.

BYOH means customers pay for GPUs themselves or pre-pay large sums, then hand over the hardware for Oracle to operate in its data centers. By the end of this quarter, customer investments through this model had accumulated to $75 billion.

This money isn’t counted as Oracle’s CapEx, but operational revenue from it is included in IaaS revenue.

In other words, Oracle is trying a “light asset” path — letting customers bear hardware costs while Oracle earns operation fees. CEO Clay Magouyrk said that current GPU utilization is at 97.5%. Another interesting stat: among expiring GPU leases, 49% of customers chose to renew and kept 92% of their GPUs, and the remaining non-renewed GPUs were immediately resold to customers on a waiting list.

Management stated the project-level ROI is "close to 30%." If this number is credible, the BYOH model can indeed drive IaaS growth without greatly expanding the balance sheet.

But note: company-wide Non-GAAP operating margin is 44.8%, boosted by the high-margin traditional Software business. Pure IaaS margins are most likely below the company average.

Management’s long-term IaaS margin target is 30-40%, which itself suggests current margins have not yet reached this range.

What does $55.7 billion CapEx tell us?

Look at CapEx next to revenue: FY2026 CapEx was $55.7 billion, revenue $67.4 billion, CapEx/revenue ratio at 83%. Operating cash flow was $32 billion (up 54% year-on-year), but entirely swallowed by CapEx, giving a full-year free cash flow of negative $23.7 billion.

Total liabilities rose from $92.6 billion to $129.5 billion, up 40% in a year. In November 2025, Barclays downgraded Oracle’s debt rating to underweight, warning it may slide towards BBB-, the last threshold for investment grade. But CreditSights gave a contrary opinion in April 2026, upgrading Oracle’s credit to Outperform, arguing that RPO provides adequate future revenue anchors.

Management made a commitment on the earnings call: no new debt in calendar 2026. Of the FY2027 $40 billion financing, half comes from ATM offerings, half from debt, but the new debt won’t be issued until 2027. This is an attempt to draw a line between growth and leverage, but whether investors buy it remains to be seen.

Break down the $40 billion: structure is more complex than it looks. Early in calendar 2026 Oracle already completed two large financings: $25 billion high-grade unsecured bonds (one-off issuance in early February, led by Goldman, attracted $129 billion in oversubscription, maturities from 3 to 40 years in eight tranches); $5 billion mandatory convertible preferred stock (6.50% coupon, conversion price about $160, led by Citi, also multiple times oversubscribed).

That means the 2026 calendar year’s $45-50 billion financing plan is basically executed. The “$40 billion” for FY2027 mentioned on the call, if it includes the $20 billion authorized but unused ATM offering capacity, then the actual incremental new debt limit is only $20 billion, and management is committed to no new bond issuance in 2026 calendar year. In other words, the supply pressure facing the credit market is much lower than the "$40 billion" headline figure suggests.

Another detail easily overlooked: the $5 billion mandatory convertibles’ dilution effect is delayed, only triggered at maturity; while ATM offerings dilute continually but flexibly, management can sell based on share price. The impact mechanisms are different — former is certain long-term dilution, latter is uncertain near-term pressure.

Income structure changes are also worth watching. Traditional Software business was down 2% year-on-year this quarter, second consecutive quarter of negative growth. SaaS growth stuck around 10% for eight quarters. Oracle is staking almost all resources on IaaS — if demand for AI compute falls short, it lacks a Plan B.

Is $638 billion in orders enough for years?

$638 billion RPO is an astronomical figure. Divide it by annualized IaaS revenue (about $23.1 billion), in theory it could "fill" 27 years of capacity. But big enterprise contracts don’t work like that.

Management disclosed RPO duration structure: 12% recognized within 12 months, 34% in 13-36 months, and 54% over 36 months.

Most orders won’t turn into revenue until three years out. Near-term revenue growth depends mostly on existing capacity and new sign-ups, not the absolute size of RPO.

Another unanswered question is cancellation terms. Management didn’t disclose the default or cancellation conditions for these big AI contracts. However, under the BYOH model, customers have already pre-paid or bought hardware, so sunk costs are high and massive cancellations are unlikely.

FY2027 revenue guidance remains at $90 billion, implying 34% growth. Q1 guidance is for revenue growth of 27-29%, Cloud growth 58-64%. Management also raised Non-GAAP EPS guidance to $8.05.

What to watch next

FY2027 has three key points to observe.

First: can Q1 IaaS growth stay above 90%? If growth slows, confidence in the $638 billion RPO will quickly fade.

Second: will BYOH model’s share continue to expand? If the scale of customer prepayments and GPU purchases keeps growing, Oracle’s own CapEx pressure will reduce and free cash flow could turn positive more quickly.

Third: credit rating direction. The divergence between Barclays and CreditSights reflects two views of Oracle’s balance sheet. If rating agencies further downgrade, Oracle’s financing costs will climb and directly erode margins.

Management laid out a long-term goal for FY2030: revenue CAGR 31%, EPS CAGR 28%. If achieved, Oracle’s revenue in five years would approach $180 billion, on par with the current scale of AWS. This is an extremely aggressive target, dependent on whether AI compute demand continues to surge as management expects.

Before the earnings, 5 investment banks raised their valuation for Oracle, with TD Cowen’s $300 target the most aggressive. But the stock move that night showed, at least in the near term, the market cares less about whether Oracle can keep high growth, and more about how much it will cost to achieve it.

Risk Disclaimer and TermsThe market has risks, investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial condition or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific situation. Investments made based on this are at your own risk.