Oracle's Cloud-Infra Revenue Just Doubled to $7.4B. The Stock's 4% After-Hours Pop Says Nobody Believes It Yet.

Oracle did not just beat — it set the record. Cloud infrastructure revenue of $7.4B was 121% higher than a year ago, RPO jumped $26B sequentially to $664B, and management raised FY27 revenue to at least $90B and non-GAAP EPS to $8.10. Even the lumpy metrics cooperated: operating cash flow tripled to $23B, AI bookings exceeded $30B in a single quarter, and Oracle deployed 850MW of new capacity while running its installed base at 97.9% utilization. On the numbers alone, this was the cleanest hyperscaler print of the cycle.

Options had priced a ±12.3% move into the print; the actual after-hours reaction was +4.1% to $159.26, less than half the implied swing. The previous regular-session close was $152.94, already down 5.4% on the day before results — meaning the stock went into earnings already trading near 52-week lows and left them there. Over 12 months, Oracle is off 53% from its September 2025 peak of $345.72; year-to-date the stock is down 21.5%.

The friction is no longer the demand side. It is the funding side. FY26 capex was $55.7B — 162% above the prior year — and free cash flow swung to negative $23.7B. Management guided FY27 net capex to $70B plus another $20–25B in prepayments, with a total debt envelope of $125B and $20B of new equity already raised through an at-the-market offering. The math is straightforward: at $70–95B of annual capex against roughly $8B of quarterly non-GAAP operating income growth, the runway only works if the $664B backlog actually converts and AI contracts keep booking.

S&P moved Oracle to BBB- on July 9, 2026 — one notch above junk — citing the AI capex profile. Moody's holds Baa2 with a negative outlook and has warned leverage could touch 5x. Fitch is the lone holdout at BBB stable, but its own threshold (3.5x EBITDA leverage, 4x with leases) is tighter than where the company sits today. Net debt/EBITDA on a TTM basis is 3.5x; a single quarter of incremental debt-funded capex could push that past S&P's 4.5x trigger and reset the clock on a second downgrade. That is why a $19.3B revenue beat moved the stock less than $7.

This is why the megacampus thesis is harder to kill than the credit narrative suggests. RPO has compounded from $138B two years ago to $455B a year ago, $523B in December, $638B at FY26 year-end, and $664B today. Oracle is not selling capacity — it is selling backlog faster than it can deploy power. The bottleneck is megawatts, GPUs, and lease-ready land, not customer demand.

On growth rate, Oracle's 121% OCI print tops Alphabet's 82%, Microsoft's 43% Azure, and Amazon AWS at 28%. But the comparison that matters for capital is hyperscaler capex — Microsoft is targeting roughly $190B in calendar 2026, and the big-five combined plan sits at $660–690B. Oracle's $70B FY27 capex is about 10% of that pool, and Oracle is the only one of the group with a single-name contract (OpenAI, $300B) roughly equal to its own market cap. The growth lead is real; the question is whether Oracle can fund the build to keep it.

The clearest beneficiaries of Oracle's $70B are the picks-and-shovels. Vertiv raised 2026 organic sales guidance to ~31% on AI data center demand and exited Q2 with a $15B backlog — and Oracle's 850MW ramp this quarter is a direct order signal. Eaton is widening its share of the power side via the Boyd Thermal deal. Upstream, NVIDIA collects on every GPU shipped into OCI Supercluster, and TSMC sits one layer back as the foundry for the Blackwell silicon. Downstream, the same $664B backlog pressures the Microsoft/Alphabet/Amazon trio to keep capex elevated to defend share — a backdrop that has lifted Vistra and power utilities through AI-driven electricity demand forecasts.

Short-term (days to quarters): the next data points are the FY27 quarterly prints — Q2 in December will show whether the 850MW capacity ramp translates into another double-digit OCI beat and whether free cash flow stays positive. The most binary catalyst is the FY27 capex absorption: management has guided to a $40B debt+equity raise envelope and said the newest $30B of AI bookings do not force another raise. If a quarter prints capex above $25B or FCF back to negative, expect another 5–10% drawdown like June's; if not, the stock re-rates from the current 7.8x EV/sales back toward the high-teens. Watch the CDS market — a move below 150 bps would be the first sign the agencies are stepping back from the fallen-angel watch.

Long-term (one to three years): the OpenAI $300B contract begins recognizing in 2027 and runs five years, layering roughly $60B of annual revenue on top of the existing book. That is the bridge between today's $19.3B quarter and a $150B+ revenue run-rate by FY30. But the bridge requires Oracle to deploy roughly 5GW of additional capacity on top of the existing footprint — a build-out that consumes NVIDIA's next two Blackwell cycles, locks in tens of billions more debt, and pushes debt/EBITDA past 5x at peak. The thesis holds if backlog conversion tracks; it breaks if either AI training demand softens or the cost of capital resets higher. The stock today prices most of the bear case and little of the bull case — which is exactly the asymmetry that explains the muted after-hours pop.