Who Pays for This

Oracle reported Q4 FY2026 after the close today. Non-GAAP EPS $2.11, beating $1.96 consensus by $0.15 — a 7.7% surprise. Revenue $19.2 billion, beating $19.1 billion. OCI grew 93%, landing squarely in the BofA/TD Cowen whisper range of 92-94%. RPO surged $85 billion to $638 billion. Every number the Street was watching came in clean.

Then Safra Catz said $70 billion capex for FY2027 and $40 billion in debt and equity financing, and the stock dropped 7%.

The Scorecard

Metric	Actual	Consensus	Delta
Non-GAAP EPS	$2.11	$1.96	+7.7%
Revenue	$19.2B	$19.1B	+0.5%
OCI (IaaS) Growth	93%	46-50% guide	Whisper: 92-94%
Total Cloud Growth	47%	46-50% guide	In range
RPO	$638B	$553B prior Q	+$85B QoQ
FY2026 FCF	-$23.7B	—	Negative

That OCI number matters. After Broadcom's June 3 report — where $16 billion AI semi guidance missed a $17.2 billion buy-side whisper and triggered a $1 trillion semiconductor selloff — the entire market was asking one question about Oracle: can the backlog convert?

It converted. 93% OCI growth, accelerating from 84% in Q3, landing within a percentage point of where BofA and TD Cowen said it needed to land. The conversion question is answered.

The New Question

But the market didn't rally on a clean beat. It sold. Because Catz revealed the cost of that conversion:

Oracle spent $55.7 billion in capex in FY2026 — already above its $50 billion projection. Q4 alone was $15.9 billion. And it's accelerating to $70 billion. Operating cash flow was $32 billion for the year, which means even with a growing top line, the gap between what Oracle earns and what it spends is widening.

To fund this, they're raising $40 billion — including a $20 billion at-the-market equity issuance that will dilute existing shareholders. The stock fell 9% at the exact moment that fundraise was announced on the call, before recovering to roughly -1% by late evening.

The RPO Footnote

One detail buried in the press release. Oracle disclosed that $75 billion of the $85 billion sequential RPO increase relates to "prepaid or customer-supplied GPU hardware" for large AI contracts. Customers are fronting the GPU cost and Oracle is booking it as backlog.

This isn't improper — it's the new mechanics of hyperscale AI infrastructure deals. But the $638 billion RPO headline overstates Oracle's organic revenue pipeline. Strip out the GPU passthrough, and the sequential increase was closer to $10 billion. Still strong. But $85 billion and $10 billion tell very different stories about demand trajectory.

CPI Collision: The Morning Before

May CPI hit 4.2% headline, matching consensus exactly. Core came in at 2.9% year-over-year with a 0.2% monthly gain — slightly below the 0.3% estimate. Energy drove over 60% of the increase at 23.5% year-over-year, the Iran war still flowing through every fuel-dependent line item.

For Oracle specifically, the CPI print was neutral to mildly positive. A surprise above 4.2% would have turned ORCL's $124 billion debt load and 5.2x D/E ratio into a rate-hike-panic headline. It didn't. But at 4.2%, rate cut expectations remain dead. The FOMC is expected to hold on June 17. Oracle's interest expense — already 28.76% of net income — isn't getting cheaper.

Guidance: Acceleration Into the Gap

Q1 FY2027 guidance: revenue growth 27-29%, cloud growth 58-64%. Both represent acceleration from Q4. Full-year FY2027 confirmed at $90 billion revenue with non-GAAP EPS raised to $8.05.

The forward numbers are strong. But after AVGO's lesson — where the market traded on the whisper, not the guide — the question is whether buy-side models will embed a $70 billion capex number that makes FCF inflection impossible in FY2027. Oracle generated $32 billion in operating cash flow this year. Even if that grows 30%, it's ~$42 billion against $70 billion in capex. The math doesn't close.

After AVGO, Before ADBE

Oracle sits between two earnings events that frame its meaning. A week ago, Broadcom's $16 billion AI semi guidance missed a $17.2 billion whisper and the market erased $1 trillion from chip stocks. Tomorrow, Adobe reports with its own conversion question — whether Figma's 46% growth and AI tools can offset slowing Digital Media ARR under a departing CEO.

The pattern across all three: strong backward-looking results, questions about the cost and sustainability of forward growth. AVGO's cost was expectation overshoot. Oracle's cost is literal — $70 billion in concrete and GPUs. Adobe's cost is execution risk.

Oracle answered the question the market was asking. The market responded with a harder one. Every number was green — EPS, revenue, OCI growth, RPO, guidance. The stock fell anyway, because the cost of delivering those numbers is growing faster than the numbers themselves.