Oracle Sees High Steady-State ROIC for OCI Infrastructure as Megawatt Ramp Nears

Oracle’s evidence for high steady-state ROIC in OCI infrastructure, and expected margin evolution as revenue ramps to contractual levels

1) Evidence Oracle cited for high steady-state ROIC despite heavy investment

A. Target/expectation: “high 20s” ROIC at steady state (project-level, back-of-envelope)

Oracle’s management explicitly framed returns for OCI infrastructure (including CPU/GPU) in terms of return on invested capital (ROIC), stating they see ROIC “in the high 20s at a steady state” once revenues have ramped for large projects at the project level. ¹
They also explained the calculation logic behind the back-of-envelope ROIC: after-tax operating margin plus depreciation, divided by gross investments (total gross CapEx at the project level). ¹

B. Contract structure aimed at preserving margin at maturity (including “bring your own hardware” and prepaid)

Management linked the projected ROIC to specific contract behaviors and structures:

• Oracle reported $67 billion in AI infrastructure contract value signed in the quarter, with the majority either “bring your own hardware” or prepaid. ²
• Management said this increased OCI “bring your own hardware or prepaid” contracts to $75 billion, and that these have “no degradation in margin compared to our other contracts.” ²

This matters because it addresses a common concern: heavy capex/infrastructure buildout could dilute margins if customer terms don’t compensate. Oracle’s stated evidence was that even for capital-provided (BYOH) or prepaid arrangements, margins are not degraded versus other contracts—supporting their ROIC thesis. ²

C. Utilization and renewals consistent with steady-state economics (multi-tenant capacity monetization)

Oracle also cited operating indicators that would tend to support steady-state profitability (and thus ROIC), especially for a multi-tenant infrastructure model:

• In Q4, 35 thousand GPUs from 59 separate customers were up for renewal, and 49% of those customers renewed for 92% of those GPUs. ²
• Management clarified that the remaining 8% of GPUs were not idle, because most GPUs were subsequently sold to other customers in the same quarter. ²
• Oracle stated global GPU utilization rate of 97.5%. ²

High utilization and the ability to keep capacity monetized through renewals/transfers are key operational supports for maintaining margin after initial construction and ramp. ²

D. Delivery scale and pace (evidence of revenue ramp timing that management tied to ROIC)

Oracle used deployment/megawatt delivery metrics to support that large projects move to contracted revenue levels (the “steady state” point in their ROIC framing):

• They said Q4 finalized an impressive FY 2026 where they delivered more than 1.2 gigawatts to customers. ²
• They added FY 2027 Q1 delivery was approaching 1 gigawatt, described as nearly the same capacity as delivered in the previous 4 quarters combined. ²

While these delivery figures are not ROIC by themselves, Oracle’s ROIC framework explicitly requires that revenues ramp for large projects before reaching high steady-state ROIC. ¹ The gigawatt delivery pace is management’s evidence that projects are moving toward that ramp stage. ²

2) How management expected margins to evolve as data center projects reach full contractual revenue

A. Near-term pressure, then rapid improvement as contractual revenue levels are reached

Oracle’s management directly discussed margin timing during the ramp-up period:

• They stated FY 2027 gross margin would step down due to timing for the ramp up of data center projects into their full revenue contribution, plus impacts from mix. ³
• They added that while these investments create pressure on near-term gross margins in the infrastructure business, they expected margin performance to improve rapidly as they reach full contractual revenue levels at the data centers. ³

So the margin path they described is:

1. ramp period: gross margin pressure/step-down due to timing into full revenue ³
2. maturity/steady state: rapid improvement as projects hit full contractual revenue levels ³

B. Cost actions to improve operating leverage (operating costs slightly negative YoY in dollars)

In addition to the gross margin recovery expectation, management described operating cost behavior:

• They expected operating costs to be slightly negative year over year in dollar terms due to efficiency actions, which they said would drive improved operating leverage. ³

This provides a second pathway for margins to improve beyond just revenue ramp: even if costs don’t fall aggressively in accounting terms, management expects them to be restrained enough (slightly negative in dollars) to enhance leverage as revenue accelerates. ³

C. Reinforcement from explicit revenue ramp guidance as megawatts come online

Management also connected margin evolution to revenue acceleration in periods where megawatts are brought online:

• They stated they expected revenues and earnings to accelerate in the second half of the year as they bring further megawatts online to fulfill customer demand. ³

This supports the “timing” mechanism they used: as capacity reaches contracted revenue contribution levels, earnings power should improve. ³

Synthesis: what Oracle is claiming is the causal chain

Across the excerpts, Oracle’s argument for high steady-state ROIC in OCI infrastructure combines:

• a quantitative ROIC expectation (high 20s) tied to a project-level earnings/depreciation and gross CapEx framework. ¹
• evidence that contract terms and capacity monetization (including BYOH/prepaid) are structured to avoid margin degradation and sustain utilization (97.5% utilization; GPUs rolled into renewals/other customers). ²
• an explicit margin timing narrative: gross margin pressure during ramp (FY 2027 step-down) that improves rapidly once projects reach full contractual revenue levels, helped by efficiency-driven operating leverage. ³

That is the specific operating and expectation set Oracle provided to support ROIC resilience despite heavy investment, and to explain how margins would evolve as contractual revenue levels are reached. ¹²³