Oracle Corporation Japan (TSE: 4716) just reported its best-ever revenue number: ¥285.1bn for FY2026, up 8.2% year-over-year, with a 31.5% operating margin that would make most Japanese industrials envious. Management is talking about AI, cloud, and digital transformation.
None of that is the story worth following here.
The Numbers Underneath the Headline
Revenue grew 8.2%, but operating profit grew only 3.4%. That gap is the whole story. The segment breakdown makes it precise:
| Segment | Revenue FY26 | YoY | Op. Margin FY25 | Op. Margin FY26 |
|---|---|---|---|---|
| Cloud & Software | ¥244.5bn | +9.6% | 38.4% | 38.1% |
| Hardware | ¥15.0bn | −3.5% | 3.6% | 3.6% |
| Services | ¥25.5bn | +2.6% | 23.5% | 20.7% |
| Corporate overhead | — | — | −¥5.3bn | −¥9.2bn (+74%) |
The core software business — 86% of total revenue — held its margin at 38%. The compression came from two places: Services margin fell 2.8pp, and corporate overhead jumped 74%, from ¥5.3bn to ¥9.2bn. That overhead surge is where the real question lives.
What Oracle Is Actually Selling
The “DX” and “AI” language in Oracle Japan’s filings obscures what is actually happening. The Cloud & Software segment is almost entirely Oracle Database licenses and annual support fees, charged to enterprises that have been running Oracle systems for two or three decades. It is not new customers choosing Oracle for new projects.
The Java licensing change in January 2023 illustrates the business model precisely. Oracle switched from per-device pricing to per-employee pricing, immediately raising costs 3x to 30x for many enterprises. Nearly 80% of organizations are now actively migrating away from Oracle Java. Seventy-three percent have undergone an Oracle Java compliance audit in the past three years — audits that arrive unannounced and routinely cause project delays and unbudgeted costs.
This is a licensing enforcement business. The “growth” in FY2026 partially reflects forced price increases on a captive installed base, not new customers choosing Oracle over alternatives.
AI migration agents — commercial tools that automate schema conversion and code refactoring from Oracle to PostgreSQL — are now mature enough to handle enterprise-scale migrations. Companies report 60–80% cost reductions post-migration. The technical moat that once made Oracle irreplaceable is thinning faster than the license revenue numbers suggest.
The Government Cloud Narrative Does Not Hold Up
Oracle’s filings prominently feature Oracle Alloy partnerships with NTT Data, Fujitsu, NRI, and SoftBank. Oracle Alloy is a licensing arrangement under which Oracle deploys its full OCI (Oracle Cloud Infrastructure) technology stack inside a partner’s own data centers. The partner operates the service under their own brand and sells it to end customers — primarily Japanese enterprises and government bodies that require data to remain within Japanese jurisdiction.
The problem is the economics of that structure. Under Oracle Alloy, Japanese partners retain the customer relationship and the service margin. Oracle collects a platform licensing fee — wholesale, not retail. In Japan’s government and financial IT market, where NTT Data and Fujitsu have spent 40 years building institutional trust, Oracle is not the prime contractor. It is the technology supplier to the prime contractor.
OCI has ISMAP certification and government cloud designation. But AWS and Google Cloud are the services actually running live government workloads. Sakura Internet, a domestic Japanese provider, achieved formal government cloud designation only in March 2026. For new AI development, enterprise developers are choosing AWS Bedrock, Azure OpenAI, and Google Vertex AI — not Oracle. Oracle’s generative AI strategy is to bundle AI into existing Oracle products, which retains existing customers but attracts no new ones.
The Alloy Bet Carries the Same DNA
Oracle Alloy may be the growth narrative, but it carries the same structural DNA as the Java licensing debacle — and that matters for anyone evaluating whether Oracle Japan’s strategy will work.
What happened with Java is now well-documented. In January 2023, Oracle unilaterally changed Java SE from per-device pricing to per-employee pricing, increasing bills by 3x to 30x overnight. Gartner used the word “predatory.” Nine out of ten Oracle Java customers are now looking to leave — 81% have migrated, are migrating, or plan to migrate. Dissatisfaction is highest in Europe, where 95% of German and 92% of French users want to switch. A 2026 Dimensional Research study found 92% of Oracle Java users concerned about pricing, up from 82% the previous year.
Oracle Alloy faces a structurally similar problem, but from the partner’s side. Partners who advised on Alloy implementations in 2024–2025 found that capacity commitments routinely ran 30–50% ahead of realistic first-year demand ramps. The financial exposure for that gap sits entirely with the partner — NTT Data, SoftBank — not with Oracle. Exit and ramp-down terms in first-contract drafts were left deliberately vague, concentrating the downside risk on the partner.
Beyond the capacity risk, there is an upgrade-timing problem. Oracle controls its product roadmap and end-of-support schedules. When Oracle announces that a given version of an Oracle application or database will lose support on a particular date, every partner running that version — and every end customer of every Alloy-based service — must upgrade on Oracle’s timeline, not their own. For NTT Data’s government ministry clients, this means Oracle’s deprecation calendar overrides Japan’s ministry IT budget cycles. The partner cannot offer its customers schedule flexibility that Oracle has not permitted.
Oracle has demonstrated that it will change licensing terms when it suits the parent — Java is proof. If Oracle Alloy pricing or terms change at contract renewal, NTT Data and SoftBank will face the same choice their Oracle Java customers faced: pay the new price or begin an expensive migration from infrastructure they’ve built their own services on top of.
The Cash Flows Tell the Real Story
The balance sheet reveals what Oracle Corporation (US parent) is actually doing with its Japanese subsidiary:
- Intercompany loans from Oracle Japan to Oracle Corp jumped from ¥110bn to ¥212bn in FY2026 — Oracle Japan is funding its parent
- FY2026 dividend: ¥858/share (¥198 regular + ¥660 special), totaling approximately ¥110bn to be paid in August 2026
- FY2027 dividend: listed as “−” — no forecast, complete silence after an extraordinary payout
- FY2027 EPS guidance: ¥525–540 per share (vs ¥495.97 actual) — modest growth, but framed as a range rather than a commitment
Oracle Corp owns approximately 82% of Oracle Japan. What the cash flow pattern shows is consistent with a parent extracting value from a profitable subsidiary before potentially restructuring it. Oracle Japan’s ¥74.7bn in operating cash flow is excellent. An increasing share of that cash is leaving Japan.
The Precedent Worth Watching
IBM Japan delisted from the Tokyo Stock Exchange in 2011. At the time, the Japan business was still profitable — the delisting was simply a decision that maintaining a listed subsidiary no longer served the parent’s interests. The current pattern is similar: a dominant US technology parent, a profitable but structurally declining Japanese subsidiary, partnerships with local giants that reduce Oracle Japan’s direct market role, and capital being systematically repatriated.
The ¥1.2 trillion investment Oracle announced for Japan’s cloud infrastructure is Oracle Corp’s investment in data centers — not Oracle Japan’s capex. Oracle Japan’s role in that buildout may ultimately be a sales channel, with infrastructure operated by Oracle Corp directly or by NTT Data and SoftBank on Alloy licenses. If that transition completes, the rationale for keeping Oracle Japan listed dissolves.
What This Means
Oracle Japan’s 31.5% operating margin is real and will not evaporate overnight. The core database maintenance business is sticky, and the ¥112bn in contracted deferred revenue provides near-term visibility. But the investment case for holding Oracle Japan as a listed equity depends on assumptions that are increasingly hard to support:
- That government and financial sector Alloy contracts generate margin-accretive revenue for Oracle Japan specifically, rather than flowing to NTT Data and SoftBank
- That the blank FY2027 dividend forecast is a timing issue, not a signal
- That the 82% parent does not eventually decide the listed structure is no longer worth maintaining
None of these are base case today. But all three are live risks that a standard “high-margin software company” framework completely misses. The headline margin tells you this business is profitable. The cash flows tell you who is capturing that profitability — and it is not the Tokyo Stock Exchange minority shareholders.
Source: Original filing (TDnet) | 日本語版
Disclaimer | This article is for informational purposes only and does not constitute investment advice.