At 81 years old, the co-founder of Oracle is not just a list of fortunes: as of August 19, 2025, his wealth reaches $302 billion. The essence lies elsewhere: from data centers to AI security rules, Lawrence J. Ellison (Larry Ellison) places Oracle at the heart of American digital sovereignty. From Project Texas to "Stargate", infrastructure becomes an instrument of public power.
From Private Fortune to Sovereignty Tool
Ellison’s trajectory illustrates industrial patience. In 1977, he co-founded Oracle and holds about 41% of the capital. Thus, he creates a software annuity through licenses and maintenance. He then reinvests this annuity in the cloud and AI. In 2025, Oracle’s stock rises nearly 50%, driven by AI-related computing charges and an aggressive positioning of Oracle Cloud Infrastructure (OCI).
The shift is strategic: by relying on public interest contracts. Oracle connects the general interest (security, continuity, localization) with its private interest (investment in data centers, recurring revenue). Sovereignty becomes a market.
Oracle as a State Auxiliary: Project Texas and Sensitive Data
At the heart of the State-platforms relationship, Project Texas establishes Oracle as a trusted third party for TikTok: isolation of American services in the Oracle cloud and enhanced access controls, as described by TikTok’s US Data Security (USDS) and discussed within the framework of CFIUS (Committee on Foreign Investment in the United States). The goal: mitigate a perceived risk of dependence on a foreign power and ensure that data remains under American jurisdiction.
This arrangement is not neutral: it outsources to a private actor a quasi-sovereign function (securing and auditing a platform with a very large audience), while raising the question of independent evaluation and democratic control.
Mini-chronology — TikTok/CFIUS and Oracle (2020–2025)
• Sept. 2020: the White House announces an agreement in principle; Oracle becomes the prospective "technology partner" of TikTok for the United States (U.S. Treasury statement).
• 2021–2022: change of administration; review of the case within the CFIUS framework. The project evolves towards a solution of operational data isolation.
• 2022–2023: gradual deployment of Project Texas: hosting and access controls under Oracle Cloud for American users (official TikTok USDS documentation).
• Spring 2025: return of hypotheses of divestiture of American activities; Oracle remains a cloud provider and candidate for an expanded role, depending on upcoming political and regulatory arbitration.
"Stargate": Industrial Policy Through Infrastructure
Announced as an accelerator of AI infrastructure in the United States, Stargate associates Oracle and OpenAI; a public agreement for an additional 4.5 GW of data center capacity is announced. Beyond the numbers, the spirit is clear: reindustrialize through digital, secure the training and inference footprint, and repatriate critical chains (energy, chips, colocation, network).
The public lever is multiple: regulation (AI safety frameworks), public procurement (trusted cloud, FedRAMP), technological diplomacy (export controls). Ellison positions himself at this interface between industrial policy and private capacity.
Policies Framing Computing Power
• White House: Executive Order on AI (October 30, 2023): initial obligations on security and testing for advanced systems, inter-agency coordination, and safeguards for public use of AI.
• NIST: AI Risk Management Framework 1.0: voluntary framework for risk management (security, bias, rights), already adopted by some clients.
• OMB/FedRAMP (M-24-15): modernization of cloud security requirements in federal agencies; upgrade of certifications and the assurance chain.
• CISA: Secure by Design: "security by design" guidelines that influence public markets and, by extension, the offerings of major cloud providers.
By aggregating these components, Washington guides demand and standardizes security. Oracle, an American actor, benefits from a clear legal footprint concerning federal law. Consequently, it gains a comparative advantage on sensitive issues.
Sidebar — The American Legal Framework for AI and Cloud
Executive: Executive Order of October 30, 2023 on a safe and reliable AI (obligations for testing, reporting, and inter-agency coordination).
Standardization: NIST AI RMF 1.0 (risk governance: security, bias, rights), basis for tenders and internal policies.
Public Procurement: OMB/FedRAMP (M-24-15): enhanced security requirements, continuous assurance chain for agency cloud.
Cyber Defense: CISA – Secure by Design: obligations for default security and responsible software practices.
Foreign Trade: BIS: export controls on advanced semiconductors (GPU/HBM) and critical equipment.
Transparency: role of the Government Accountability Office (GAO) and audits for monitoring data centers and federal digital policies.
EU–United States Comparison: Standards, Sovereignty, and Public Procurement
Regulation Approach
• European Union: the AI Act adopts a risk-based logic, with targeted bans, enhanced obligations for high-risk systems, and innovation sandboxes. Progressive implementation between 2025 and 2026, under the supervision of national authorities and a European AI office.
• United States: no equivalent federal framework law; the Executive Order (10/2023) sets safeguards and delegates to agencies (NIST, CISA, OMB) the production of frameworks and public sector compliance.
Cloud, Localization, and Portability
• EU: NIS2 tightens security and incident reporting obligations. DORA imposes operational resilience on the financial sector and supervision of critical cloud providers. The Data Act enshrines reversibility and interoperability of services (anti-lock-in) from September 2025. The EUCS scheme (cloud certification, ENISA) is in finalization and could include legal sovereignty requirements (immunity from non-EU laws). Transatlantic transfers rely on the Data Privacy Framework (2023), subject to CJEU review.
• United States: FedRAMP/OMB structures public procurement and federal cloud security; data localization mainly involves contractual clauses and sectoral regimes.
Public Procurement and Competition
• EU: increasing demands for trusted cloud and portability/interoperability clauses to avoid contractual lock-ins. The DMA and competition law aim to limit the structural positions of major players.
• United States: orientation by technical frameworks, framework contracts, and audits, with a logic of performance and security rather than localization.
Energy and Environment
• EU: the energy efficiency directive (recast 2023) introduces reporting obligations for data centers and sobriety indicators (e.g., efficiency, water). CSRD standards strengthen the disclosure of climate impacts.
• United States: monitoring by DOE/LBNL, local devices (utilities, PPA), and eligibility for certain incentives depending on the states.
Convergences and Divergences
• Convergence on security by design, auditability, and risk management.
• Divergence on legal sovereignty: the EU formalizes territorial control and reversibility requirements; the United States favors an agency-centered and contractual approach.
Energy, Water, and Public Finances: The Democratic Cost of Gigawatts
The electric demand of data centers has tripled over ten years and could double or triple again by 2028, according to the Department of Energy (LBNL report 2024). Locally, projects require heavy connections, water, and tax incentives.
Democratic question: who pays? Between regulated tariffs, subsidies, and network arrangements, part of the cost is socialized. However, direct jobs remain limited. Budget transparency, environmental conditionality, and independent evaluation become criteria of legitimacy.
Trade and Technological Diplomacy: Chip Control, Securing Chains
Sovereignty also involves mastering silicon. Since 2022, the Bureau of Industry and Security (BIS) has strengthened export controls on advanced circuits (GPU and HBM) to certain destinations, to limit access to advanced training capabilities. These geopolitical measures reconfigure supplies in chips, servers, and memory. Thus, the cost and pace of infrastructure deployment.
A Democratic Debate to Structure
The power of infrastructure calls for a counter-power. Several projects are necessary:
- Public accounts: publish, ex ante and ex post, the costs and benefits of installations (energy, water, taxation), based on common frameworks (DOE, LBNL).
- Procedure: frame projects with audits and public consultations, and with harmonized cyber requirements (NIST, CISA, FedRAMP).
- Platform transparency: Require independent reports on security, data usage, and impact of models. Especially for "general interest" services (information, health, and education).
- Public procurement: avoid contractual lock-ins through multicloud and open standards.
The GAO has long advocated for tighter governance of federal data centers: a useful benchmark for the AI era.
Style and Networks: The Art of "Being at the Table"
A Republican sympathizer, Larry Ellison has cultivated access to decision-makers. He supported Donald Trump, participated in announcements around AI, and served on the Tesla board (2018-2022). His method: mix political alliance, technical proposal (cost, performance, security), and execution capacity (rapid deployment of data centers). The result is an interdependence: the State needs credible operators; Oracle needs a stable public framework to invest on a large scale.
What the "Ellison Case" Reveals
Digital sovereignty is not declared; it is built. By establishing itself on the infrastructure layer (computing, storage, network), Oracle is betting on the power that now matters: that of the available computing time, under American law. For citizens, the stakes are high: freedoms (data governance), public finances (implementation costs), climate (center footprint), competition (avoiding concentration). It is under these conditions that Ellison’s gamble can be considered a public good as much as a private success.
Biographical Highlights
• Full Name: Lawrence Joseph Ellison
• Birth: August 17, 1944, New York
• Oracle: co-founder (1977), CEO until 2014, executive chairman and CTO since 2014
• Ownership: about 41% of Oracle
• Net Worth: $302 billion (August 19, 2025, Bloomberg index)
• Notable Achievements: purchase of Lānaʻi (2012); victories in the America’s Cup (2010, 2013); launch of SailGP (2019)
