Petrobras AI Infrastructure 2025: How a $500M Data Center Investment Signals a New Strategy

Petrobras’s Commercial AI Projects: A Strategic Shift from Tactical Buys to Foundational Infrastructure

Petrobras has transitioned its AI strategy from acquiring individual high-performance computing systems to building a dedicated, large-scale infrastructure ecosystem designed for long-term energy market dominance. This shift is evident in the scale and nature of its investments, moving from tactical processing power upgrades to the construction of a strategic, proprietary AI factory.

• Between 2021 and 2024, the company focused on deploying specific supercomputers to solve immediate computational challenges. This included the 2023 activation of the Tatu supercomputer, a $6.8 million system with 224 NVIDIA A100 GPUs, and the 2024 announcement of an $89 million purchase of five new supercomputers from Lenovo to process raw seismic data.
• Starting in 2025, the strategy evolved toward building a foundational asset. The centerpiece of this change is the October 2025 agreement with Elea Data Centers, a 17-year, $500 million contract to construct a new 30 MVA data center in São Paulo specifically for AI workloads.
• This infrastructure-first approach demonstrates a clear intent to leverage the company’s unique pre-salt data assets at an industrial scale, creating a durable competitive advantage that is difficult for peers relying on generalized cloud services to replicate.

Analyzing Petrobras’s Multi-Billion Dollar AI Investment Strategy

Petrobras’s investment strategy escalated significantly in 2025, moving from targeted HPC procurements to a long-term, multi-billion dollar commitment to foundational AI infrastructure. This financial pivot underscores that the company views AI not as an IT project, but as a core pillar of its future profitability and operational efficiency, as highlighted in its $109 billion investment plan for 2026-2030.

• A clear escalation in investment scale is visible when comparing the $89 million deal for a Lenovo supercomputer fleet in 2024 to the landmark $500 million contract awarded to Elea Data Centers in October 2025.
• The 17-year term of the Elea data center contract signals a strategic departure from short-term compute leasing, locking in a long-term commitment to a proprietary, centralized AI processing environment.
• Further investments in 2025, including a tender awarded to Atos Bull for a supercomputer with next-generation NVIDIA Hopper and Blackwell GPUs and a $28 million hardware supply contract with Lenovo, reinforce this focus on building a comprehensive, state-of-the-art computational ecosystem.

Table: Key AI Infrastructure Investments by Petrobras (2024-2025)

How Petrobras’s Partner Ecosystem Supports its AI Infrastructure Build-Out

Petrobras is assembling a specialized partner ecosystem to construct and equip its AI infrastructure, deliberately selecting partners with deep expertise in HPC, data center construction, and cutting-edge processing technology. This focused approach ensures the company can build a purpose-driven environment optimized for its unique large-scale energy datasets.

• While partnerships before 2025 focused on acquiring compute hardware, such as the 2024 HPC fleet deal with Lenovo, the alliances in 2025 became more foundational and strategic in nature.
• The cornerstone partnership is the October 2025 agreement with Elea Data Centers, which is not a simple vendor contract but a long-term collaboration to build a custom, high-capacity AI facility from the ground up.
• To equip this facility, Petrobras is working with Atos Bull and NVIDIA to secure access to next-generation Hopper and Blackwell GPUs, ensuring its infrastructure remains at the technological forefront for complex AI model training.

Table: Key AI Infrastructure Partnerships of Petrobras (2024-2025)

Petrobras’s Geographic Focus: Centralizing AI Power in Brazil

Petrobras is deliberately consolidating its most powerful AI and HPC infrastructure within Brazil, creating a centralized hub to directly process proprietary data from its highly valuable pre-salt exploration and production operations.

• While earlier HPC deployments between 2021 and 2024, such as the Tatu supercomputer, were located in Brazil, the investments made in 2025 represent a significant centralization and scaling of this strategy.
• The selection of São Paulofor the new30 MVAdata center built byElea Data Centerssolidifies this approach, establishing the company’s primary AI factory in a key Brazilian economic hub, close to its corporate decision-making centers.
• This Brazil-centric infrastructure strategy enablesPetrobrasto maintain tight security and control over its sensitive geological data, ensuring its most valuable digital asset is processed in-country and in close proximity to its core offshore assets.

Technology Maturity: Petrobras’s Move to a Custom-Built, Industrial-Scale AI EnvironmentPetrobras’sAI infrastructure strategy has matured from procuring commercial off-the-shelf HPC systems to commissioning a custom-built, industrial-scale supercomputing environment designed for the future of AI in energy.

• During the 2021–2024 period, the company’s focus was on acquiring and deploying available technology, such as the Tatu supercomputer powered by commercial NVIDIA A100 GPUs, to meet existing computational demands.
• The year 2025marks a decisive shift toward a strategic build-out phase. TheElea Data Centersproject, with its17-yearterm, massive30 MVAcapacity, and specializedliquid coolingtechnology, is not a standard procurement but the creation of a purpose-built, strategic asset.
• This technological maturation is further validated by the company’s commitment to procuring next-generationNVIDIA Hopper and Blackwell GPUsthrough its tender withAtos Bull, ensuring its infrastructure is capable of handling the increasingly complex AI models required for future energy exploration and operations.

SWOT Analysis: Evaluating Petrobras’s Infrastructure-First AI StrategyPetrobras’sprimary competitive strength lies in its commitment to building a proprietary, powerful AI infrastructure, yet this capital-intensive strategy introduces a long-term dependency on a concentrated group of specialized partners.

• The decision to build a dedicated “AI factory” givesPetrobrasa significant advantage in processing its unique pre-salt data, creating a formidable competitive moat.
• However, the high capital outlay and long-term contracts, such as the17-yeardeal withElea Data Centers, create financial rigidity and increase reliance on key technology providers likeNVIDIA.
• This contrasts with the more agile, cloud-based strategies of competitors, who can adapt to technological shifts without the same level of sunk costs.

Table: SWOT Analysis for Petrobras’s AI Infrastructure Strategy

Petrobras Accelerates Reservoir Simulation with NVIDIA CPU

According to data from NVIDIA, Petrobras is speeding up its linear solvers for reservoir simulation. The chart highlights the specific solver scalability achieved on the Buzios dataset using the NVIDIA Grace CPU.

(Source: NVIDIA Developer)

SWOT Category	2021 – 2024	2025 – Today	What Changed / Resolved / Validated
Strength	Acquired individual AI-capable supercomputers like Tatu to enhance geophysical modeling capabilities.	Committed $500 million to a dedicated, long-term AI data center with Elea Data Centers and secured next-gen NVIDIA GPUs.	The strategy shifted from tactical hardware purchases to building a proprietary, industrial-scale “AI factory,” creating a durable competitive moat around its unique pre-salt data.
Weakness	Relied on vendors like Lenovo and NVIDIA for standard HPC system procurement.	Locked into a high-capital, 17-year contract with Elea Data Centers, increasing financial commitment and dependence on a few key partners.	The scale of investment increased both the potential reward and the risk, creating long-term dependency and reducing financial flexibility compared to asset-light cloud strategies.
Opportunity	Utilized HPC primarily to accelerate traditional exploration and production workflows.	New infrastructure unlocks the ability to apply large-scale AI across the entire value chain, from new energy ventures to corporate functions, evidenced by a $120 million tax automation saving.	The opportunity expanded from optimizing existing E&P tasks to enterprise-wide transformation, driven by the massive processing power of the new infrastructure.
Threat	Competitors were adopting AI through generalized cloud services.	Peers like NextEra Energy are deepening partnerships with flexible cloud providers like Google Cloud for AI and data platforms.	The threat from agile, cloud-native competitors has become more defined, as they can adopt new AI technologies without the sunk costs of a long-term, custom-built data center.

2026 Outlook: Translating Infrastructure Investment into Measurable Returns

The critical action forPetrobrasin the coming year is to demonstrate tangible returns from its massive infrastructure investment by scaling high-value AI applications and proving the competitive superiority of its proprietary “AI factory” model.

• The primary measure of success for the $500 million Elea data center and the newAtos Bullsupercomputer will be their ability to accelerate strategic R&D projects, such as thefour-yearreservoir characterization initiative withEliis, and deliver quantifiable efficiency gains in exploration.
• Investors will be closely watching for evidence thatPetrobrascan replicate the spectacular$120 million** ROI achieved with its tax automation project in other corporate and operational domains, validating the enterprise-wide potential of its AI investment.
• The company must decisively prove that its capital-intensive, infrastructure-first strategy yields a greater long-term competitive advantage than the more flexible, cloud-based AI approaches being pursued by industry peers.

Frequently Asked Questions

What is the main change in Petrobras’s AI strategy starting in 2025?
Before 2025, Petrobras’s strategy was to acquire individual high-performance computing (HPC) systems to solve immediate computational challenges. Starting in 2025, the strategy shifted to building a foundational, large-scale AI infrastructure. The centerpiece of this change is a $500 million, 17-year contract with Elea Data Centers to construct a dedicated AI data center, signaling a move from tactical purchases to building a long-term, proprietary “AI factory”.

Why is Petrobras investing $500 million in a new data center instead of using cloud services?
Petrobras is building its own dedicated AI data center to create a durable competitive advantage. This allows the company to process its unique and massive pre-salt geological data at an industrial scale, maintain tight security and control over its most valuable digital assets, and create a purpose-built environment that is difficult for competitors using generalized cloud services to replicate.

Who are the key partners helping Petrobras build its new AI infrastructure?
Petrobras has assembled a specialized ecosystem of partners. The key partners include Elea Data Centers, which is building and operating the new $500 million AI facility; Atos Bull and NVIDIA, which are supplying a new supercomputer with next-generation Hopper and Blackwell GPUs; and Lenovo, which provides supercomputer fleets and other supporting hardware.

What kind of advanced technology will be in the new Petrobras AI infrastructure?
The new infrastructure will be state-of-the-art. The data center built by Elea will have a massive 30 MVA capacity and will use specialized liquid cooling technology. To power its AI models, Petrobras is acquiring a new supercomputer from Atos Bull that will incorporate next-generation NVIDIA Hopper and Blackwell GPUs, ensuring it can handle complex AI training and inference tasks for years to come.

What is the main goal for Petrobras’s AI strategy in 2026?
In 2026, the primary goal is to demonstrate tangible returns on its massive infrastructure investment. Petrobras aims to scale high-value AI applications in areas like reservoir characterization and exploration to achieve quantifiable efficiency gains. The company needs to prove that its capital-intensive, proprietary AI factory model delivers a superior long-term competitive advantage compared to the more flexible, cloud-based approaches of its competitors.

Experience In-Depth, Real-Time Analysis

For just $200/year (not $200/hour). Stop wasting time with alternatives:

• Consultancies take weeks and cost thousands.
• ChatGPT and Perplexity lack depth.
• Googling wastes hours with scattered results.

Enki delivers fresh, evidence-based insights covering your market, your customers, and your competitors.

Trusted by Fortune 500 teams. Market-specific intelligence.

Explore Your Market →

One-week free trial. Cancel anytime.

Related Articles

If you found this article helpful, you might also enjoy these related articles that dive deeper into similar topics and provide further insights.

• E-Methanol Market Analysis: Growth, Confidence, and Market Reality(2023-2025)
• Battery Storage Market Analysis: Growth, Confidence, and Market Reality(2023-2025)
• Carbon Engineering & DAC Market Trends 2025: Analysis
• Climeworks 2025: DAC Market Analysis & Future Outlook
• Climeworks- From Breakout Growth to Operational Crossroads

AI Infrastructure AI in the energy sector Atos Bull supercomputer Elea Data Centers Energy digital transformation High-performance computing in oil and gas Petrobras AI Strategy Petrobras data center investment 2025 Petrobras NVIDIA Blackwell GPUs Supercomputing for seismic data