Brazilian banks are no longer just adopting AI, they are building it. The decision to develop proprietary AI platforms is forcing a fundamental rethink of on-premises data center infrastructure, and the window to act is now.
In 2025, Brazilian banks' total technology budget is expected to reach R$47.8 billion, with investment in AI, analytics, and big data growing 61% compared to 2024. Globally, the scale of this shift is equally striking: 96% of the 650 CIOs and CISOs surveyed in the State of the Application Strategy 2025 report are already developing, or initiating, projects to build their own AI models. The era of simply consuming AI tools is giving way to the era of owning them.
This transition is not incremental. It is structural. At Vertiv, we are seeing this shift firsthand: financial institutions across Brazil are requesting our expertise not just to maintain existing infrastructure, but to fundamentally redesign it for AI workloads.
What Banks Are Building Toward
Proprietary AI platforms require dedicated budgets, specialized teams, and multiple cycles of testing and refinement. The end goal is a hybrid AI model — one where connections to public AI are protected through controlled gateways — enabling institutions to drive productivity while delivering better experiences to customers, partners, and employees.
The business case is clear. In the finance sector, in-house AI platforms are targeting three high-value outcomes:
Greater efficiency and cost savings: AI optimizes loan processing, fraud detection, and customer service.
Improved risk management: AI algorithms identify patterns across large datasets, reducing defaults, lowering financial provisions, and improving profit margins.
Increased revenue generation: AI enables personalization of financial products and services at the scale of millions of transactions per second, driving customer satisfaction and loyalty.
The financial sector spent $35 billion on AI in 2023. That figure is expected to reach $97 billion by 2027. According to the World Economic Forum, 70% of global financial executives believe AI is critical for business innovation over the next five years.
Scaling these outcomes demands critical digital infrastructure built for the power and thermal intensity of AI factories. High-density AI workloads are redefining requirements for power distribution, liquid cooling, and systems integration. A system-level approach can help improve energy and thermal efficiency, strengthen resilience through infrastructure designed to operate as a unified ecosystem, and reduce deployment timelines to accelerate revenue-generating AI initiatives.
AI Platforms Are Rewriting the Infrastructure Equation
Financial institutions’ journeys toward AI integration are increasingly built on hybrid infrastructure. According to the F5 2026 State of Application Strategy Report: Financial Services, financial organizations continue to distribute applications across on-premises environments, public and private clouds, SaaS, colocation facilities, and edge locations. Where business-critical applications run is therefore increasingly fluid, shaped by factors such as flexibility, resilience, service optimization, regulatory requirements, data sovereignty, and latency.
According to the FEBRABAN-Deloitte Banking Technology Survey 2026, Brazilian banks' technology budgets are expected to reach R$50.4 billion this year, an 8% increase over 2025. Of this amount, approximately R$3 billion is earmarked for AI, analytics, and big data, also representing an 8% increase year over year. Globally, the F5 2026 State of Application Strategy Report reveals that 78% of organizations now operate their own AI inference infrastructure, highlighting the growing importance of internally managed AI capabilities. The shift toward greater control over AI workloads is reshaping infrastructure strategies.
On-premises Data Center Retrofit Projects
This reality is prompting retrofit projects in on-premises data centers in the Brazilian banking sector. The era of the "blank check" for banks to adopt public cloud is, in some cases, giving way to a new look at the role of the institution's own data center. The business decision to develop its own AI platform consolidated the need to renovate the financial institution's on-premises data center.
Vertiv is actively supporting retrofit projects in Brazilian banks, replacing previous generations of critical power and cooling infrastructure with solutions engineered for the density and efficiency demands of AI applications. These are technologies that continue to work for traditional compute loads but were not developed to handle the high data processing density typical of AI applications. Today, a bank's on-premises data center might handle 5 kW per rack, in the medium term it could reach 10, 20, or 25 kW. In a few years, it could reach 100 kW.
As this data center's profile is to support corporate AI applications, not large-scale public AI platforms, the data processing demand is lower. When proprietary AI platforms expand, banks may need to adopt higher capacity power systems and liquid cooling systems to support the high-density requirements of AI applications.
The Challenge of Reducing Energy Consumption
A key aspect of the transformation banks are currently undergoing is energy consumption reduction. Lowering the PUE metric is critical in this evolution. PUE, or Power Usage Effectiveness, is a measure used to evaluate data centers' energy efficiency, calculating the relationship between total energy used by the data center and energy used by IT equipment. The closer to 1 the PUE value is, the more efficient the data center, indicating that a smaller proportion of total energy is spent on support infrastructure (like cooling) and a larger proportion is used for IT equipment operation. Tracking a single data center's PUE over time reveals the facility's performance. According to the Uptime Institute Global Data Center Survey 2026, the industry-wide average annual PUE reached 1.52, reflecting gradual improvements in data center energy efficiency.
Best business and environmental practices have led data center managers to seek to lower PUE. This applies to financial sector on-premises data centers and their AI workloads. Data center retrofit can include solutions, such as Uninterruptible Power Supply (UPS) systems and cooling systems that can help provide energy savings compared to previous-generation power and cooling systems. In addition, a 360º service approach starting from the design phase through the commissioning phase, and 24x7 monitoring, maintenance, and lifecycle services, can help optimize infrastructure performance and provide additional efficiency gains through system design, proactive support and ongoing optimization.
This evolution is happening right now in Brazil and marks a new era for banks' on-premises data centers. Critical digital infrastructure models designed to support high-density AI applications are coming into play. In this journey, Vertiv ™ solutions and services can help to accelerate the innovation that AI brings to banking. The goal is to help support the competitiveness of organizations that sustain Brazil's economy.
About the author
Felipe Vasco da Silva, Technology Applications Director at Vertiv LATAM
With 16 years of experience in technology, data center, and energy sectors, Felipe Vasco adopts a consultative approach with the sales team, going beyond Vertiv™ solutions and creating value-added proposals that impact overall data center management. In his role, he is responsible for Vertiv LATAM Technology Applications. His main tasks include guiding and monitoring team performance and sales indicators, as well as evaluating and optimizing best practices. Before joining Vertiv, he worked at Schneider Electric, Green4T, and Aceco TI. Felipe holds a degree in Electrical Engineering with emphasis in Electrotechnics from UNIP (Universidade Paulista). Additionally, he has a Post-Graduate degree in Project Management from Mackenzie Presbyterian University and an International ATD (Accredited Tier Designer) Certification from the Uptime Institute.
