The Cost of Opacity
There is a notable gap in the Pix Direto content market: almost nobody talks about real numbers. Most articles and presentations use expressions like 'significant investment', 'considerable costs' or 'robust infrastructure' — terms that convey seriousness, but do not help anyone make a decision.
This opacity has a concrete cost. Fintechs that choose the direct model without clarity on the real TCO tend to underestimate CAPEX, discover OPEX the hard way — and, in some cases, reach go-live with an operation whose monthly cost invalidates the margin that motivated the migration.
This article attempts to be the opposite of that. We present cost ranges based on market references for each investment component — implementation CAPEX, recurring OPEX, regulatory and operational costs. And we close with a comparative TCO simulation between the direct and indirect models, for two real fintech profiles, which clearly shows when the ROI adds up and when it doesn't.
⚠️ Methodological warning: The values presented in this article are market reference ranges for Brazil in 2024–2025, based on cost compositions of similar projects. They are not quotes. Each project has specific variables — technical maturity of the team, existing stack, scope of features at go-live, and build vs. buy decisions — which can move the numbers significantly up or down.
📌 Context: If you are still deciding whether the direct model makes sense for your fintech, we recommend starting with the strategic comparison in the article Direct Pix vs. Indirect Pix: which model makes more sense for your fintech? — which includes the decision matrix by company profile before getting to the numbers.
CAPEX: o custo de implementação
O CAPEX do Pix Direto é o investimento necessário para construir e certificar a infraestrutura até o go-live. Ele cobre desenvolvimento técnico, integrações obrigatórias, infraestrutura cloud, segurança e o processo de homologação no ambiente do Banco Central.
A tabela abaixo detalha cada componente em dois cenários: conservador (build interno com time experiente, escopo de go-live enxuto) e agressivo (build interno com curva de aprendizado, escopo amplo de features no lançamento).
Componente | O que inclui | Cenário conservador | Cenário agressivo |
|---|---|---|---|
Infraestrutura cloud (setup inicial) | Provisionamento de ambientes redundantes (produção + DR), load balancers, redes privadas, zonas de disponibilidade. | R$ 80k – R$ 250k | R$ 150k – R$ 600k |
Desenvolvimento da camada SPI / ISO 20022 | Implementação do protocolo de mensageria, parsers, validadores, handlers de erro e retry logic. | R$ 200k – R$ 500k | R$ 500k – R$ 1,5mi |
Integração DICT | APIs de registro, consulta e portabilidade de chaves Pix. | R$ 80k – R$ 200k | R$ 200k – R$ 500k |
Integração RSFN | Contratação de provedor homologado, configuração de rede dedicada, testes de conectividade. | R$ 40k – R$ 100k | R$ 100k – R$ 250k |
Sistema de antifraude | Desenvolvimento ou licença de solução de prevenção à fraude + integração ao mecanismo FRAUD do BC. | R$ 100k – R$ 300k | R$ 300k – R$ 800k |
Certificados ICP-Brasil | Aquisição e setup de certificados digitais para autenticação nas mensagens ao BC. | R$ 10k – R$ 30k | R$ 30k – R$ 80k |
Observabilidade e monitoramento | Stack de logs, métricas, alertas e dashboards (ex: Datadog, Grafana, ELK). | R$ 40k – R$ 100k | R$ 100k – R$ 300k |
Segurança e pen test | Auditorias de segurança, testes de intrusão, gestão de vulnerabilidades. | R$ 50k – R$ 150k | R$ 150k – R$ 400k |
Homologação e certificação BC | Horas de engenharia dedicadas ao ambiente de homologação do BC, correções e retestes. | R$ 80k – R$ 200k | R$ 200k – R$ 500k |
TOTAL CAPEX ESTIMADO | Build interno, sem fornecedor de plataforma | ~R$ 680k – R$ 1,8mi | ~R$ 1,7mi – R$ 4,9mi |
💡 O que move o CAPEX: O fator que mais impacta o custo de implementação é a experiência do time com mensageria financeira e ISO 20022. Um time que nunca trabalhou com esse protocolo vai levar 2–3x mais tempo na integração ao SPI do que um time com experiência prévia — o que se traduz diretamente em custo de horas de engenharia.
O que o CAPEX não inclui
As faixas acima cobrem o custo técnico da implementação. Elas não incluem:
Custo do processo regulatório de autorização do BC: assessoria jurídica especializada para preparação da documentação e acompanhamento do processo. Faixa adicional estimada: R$ 80k – R$ 300k dependendo da complexidade do caso.
Horas de gestão de projeto: PM dedicado, scrum master, stakeholder management. Em projetos de 18–24 meses, esse overhead é relevante.
Treinamento e capacitação: curva de aprendizado do time em SPI, ISO 20022 e regulação do BC.
Custo de oportunidade: horas de engenharia que poderiam estar em outros produtos durante os 18–24 meses de implementação.
OPEX: o custo mensal recorrente
If CAPEX is the cost that appears in the initial business case, OPEX is the cost that determines whether the operation is sustainable in the long term. And this is where most initial estimates fail: the OPEX of Direct Pix is not just cloud infrastructure — it is the sum of several components that, individually, seem reasonable, but together form a significant cost line.
Component | What is included | Conservative scenario | Aggressive scenario |
|---|---|---|---|
Cloud infrastructure (recurring) | Compute, storage, network, redundancy, active DR. | BRL 30k – BRL 80k/month | BRL 80k – BRL 250k/month |
Technical team (engineering + SRE) | 2–4 dedicated engineers for the Pix platform + structured on-call. | BRL 40k – BRL 100k/month | BRL 100k – BRL 300k/month |
Compliance and risk | Compliance analyst(s) dedicated to Pix operations, regulatory reporting. | BRL 15k – BRL 40k/month | BRL 40k – BRL 120k/month |
Anti-fraud (operations) | Continuous monitoring, model tuning, case review, integrations. | BRL 15k – BRL 50k/month | BRL 50k – BRL 150k/month |
Observability tools | Monitoring licenses, APM, log management. | BRL 5k – BRL 20k/month | BRL 20k – BRL 60k/month |
Certificates and RSFN | Renewal of ICP-Brasil certificates + monthly fee for RSFN provider. | BRL 3k – BRL 8k/month | BRL 8k – BRL 20k/month |
Auditing and legal advice | Periodic security and compliance audits, regulatory consulting. | BRL 10k – BRL 30k/month | BRL 30k – BRL 80k/month |
TOTAL ESTIMATED OPEX | Per month, in full operation status | ~BRL 118k – BRL 328k/month | ~BRL 328k – BRL 980k/month |
The invisible cost: the dedicated technical team
The most frequently underestimated OPEX component is not infrastructure — it is the team. Operating Direct Pix with a 99.5% SLA requires engineers who deeply understand the platform and are available to act at any moment. This is not compatible with a model where Pix is just one more responsibility for an overloaded platform team.
The market reference is 2 to 4 dedicated (or heavily allocated) engineers for the Pix platform — plus an SRE or equivalent for structured on-call. Factoring these headcounts into the OPEX from the start is essential for the business case to reflect operational reality.
The cost of continuous compliance
Compliance is not a project — it is a permanent function. BCB Resolution No. 429/2024 and its future updates, periodic reports to the BC, fraud monitoring, and the management of indirect participants (if any) require the continuous dedication of at least one specialized compliance analyst. In smaller fintechs, this function is usually accumulated with other responsibilities — which works until a regulatory incident reveals the cost of that decision.
🔗 In-depth: The technical requirements that drive these OPEX costs — SPI, DICT, RSFN, anti-fraud, and 24/7 availability — are detailed in the article How to become a direct participant in Pix: technical and regulatory requirements.
Build vs. Buy vs. Parceiro tecnológico: qual abordagem tem melhor TCO?
The decision of how to implement Pix Direto — whether to build internally, license a specialized platform, or partner with a provider that delivers the direct model as a service — directly impacts CAPEX, timeline, and the project's risk profile.
There is no universal answer. The right approach depends on the institution's profile, the size of the technical team, the available budget, and — most importantly — long-term strategic ambition.
Criterion | Build (internal) | Buy (platform) | BaaS / Full Partner |
|---|---|---|---|
Implementation CAPEX | BRL 680k – BRL 4.9m | BRL 200k – BRL 800k (license + setup) | BRL 1.5m – BRL 5m+ (all-inclusive) |
Time to go-live | 18–24 months | 10–16 months | 12–20 months |
Technical risk | High — depends on the maturity of the internal team | Medium — shared with vendor | Low — vendor assumes infrastructure responsibility |
Customization | Total | High, within platform scope | Contract-dependent |
Asset ownership | 100% internal | Partial — internal code + licensed platform | Zero — outsourced service |
Recurring cost (OPEX) | High — internal team + own infrastructure | Medium — license + reduced team | Variable — usually fee-per-transaction or monthly fee |
Vendor lock-in | Zero | Medium — critical platform | High — vendor is the critical path |
Ideal for | Large FIs with a robust team and BaaS platform ambitions | Fintechs at scale with technical team and relevant go-live urgency | Fintechs wanting a direct model without building infrastructure |
When build makes sense
Building internally makes sense when the institution has a robust technical team with experience in mission-critical financial systems, when the ambition is to become a BaaS platform and offer Pix as a service to third parties — which requires full control over the architecture — and when the budget allows absorbing a higher CAPEX in exchange for full ownership of the asset.
When buy makes sense
Licensing a specialized platform makes sense when the go-live timeline is a critical variable, when the technical team has integration capabilities but not from-scratch building capacity, and when the necessary customization is within the scope of what the available vendors offer.
When a full partner makes sense
Hiring a partner that delivers the direct model as a service makes sense when the priority is to reach the direct model with the lowest possible operational risk — even if this implies vendor dependency and higher recurring costs. It is the most suitable option for fintechs that need the direct model for strategic or product reasons but lack the internal capacity to build and operate the infrastructure.
⚠️ Common Pitfall: Deciding to build based on technical pride — without conducting a formal RFI of platforms and partners available in the market. The opportunity cost of building everything from scratch when mature platforms are available can mean months of delay and millions of BRL in additional CAPEX.
Simulação de TCO: quando o ROI efetivamente fecha
The most important question in the financial analysis of Direct Pix is not 'how much does it cost?' — it is 'at what volume does the cost of the direct model become lower than the cost of the indirect model?'. This is the break-even analysis that defines if and when the migration makes economic sense.
The simulation below compares two fintech profiles over a 36-month horizon, using the conservative CAPEX and OPEX scenario from the previous tables:
Variable | Fintech A — 5M transactions/month | Fintech B — 20M transactions/month |
|---|---|---|
Monthly Pix transaction volume | 5 million/month | 20 million/month |
Current average fee (indirect model) | R$ 0.04/transaction | R$ 0.03/transaction |
Current monthly cost (indirect model) | R$ 200,000/month | R$ 600,000/month |
Estimated CAPEX (conservative build) | R$ 1.2M | R$ 2.5M |
Estimated monthly OPEX (direct model) | R$ 180,000/month | R$ 320,000/month |
Monthly savings with direct model | R$ 20,000/month | R$ 280,000/month |
CAPEX Payback (months) | ~60 months (5 years) | ~9 months |
TCO over 36 months — indirect model | R$ 7.2M | R$ 21.6M |
TCO over 36 months — direct model | R$ 7.68M (CAPEX + OPEX) | R$ 14.02M (CAPEX + OPEX) |
Verdict over 36 months | 🔴 Indirect is still cheaper | 🟢 Direct saves ~R$ 7.6M |
What the simulation reveals
For Fintech A, with 5 million transactions per month, the direct model does not yet achieve ROI within 36 months. This does not mean the decision is wrong — it means the decision needs to be justified by other drivers beyond cost: product autonomy, locked roadmap, platform strategy. If the only argument is fee savings, the volume does not support it yet.
For Fintech B, with 20 million transactions per month, payback occurs in about 9 months — and cumulative savings over 36 months exceed R$ 7.5 million. In this scenario, the question is not 'is it worth it?' but rather 'why haven't we started yet?'.
Variables that move the break-even
Current partner fee: the higher the fee charged today, the faster the break-even. Renegotiating the fee before deciding on the direct model can completely change the analysis.
Volume growth rate: fintechs with accelerated transaction growth see the break-even approach quickly. TCO should be calculated with projected volume, not just current volume.
Build vs. buy decision: a specialized platform can reduce CAPEX by 40–60% compared to internal build, significantly accelerating break-even.
Monetization of indirect participants: if the institution becomes a reference direct participant for other fintechs, the revenue generated by this model completely changes the financial equation — and is not included in this simulation.
📊 Important: This simulation uses an average reference fee. Your current partner's actual fee may be significantly different. Before using these numbers in an internal business case, replace the average fee with your actual cost per transaction — the difference can be decisive.
O custo que não aparece na planilha
Every TCO analysis captures direct costs reasonably well. What rarely shows up — but is often decisive — is the opportunity cost of indirect model dependency.
Revenue not generated due to non-existent features
Every feature that the partner does not offer and that blocked a launch has a revenue cost. Dynamic billing that wasn't implemented, payment split that the partner didn't support, scheduled Pix with customized logic that sat on the roadmap for 18 months. This value is rarely quantified — but in fintechs where the financial product is core, it can exceed the operational cost of the direct model.
Customers lost due to inadequate SLA
Incidents with the partner that brought down your operation have an NPS, churn, and reputation cost that does not appear in OPEX. Every hour of downtime at a critical time — weekend, end of the month, high-volume dates — has a real cost that needs to be considered in the analysis.
Lost bargaining power
A fintech that relies on a single direct participant to operate its Pix has no real bargaining power when renegotiating fees. The cost of maintaining this position over time — in terms of fees paid above what would be possible in a scenario of alternatives — is cumulative and silent.
Conclusão: o custo real é o custo de não saber o custo
A análise de TCO do Pix Direto raramente é simples — mas é sempre possível. O que inviabiliza decisões boas não é a complexidade dos números, mas a ausência deles.
Para Product Managers e Analistas que estão conduzindo essa análise internamente, a recomendação prática é construir o modelo financeiro com as faixas apresentadas neste artigo, substituindo os valores de referência pelos custos reais onde você tiver acesso — fee atual do parceiro, custo-hora do time de engenharia, orçamentos de infraestrutura cloud. Cada variável que você torna concreta melhora a qualidade da decisão.
E se a análise mostrar que o volume atual ainda não suporta o modelo direto economicamente, isso não encerra o debate — significa que a decisão precisa ser avaliada com o volume projetado, não o atual, e com os vetores estratégicos (autonomia de produto, estratégia de plataforma) pesando ao lado do custo.
🚀 Próximo passo: Com o business case financeiro mapeado, o próximo nível de aprofundamento é entender a arquitetura técnica que gera esses custos — SPI, DICT, Conta PI e o fluxo completo de uma transação Pix. Esse é o tema do próximo artigo: SPI, DICT e Conta PI: a infraestrutura por trás do Pix Direto explicada.
Leituras relacionadas neste cluster
→ Direct Pix: what it is, how it works, and when it is worth it
→ Direct Pix vs. Indirect Pix: which model makes more sense for your fintech?
→ How to become a direct participant in Pix: technical and regulatory requirements
→ SPI, DICT, and PI Account: the infrastructure behind Direct Pix explained
