AI Revolution: How to Identify the Next NVIDIA in the Brazilian Market

In January 2023, one share of NVIDIA was worth around $14. By early 2026, the same stock was trading above $180. A gain of +1,190% in just over three years. In the fourth quarter of 2025, the company reported $68.1 billion in revenue — a figure that rivals the GDP of entire countries.

These figures are not curiosities. They are a signal about what happens when a technology stops being a promise and becomes infrastructure.

The relevant question for the disciplined investor is not "how much did NVIDIA rise." It is: where are the assets in Brazil that could follow a similar trajectory?

The size of the opportunity in numbers

The global artificial intelligence market moved $538 billion in 2026, growing at an annual rate of 37.3% according to industry projections. For context: this is a market that roughly doubles in size every two years.

Brazil is not on the sidelines of this movement. Brazilian spending on AI reached $2.4 billion in 2026, a 30% increase over 2024. A survey of Brazilian executives found that 78% of companies are expanding AI investments in the current period.

At the federal level, the government launched the Brazilian Artificial Intelligence Plan with a commitment of R$ 23 billion through 2028. Part as subsidy, part as public procurement, part as infrastructure. But the practical effect is the same: creating predictable demand and volume for companies that are positioned in the sector.

Why NVIDIA went so far beyond expectations

Understanding the NVIDIA case helps calibrate what to look for in Brazilian companies.

NVIDIA was not just a chip company. It was the company that built the infrastructure. GPUs are the building blocks of AI: without them, language models, computer vision, and recommendation systems cannot run at scale. When demand for AI exploded, demand for GPUs exploded with it. And because NVIDIA controlled the entire ecosystem — hardware, software, and development tools — the margin it captured was disproportionate.

The lesson: companies that capture disproportionate value in technology cycles tend to be those that control the infrastructure layer or the critical integration points, not necessarily those that develop the most visible end application.

In the Brazilian context, the equivalent is not a chip company. It is a company that controls the enterprise management software that needs to be updated, the financial platform that needs to integrate AI, or the data infrastructure provider that underpins the applications above it.

Investment angles in Brazil

TOTS3: control of the management layer

TOTVS is the largest enterprise management software company in Latin America, with a presence in more than 120,000 Brazilian companies. In 2025, the company committed R$ 600 million to AI, and the results showed: net income grew 26% in the year.

TOTVS's business model is structurally favorable to the AI cycle. It does not need to convince companies to adopt management software — they already use it. It needs to convince those same companies to upgrade their systems to AI-integrated versions. The switching cost is high for the customer, which creates recurring and predictable revenue as adoption grows.

TOTS3 operates at the management layer. Whoever controls the ERP controls where the data lives. And data is the primary input for any AI application.

NVDA: still with structural room

Brazilian investors have access to NVDA through BDRs (NVDC34) or international accounts. The question is not whether NVIDIA has already risen too much — that is a timing question that any honest analyst will admit cannot be answered with certainty. The structural question is different: the chip's role in the AI ecosystem remains irreplaceable over the relevant horizon.

Competitors like AMD and Intel have advanced, but the advantage of the CUDA ecosystem — the development platform that locks researchers and companies into NVIDIA's architecture — is difficult to replicate over short cycles. This is not a performance guarantee. It is a variable that the fundamental investor needs to weigh.

Nubank and AI-driven financial infrastructure

Nubank (NU) is a different case. It is not an AI company, but it is a company that uses AI as a competitive advantage at scale. Credit models, fraud detection, product personalization — each of these vectors impacts margin and customer retention.

With more than 110 million customers across Latin America, the volume of data Nubank processes is a real competitive barrier. Traditional financial companies have comparable data volumes, but rarely have the engineering capacity to extract value at the same pace.

Data infrastructure and connectivity

A less-discussed but equally relevant layer is the infrastructure that underpins all of this: data centers, high-speed connectivity, and cloud services. In Brazil, companies such as Oi Soluções, Locaweb, and regional colocation players are positioned to capture derived demand — not from the AI product itself, but from the need to process and store data growing at an exponential rate.

How to identify companies with real potential

The most useful question an investor can ask when analyzing an "AI company" is: does it use AI as a marketing argument or as a driver of results?

The test is straightforward. Look for answers to three questions in the financial statements and earnings calls:

1. Is the margin improving? If AI is being used for operational efficiency, this should appear in gross or operating margins over two to four quarters. A narrative without margin improvement signals that the technology is still a cost, not a lever.

2. Is recurring revenue growing? Companies that embed AI into existing products (like TOTVS migrating customers to newer versions) tend to increase ARR (annual recurring revenue) without needing to acquire new customers. This is the cleanest indicator of real adoption.

3. Is management allocating capital consistently? A R$ 600 million commitment to AI, like TOTVS's, is a signal. The question is whether the plan is being executed — and that is visible quarter by quarter, not in a press release.

The risk that most people ignore

Technology cycles create three types of companies: those that define the infrastructure, those that build on top of the infrastructure, and those that are replaced by the infrastructure.

In Brazil, the greatest risk is not missing out on the AI sector. It is buying companies that appear to be beneficiaries but are actually being disintermediated by the cycle. Sectors such as manual accounting, unstructured data services, and traditional customer service are subject to margin compression as AI tools reduce the unit cost of these activities.

Correctly identifying which side of that line a company sits on — beneficiary or disintermediated — is the central analytical work of this cycle.

The disciplined approach

At Royal Binary, the analysis of technological opportunities follows the same method that Sidnei Oliveira applies to any investment thesis: data first, narrative second.

The AI cycle is real. The sector's growth figures, corporate commitments, and government investment are verifiable. What is not automatic is the translation of sector growth into investor return — that part depends on knowing which assets convert growth into profit, and which merely convert it into cost.

The difference between identifying NVIDIA in 2023 and buying any technology stock is precisely that: understanding who controls the infrastructure, who has recurring revenue, and who has the switching cost that prevents customers from leaving. These are criteria that feel anachronistic against the "revolution" narrative, but they are the ones that determine returns.

Past results do not guarantee future returns. Returns are variable income.