Brazil is about to get a whole lot more AI venture capital. Watch this.

 This Week’s Essay

For the past 24 months, AI capital flowed into a very small number of companies.

OpenAI. Anthropic. xAI. Google. Meta.

More GPUs. More data centers. More capital. More concentration.

And for a while, it made perfect sense. If you believed foundation models would become winner-take-most platforms, then concentration at the model layer was rational. The best labs attracted the best talent, the best GPUs, the best data partnerships, and eventually the best distribution.

But the power law at that layer is now drawn.

The next marginal dollar into another foundation model is not where the 10x lives.

The 10x lives in vertical AI. Embedded into workflows. Where the moat is compliance, proprietary data, distribution, customer relationships, and the operational infrastructure built around them.

And if that thesis is right, Brazil becomes one of the most important AI markets in the world.

—  —  —

The foundation model race is becoming infrastructure.

One of the most under-discussed shifts in AI right now is that frontier models are improving rapidly — but the gap between them is compressing.

For most enterprise use cases, “good enough” intelligence is arriving extremely fast.

That changes the economics. When models become abundant, differentiation moves elsewhere: proprietary data, distribution, embedded workflows, compliance layers, operational integration, customer trust, speed of deployment.

The value migrates from intelligence itself → toward applied intelligence.

This is why the smartest capital in the world is suddenly obsessed with vertical AI.

Not generic copilots. Not wrapper companies. Not “AI for everything.”

Deeply embedded systems built around messy industries. Legal. Healthcare. Insurance. Accounting. Logistics. Financial operations. Industrial workflows.

The categories where data is fragmented, regulation is painful, processes are manual, and incumbents still run on PDFs, WhatsApp groups, Excel files, and human coordination.

That’s exactly where Brazil becomes globally relevant.

—  —  —

Brazil’s dysfunction is becoming an AI moat.

For years, global investors viewed Brazil’s complexity as a tax. Too regulated. Too fragmented. Too operationally messy.

AI inverts that equation.

Modern AI systems thrive in environments with repetitive operational workflows, fragmented data, unstructured information, human-heavy coordination, localized regulation, and high labor intensity.

Brazil has all of it at massive scale:

• One of the world’s most complex tax systems

• Hyper-regulated financial and healthcare sectors

• Enormous SMB fragmentation

• Deeply offline operational workflows

• Massive legal infrastructure

• Extremely high WhatsApp penetration

• Millions of businesses still operating with limited software penetration

What traditional software viewed as friction, AI views as training terrain.

The inefficiency itself becomes the moat.

And the strongest vertical AI businesses won’t look like SaaS. They’ll look like operational infrastructure — workflow monopolies, decision engines, embedded financial rails, AI-native coordination systems. The product won’t assist work. It will execute it.

That matters enormously in Brazil, because many industries skipped software modernization entirely. AI isn’t competing against best-in-class software stacks here. It’s competing against operational chaos.

Which is where the leapfrogging happens.

—  —  —

Brazil has distribution rails most countries don’t.

Global investors still underestimate how advanced Brazil’s financial and communication infrastructure has become — especially relative to other emerging markets.

Pix changed the speed of money movement in the country. WhatsApp became the interface layer for commerce. Open Finance is creating interoperable financial data infrastructure.

This combination matters more than people realize. Because AI products scale faster when customer acquisition is cheap, communication is consolidated, payments are instant, financial data is interoperable, and users already live inside digital ecosystems.

Brazil checks all five boxes.

Brazilian founders can iterate faster, distribute faster, collect workflow data faster, automate faster, and monetize faster than many peers globally — especially in SMB-heavy verticals.

—  —  —

AI is also compressing the historical disadvantages of emerging markets.

This may be the most important point of all.

Historically, building globally relevant software companies from emerging markets was structurally harder. Talent density was lower. Capital access was weaker. Distribution was fragmented. Engineering costs were high. Scaling teams required enormous capital.

AI compresses every one of those disadvantages.

Small teams now produce software leverage that previously required hundreds of engineers. That changes the math for any country with strong technical talent, lower operating costs, ambitious founders, and improving infrastructure.

Brazil fits the profile.

And the smartest investors already know it.

The obvious example is Enter, a Brazilian legaltech automating legal defenses for large enterprises. It just raised $100M+ at a $1.2B valuation — 3.5x in roughly seven months — led by Founders Fund, with Sequoia and Ribbit Capital. That makes Enter Latin America’s first AI unicorn.

Then there’s Comp, a Brazilian HR tech startup that just landed Khosla Ventures’ first-ever investment in Brazil.

On the surface, this looks like the 2021 LATAM boom all over again.

It isn’t.

That cycle was driven by zero rates, crossover momentum, fintech multiple expansion, and macro liquidity.

This cycle is thesis-driven.

The investors leaning into Brazil right now are tier-1 leads. They’re underwriting vertical-specific theses, not “Brazil exposure.” They’re infrastructure-oriented. And they’re writing checks in a higher-rate environment — which means they’ve actually underwritten the returns.

That last point matters more than the first three combined.

—  —  —

The real opportunity isn’t local champions.

Most people still underwrite Brazil with an outdated framework: large domestic market → local winner.

The more interesting outcome is the emergence of globally exportable AI companies built from Brazil.

Especially in categories where Brazil already has operational depth: fintech infrastructure, healthcare operations, legal workflows, SMB financial management, logistics coordination, industrial systems, insurance infrastructure, HR and payroll complexity.

Because if you can build AI systems that survive Brazil’s complexity, they can probably survive almost anywhere.

Brazil isn’t a local market story anymore.

It’s an AI proving ground.

That’s a very different investment thesis.

—  —  —

What happens next.

US VCs aggressively lead Brazilian vertical AI rounds across legal, HR, healthcare ops, financial ops, and logistics.

Founders build for global markets from day one — not Brazil first, then expand.

Brazil becomes a net exporter of AI companies.

It’s a very interesting time to invest in Brazil tech.

Watch the cap tables.

This Week’s Number

$ 725 billion.

That’s what the four largest hyperscalers — Amazon, Microsoft, Alphabet, and Meta — are now projected to spend on capex in 2026 alone. Up 77% from last year’s record $410B. Roughly 75% of it ($545B) is going directly to AI infrastructure: GPUs, data centers, power, cooling.

For context: that single-year figure is comparable to the entire annual GDP of Belgium or Sweden — and exceeds that of Argentina or Norway.

Capital intensity at Oracle (86%), Meta (54%), and Microsoft (47%) now resembles utilities and industrials more than software companies. UBS estimates this level of capex will consume nearly 100% of operating cash flows in 2026 — versus a 10-year average of 40%.

The AI trade isn’t a software trade anymore.

3 Major Shifts In AI This Week I’m Paying Attention To

1. Long Lake just bought Amex Global Business Travel for $6.3B in cash.

Long Lake — an AI-focused buyout vehicle backed by General Catalyst, Alpha Wave, Elad Gil, D1, Thrive, and Koch Equity — agreed to take Amex GBT private in an all-cash deal at a 60% premium. The transaction closes in H2 2026.

The thesis writes itself.

Corporate travel is one of the largest enterprise workflows in the world. It’s fragmented, messy, data-rich, labor-intensive. Constant disruptions, approvals, policy enforcement, expense reconciliation.

In other words: perfect terrain for AI agents.

Long Lake is effectively betting that travel booking becomes autonomous, disruption management becomes AI-native, and more travel work gets handled by software agents rather than humans.

And Amex GBT already has the substrate that makes that bet work — enterprise distribution, embedded workflows, customer relationships, transaction data, infrastructure in 140+ countries.

The broader pattern: AI capital is moving from “model companies” toward workflow monopolies with data.

2. Anthropic’s secondary market valuation just blew past $1 trillion — at times higher than OpenAI.

The numbers: Anthropic closed its February 2026 round at $380B post-money. Three months later, secondary markets are pricing it at roughly $1T on Forge Global. Some bidders have offered $1.15T. OpenAI, by comparison, is trading at $880B on the same platform — barely above its $852B primary.

The repricing is being driven by one number: annualized revenue jumped from $9B at the end of 2025 to $30B by March 2026. A 233% increase in a single quarter, fueled largely by Claude Code adoption.

The underrated part is what investors are actually buying.

The market is starting to value AI companies based on economic architecture, not just model quality. Who becomes cash-flow positive. Who controls enterprise workflows. Who monetizes agents at scale. Who secures compute long term.

The AI race is no longer just “who has the smartest model.” It’s increasingly: who controls compute, who controls enterprise distribution, and who can finance infrastructure at unprecedented scale.

Whether $1T is rational is a separate question. The repricing is real.

3. Coatue launched Next Frontier — to buy land and build the physical substrate underneath AI.

Coatue Management — a $70B firm with major positions in Anthropic, OpenAI, xAI, and CoreWeave — just launched a new vehicle called Next Frontier. Its mandate: acquire land near major power sources and convert those parcels into AI data centers. Tens of billions in planned spend.

Its first project is a campus in Indiana, in joint venture with Fluidstack — the cloud infrastructure startup that recently signed a $50B deal to build data centers for Anthropic.

That’s a gigantic shift. One of the most sophisticated tech investors in the world is moving from investing in AI companies → to owning the physical substrate underneath them.

Because the bottleneck in AI is no longer just software.

The bottlenecks are now: power, land, cooling, chips, grid access, permitting.

The AI trade is becoming an industrials trade.

This is why Blackstone, Brookfield, Coatue, sovereign wealth funds, and infrastructure investors are suddenly moving aggressively into power and data centers. Everyone wants exposure to foundation models. But some of the most durable money may accrue to energy, compute infrastructure, cooling, grid optimization, and networking.

The AI boom is starting to look less like SaaS 2.0 and more like railroads + telecom + industrial buildout.

What I’m Loving 

“AI Founder Mode” — Brian Chesky on Invest Like the Best. A few ideas worth the listen: the next wave of AI will be consumer, not enterprise — and there are two types of professionals who won’t survive the AI transition: pure people managers, and leaders who resist getting back into the details. They also dig into Brian’s eleven-star experience exercise as a path to product-market fit, why founders rarely make good early CEOs, and what changed for him when he stopped chasing adulation and started making things for the love of making them. Lots of arguments here that cut against the prevailing wisdom.

Thanks for reading,

Olga 

 P.S. If this issue was valuable to you please share it with a founder who needs to hear it. Let’s build LATAM’s next tech leaders—together

🎙 The J Curve  is where LATAM's boldest founders & investors come to talk real strategy, opportunity and leadership.