Databricks is staffing São Paulo and Mexico City for a product that competes with Oracle and SQL Server

The LATAM Bet Nobody Noticed

Databricks reported over 80% year-over-year growth in its LATAM business in October 2024. By July 2025 that figure had accelerated past 150%. Brazil alone grew more than 150% over two years, the local team more than doubled and is projected to clear 200 employees by the end of 2025. When a company triples its São Paulo office space and opens new offices in Heredia, Costa Rica, and Mexico City in the same 24-month window, it is placing infrastructure bets, not testing the water.

Grupo Bimbo, one of the world's largest bakery companies, processes hundreds of thousands of daily transactions on the Databricks Data Intelligence Platform. Banco Bradesco, iFood, Natura &Co, PicPay, Cemex, and Coca-Cola FEMSA are all live. These are production workloads — governance, real-time analytics, operational AI — not pilot programs. "We are seeing unprecedented demand for data and AI from companies across the region," said Marcos Grilanda, Databricks' Vice President and General Manager for Latin America.

The physical footprint tells the story in concrete. The new São Paulo office sits on Avenida Brigadeiro in the Faria Lima Corporate Building, offering three times the collaborative space of the previous location. The second LATAM hub is slated to open in the first half of 2025, with the Data Intelligence Platform becoming available in the Azure Mexico data center around the same time. Databricks' Mexican enterprise customers currently run their workloads in U.S. cloud regions on AWS, Azure, and GCP; the local Azure data center changes the latency and data-sovereignty calculus entirely.

Job-board data confirms the hiring is tilting toward the new product line: 55 Databricks roles added in the past seven days include Lakebase Sales Specialist postings in both Mexico City and São Paulo, plus a Director, Lakebase Sales Specialists in São Paulo. The initial Mexico City push targets go-to-market hires, but the broader regional strategy clearly spans technical, customer-facing, and operations roles.

"We are looking forward to helping more Mexican enterprises unlock the power of data intelligence and contributing to the data and AI ecosystem by helping train and upskill local businesses." — Marcos Grilanda, VP and GM for Latin America, Databricks

That quote reads differently when you map it against the investment cadence. Databricks opened its first LATAM office in São Paulo two years ago. It followed with San José, then Heredia, then the Mexico City announcement, then the expanded São Paulo flagship. Each step locks in local talent before competitors arrive. The "upskilling" language is not corporate goodwill; it is pipeline construction for a region where enterprise-data engineering talent is abundant but platform-specific experience is scarce. Whoever trains that workforce first owns the account base.

How Lakebase Reshapes the Regional Hire

Databricks' product pitch has always been analytics at scale. Its Latin American job listings tell a different story now. The June 11 launch of Lakebase (a fully managed Postgres operational database built on Neon technology, which Databricks acquired in 2024) has started pulling the company's regional hiring away from pure sales and toward engineers who can ship and support transactional workloads.

In the past seven days alone, Databricks posted a Lakebase Sales Specialist role for Mexico City, another for São Paulo, and a Director, Lakebase Sales Specialists position also based in São Paulo. These aren't generalist AE roles. They require enough fluency in OLTP architecture, Postgres internals, and lakehouse integration to sell a product that didn't exist a month ago.

This mirrors the broader hiring pattern LinkedIn shows: a Sr. Director, Field Engineering (Lakebase) role for North America, tasked with scaling a global field engineering organization past 50 solutions architects, with explicit requirements for PostgreSQL, real-time streaming, and AI/ML state frameworks. The LATAM-specific listings (a Sr. Engagement Manager and a Field Technical Program Manager, both in Forward Deployed Engineering) map the same template onto the region.

Lakebase separates compute and storage using Neon's architecture, targets sub-10ms latency and over 10K QPS, and ships with branching, which are copy-on-write database clones that let AI agents experiment without risking production data. It auto-syncs with lakehouse tables and includes an online feature store for model serving. The design choices matter for hiring: these are infrastructure problems, not CRM problems. Databricks needs people who understand why an agent's state layer might need a transactional database instead of just another S3 bucket.

The partner ecosystem reinforces this. Accenture, Deloitte, EPAM, Fivetran, Confluent, and dbt Labs all signed on at launch. Each of those firms will need implementation talent in-region, and most already have São Paulo or Mexico City practices. The engineering gravity isn't coming only from San Francisco.

Databricks CFO Dave Conte told investors at the Data + AI Summit on June 11 that annualized revenue will hit $3.7 billion by July. The company is hiring 3,000 people this year. How much of that headcount lands in Latin America remains undisclosed, but the Lakebase listings make one thing clear: the region isn't just getting a quota-carrying sales team. It's getting asked to sell, deploy, and support a product that competes directly with Oracle and Microsoft SQL Server — in a market those incumbents have owned for two decades. That demands engineers, not just account executives.

Seattle Spillover

Databricks signed a 160,000-square-foot lease in a new downtown Bellevue skyscraper in June 2026, more than doubling its Puget Sound office holdings and pushing its total regional footprint past 270,000 square feet. The Four106 lease is the largest piece of a piecemeal expansion that has been running since at least 2024, with offices at Sixth & Lenora in Seattle, multiple floors at City Center Bellevue, and roughly 45,000 square feet added at West8 earlier in 2025. The Puget Sound Business Journal first reported the deal; Connect CRE confirmed the square footage.

The timing matters. Seattle's Eastside office vacancy hit 25.7 percent last quarter after Microsoft's exit from the Bravern towers, which left nearly 750,000 square feet empty. Databricks is expanding into that slack, and doing it while simultaneously standing up new offices in Mexico City and São Paulo. The two moves share a logic.

Both markets offer engineering talent at a discount to the Bay Area. Both sit in time zones that overlap with Latin American and European working hours. And both are anchored by existing hyperscaler presence (Microsoft and Amazon in Seattle; AWS, Google, and Azure across LATAM), which means the cloud infrastructure Databricks runs on is already built out. Hiring nearly 50 open positions across the Seattle-Bellevue area while listing Lakebase Sales Specialist roles in São Paulo and Mexico City on its own careers page is not a coincidence. It is one distributed engineering strategy running on two continents.

For engineers tracking where Databricks actually builds rather than where it sells, the signal is clear: the company is assembling a Pacific-time-zone engineering spine that stretches from Bellevue to São Paulo, with Mexico City in between. The next question is which teams sit where, and the job board suggests LATAM gets sales and solutions architecture while Seattle absorbs broader engineering work. Watch the split as Lakebase ramps.

Nokia's Proof of Concept and What It Signals

The Nokia-Databricks partnership moved from press release to proof of concept in June 2026, and the fine print of what they actually tested reveals more about Databricks' LATAM trajectory than the announcement itself. Nokia and Databricks built a unified, substrate-agnostic data platform purpose-built for autonomous telecom networks. It is the kind of system that ingests network telemetry at Tier-1 carrier scale and runs the same data pipelines across Databricks and open-source environments built on Apache Flink, Kafka, and Iceberg without rewriting code. That last detail matters: it means a Brazilian or Mexican telco with a decade of legacy OSS/BSS sprawl can adopt Databricks' platform without torching its existing stack.

Nokia engineers developed a platform-independent data transformation layer in Python that decouples business logic from infrastructure-specific connectors. A compiler then translates those workflows into native execution formats (Delta Live Tables on Databricks, Flink SQL on open-source clusters), automatically. The system also demonstrated AI-assisted generation of new data products through natural-language prompts and zero-copy data sharing across operational domains, with multi-agent AI systems performing root-cause analysis and cross-domain correlation. "This is a big step as we work toward building the types of data foundations required for next-generation autonomous networks," said Oguz Sunay, Nokia's CTO for AI and autonomous networks.

Why should LATAM watchers care? Because telco is infrastructure, and infrastructure buyers in Latin America sign contracts measured in years, not pilot quarters. A Nokia-Databricks deployment at a major LATAM carrier (América Móvil, Telefônica Brasil, or any of the region's top-five operators) would anchor Databricks' platform in production traffic that doesn't turn off. The engineering roles required to support that kind of deployment, including data platform engineers, Flink and Kafka specialists, and solutions architects who understand both telecom protocols and Lakebase operational databases, are precisely the roles now appearing on job boards across the region.

There is also a competitive angle worth tracking. Nokia has a parallel partnership with Google Cloud for telco AI, and AWS is pushing its own telco data frameworks. Databricks' counter-position is cloud-agnosticism: run the same pipelines on any infrastructure, including on-prem, which is non-negotiable in markets where data sovereignty rules keep workloads out of hyperscale regions. That pitch lands harder in São Paulo than in Virginia, precisely because Brazilian and Mexican regulators have kept local infrastructure requirements tight. Follow the Nokia reference architecture: when named telcos start citing it in procurement RFPs, the hiring wave follows six to twelve months later.

The Power-Bill Wager

Naveen Rao, the former head of AI at Databricks, has made a claim so large it functions as a stress test for the entire enterprise AI thesis. His new company, Unconventional AI, promises to cut the power cost of AI inference by a factor of 1,000x. If it works, the LATAM talent build stops being a straightforward hiring story and becomes a cost-arbitrage play where Mexico City and São Paulo matter precisely because they sit far from the most expensive electricity grids on earth.

Rao's pitch, detailed in a TechCrunch report from June 25, 2026, rests on an oscillator-based computing architecture that replaces conventional transistor switching with a different physical mechanism. The company's first public artifact is Un-0, an image-generation model that runs on a software simulation of those oscillator chips and produces output comparable to Stable Diffusion and OpenAI's GPT Image 1. Rao said the system would eventually run AI models at "1/1000 of power," with prompts entering over a network cable and inferences exiting at a fraction of today's energy cost. The company, which counts fewer than 50 employees, plans to release physical chip schematics and build a full inference stack from scratch.

The claim is staggering, and it should be treated as such. Hardware gains demonstrated in simulation routinely collapse once silicon meets memory bottlenecks, I/O limits, and thermal reality. Rao himself acknowledged the infrastructure is still being built. But the direction of travel matters more than the timeline. Databricks' own platform is already used by energy companies to cut waste and lower per-unit power costs. Rao's bet is that the inference side of the AI buildout, the part that never stops running once a model is deployed, will hit a wall so hard that only a fundamental architectural break can clear it.

That is what makes the LATAM offices more than a sales play. If inference power costs drop by even a fraction of the 1,000x target, the economics of running always-on enterprise AI shift dramatically. Regions with lower industrial power costs, growing engineering pools, and proximity to markets that lack their own mega-data-center infrastructure become deployment locations, not just support outposts. Both cities are already where Databricks is hiring Lakebase Sales Specialists and a Director of Lakebase Sales Specialists, roles tied to the operational-database product that sits at the intersection of data intelligence and real-time infrastructure. Those hires are the front edge of a bet that the next wave of AI deployment will be distributed, power-constrained, and built by people who understand local grids as well as local customers.

"AI scaling is hard because of energy. It's going to be the fundamental limit in the next few years," Rao said. If he is even directionally right, the engineers and solution architects Databricks is hiring across Latin America are not staffing a regional outpost. They are building the workforce for an industry that has just discovered its most expensive input.

What Engineers and Operators Should Watch

Zero G Talent's board shows Databricks added 55 roles in the past week alone, including Lakebase Sales Specialist posts in both cities and a Director of Lakebase Sales Specialists in São Paulo, a sign the LATAM build is shifting from founder-led sales to layered management. Watch for whether those individual contributor roles backfill quickly; repeat postings signal either territory expansion or attrition problems nobody talks about publicly.

The Solutions Architect posting that sat on LinkedIn and Built In through early 2025 tells you what the job actually looks like on the ground. Based in Heredia, Costa Rica, it required 50% travel to Mexico and other Spanish-speaking markets and covered roughly 20 enterprise accounts as the primary technical voice in pre-sales. The ask was production programming in Python, R, Scala, or Java plus hands-on experience with Hadoop, NoSQL, MPP, OLTP, and OLAP. That is not a sales engineer who demos; it is someone who designs and extends production data systems alongside the customer's team. If Databricks posts a second or third copy of that role from the same Heredia hub, the Costa Rica base is doing more than covering Central America.

Compensation data from Levels.fyi gives you a benchmark to evaluate any offer:

These are U.S.-anchored numbers (LATAM-based roles will almost certainly adjust downward for local cost of living), but the spread tells you where Databricks' internal leveling sits. If a Mexico City or São Paulo offer comes in dramatically below these bands, you are being asked to take a regional discount on a company that pays U.S. rates for remote U.S. roles. Know your walk-away number before negotiating.

The partner ecosystem is the signal most people miss. Compass UOL was named Databricks' 2025 LATAM Delivery Excellence Partner after five years of collaboration on global projects formalized in 2023. That means there is already a channel of consultants across the region who know the platform, which lowers the activation energy for enterprise buyers who want to pilot before committing. Watch the Databricks Partner Directory for new LATAM-based resellers and system integrators, as each one is a leading indicator that deal flow in the region is large enough to sustain a local services economy. If you are an independent operator or a small shop, the partner program is a faster path to Databricks-adjacent revenue than waiting for a direct hire.

The telco pipeline is the one to track for 2026. The Nokia autonomous-networks demonstration on Databricks' platform is not a lab experiment; it is a buyer with a real infrastructure budget validating the stack for production. Telco contracts are long, capital-intensive, and sticky. If one major LATAM carrier signs, expect procurement teams at the others to start evaluating within two quarters. That creates demand for engineers who understand both the data platform and the operational constraints of network infrastructure, a narrow profile that commands a premium.

One practical note: Databricks' São Paulo office expansion, announced July 1, 2025, came with a 150% two-year growth in Brazil headcount and a Data + AI World Tour stop in São Paulo on September 3, 2025. The World Tour events are recruiting theater disguised as user conferences. If one comes to Mexico City, attend, because the hallway conversations will tell you more about the team's actual roadmap than any job description.

Working in AI? Zero G Talent tracks the openings: browse AI jobs, openings at Databricks, and the people building the field.