Context
Brazil is a hard place to buy a serious workstation. Import friction, thin support, and machines specced on paper instead of for real workloads. Professionals were paying premium prices for systems that still bottlenecked under their actual jobs.
Problem
Performance was being sold as a spec sheet. But a render farm, a CAD seat, a simulation rig, and an AI workstation each fail in completely different ways. The market optimized for the number on the box, not the experience under load.
What I did
- Founded and scaled a B2B high-performance computing company from a single workbench to a national operation.
- Helped more than 200,000 people figure out the right machine for their actual workload, which became a dataset of real pain points we study to this day.
- Built the sales, service, logistics, and support systems that let thousands of customers trust the brand with mission-critical machines.
- Engineered systems around workloads, thermals, drivers, storage topology, and stability, rather than headline specs.
What I learned
- Performance is a promise, not a benchmark. Customers buy the absence of problems.
- The bottleneck is almost never the part you would put on the box.
- Trust compounds. Service and proof are part of the product, not overhead.
- Hardware taught me the limits of hardware, which is why I am now building software.
Razor started with a simple frustration. The people doing the most demanding work in the country had the worst options for the tools they depended on. A structural engineer running simulations overnight, a studio rendering on a deadline, a lab pushing a workload no consumer machine was designed for. They were all buying on faith.
What we actually built
We did not sell components. We sold a guarantee that the machine would get out of the way. That meant matching the system to the workload, validating it under the conditions it would actually face, and standing behind it with service that treated downtime as the real enemy.
Helping people choose also gave us something rare. Over the years we guided more than 200,000 people toward the right machine for their work. Most of them never became customers, but every conversation taught us where computers actually let people down. That body of real-world pain points is the foundation for what I am building next.
Why it mattered
Across more than a decade, that approach turned into thousands of systems in the field and thousands of customers who came back. But the most valuable output was not revenue. It was a deep, hands-on understanding of why computers disappoint the people who depend on them.