The alternative is renting the decision from whoever sold the compute. For any large enterprise, that conversation is not theoretical. It is happening right now, at every company trying to move from pilot to production.
What is missing is not the model. It is the control plane around the model, the system that decides where AI runs, proves every decision, and meters every cost. That control plane does not exist yet. It is what we are building.
Jensen Huang is right: every company will run AI in production. What he leaves unanswered is who controls it, under whose law, on whose infrastructure, by whose rules. Saptiva AI exists so that the answer to each of those questions is: yours.
The founders of Saptiva AI spent thirteen years (2007 to 2020) building and operating Quiubas Mobile, bootstrapped with no venture capital and no institutional backing, until it became the messaging backbone across Latin America and was acquired by Twilio in 2020. Saptiva AI is what happens when the same team, with the same operating philosophy, starts again on a bigger problem.
Angel and Jesus Cisneros founded Quiubas Mobile and ran it for thirteen years. No VC. No institutional money. Just the patient work of operating carrier-grade infrastructure for banks, telecoms, and enterprises across 61 countries. By the time of the Twilio acquisition in 2020, Quiubas was the dominant messaging backbone across the region.
What this taught us matters more than the outcome. We learned how to operate mission-critical infrastructure at scale, under compliance regimes that do not forgive mistakes, for customers whose trust is earned in quarters and lost in minutes. We learned the real shape of doing business with enterprises in Latin America.
After the Twilio acquisition, the obvious path was to step back. We saw something else: the same enterprises we had spent over a decade serving began to ask a new question. How do we actually deploy AI, given what we handle and who watches us? The answer, in every conversation, was the same: nobody had built the control plane that made it possible.
Hyperscalers were optimized for US and EU compliance. Systems integrators were running project-based pilots that never compounded. Internal teams were stuck in multi-year builds. Every enterprise AI initiative was a one-off.
We started Saptiva AI to build what did not exist yet. Application layer that deploys in two weeks. frIdA, the control plane that decides where AI runs and proves every decision. Infrastructure layer that operates across cloud, hybrid, on-premise, and air-gapped modes. One platform. One audit trail. Deployment mode as a runtime concern, not a replatforming event.
Today, Saptiva AI is in production at Rappi, Banco Invex, MultiMoney, Universidad Iberoamericana and ACMES. KAL, the first open-weight Mexican LLM, was built on it. This is the beginning, not the end state.
Meet the team →We do not cite Quiubas to brag. We cite it because when an enterprise in this region is considering handing its AI infrastructure to us, the relevant question is not what we have promised. It is what we have actually operated.
Market share
in Mexico
Countries
of operation
Acquired
by Twilio
We built Latin America's messaging backbone before APIs were standardized. We are building the control plane for AI, from Mexico, for the world.
Every product decision, every hiring decision, every customer conversation eventually lands on a principle. These are the ones we return to.
Pilots are how enterprise AI goes to die. We ship to production in weeks and stay until it runs.
Where it runs, what it decided and what it cost are not features we added. They are the reason the platform exists. Everything else is downstream of getting this right.
The customer chooses the model, the cloud, the deployment mode. The platform follows. If we can only win by trapping people inside our stack, we have not earned the win.
We are not a GPU reseller. We are not a model company. We are not a consultancy. We publish what we run, and we are honest about what we do not yet do. Confidence does not require overclaiming.
We write about our customers only as carefully as they would write about themselves. Volumes, metrics and names are theirs to disclose. The willingness to stay quiet is part of why named customers let us name them.
Quiubas was thirteen years. Saptiva AI is a longer arc. We are building infrastructure, not a growth chart. The decisions we make about customers, technology, and team reflect that.
If you're an enterprise evaluating AI infrastructure, a technologist considering the team, or an institution thinking about what owning your AI means in practice, there's a path in for each.