Reka gets $110M from NVIDIA and Snowflake

For Indonesia, the story is bigger than one founder. It is evidence that the country’s best technical talent can graduate from being consumers of imported AI to builders of exportable AI infrastructure. Dani, an ITB graduate and former Google DeepMind researcher, has turned that ambition into Reka, a company making multimodal models that can process text, image, audio and video in one system. Its products include Reka Flash, a 21-billion-parameter model, as well as tools for visual search, web-based research and model efficiency. That emphasis on efficiency is critical: as enterprise AI shifts from experimentation to deployment, the winners will be the models that can deliver capability without crushing inference costs.

This is where the market underestimates the opportunity. The AI trade has largely centered on the largest U.S. platforms and the semiconductor names supplying them, but the real economic value chain is expanding. Every enterprise that wants its own AI stack needs foundation models, tuning tools, retrieval layers and lower-cost inference. That creates room for a new class of beneficiaries beyond the obvious hyperscalers. If Reka can translate research credibility into enterprise adoption, it could become a regional anchor in the same way infrastructure software platforms became essential picks-and-shovels in earlier cloud cycles.

The funding from NVIDIA and Snowflake is especially telling. NVIDIA is not just a chip supplier; it is a strategic validator of where AI demand is headed. Snowflake’s participation signals that data platforms still see a future in specialized model partnerships, not just commoditized model access. For investors, that matters because it points to a broader ecosystem trade: the AI buildout is no longer just about chips, but about every layer that helps companies train, serve and operationalize models more efficiently.

Dani’s Endeavor selection also matters for execution, not just prestige. Global founder networks can accelerate hiring, open enterprise doors and shorten the path to strategic partnerships. For a startup trying to move from a promising lab into a durable global business, those relationships can be as valuable as capital. Endeavor itself says its entrepreneurs have generated more than $111 billion in annual revenue and millions of jobs, which underlines why access to that network can change the trajectory of a company like Reka.

Investors should read this as another sign that the AI opportunity is widening geographically and structurally. The market still prices most AI upside through a handful of U.S. mega-caps, but the next wave of returns may also come from the infrastructure enablers, semiconductor suppliers and cloud platforms that support a global ecosystem of model builders. NVIDIA remains the clearest public-market beneficiary, even after its powerful run, because frontier AI outside the U.S. still relies on its stack. Microsoft remains another obvious winner as enterprise AI spending migrates deeper into production workflows, while data and compute ecosystems stand to benefit from every new frontier model that comes to market.

The bigger takeaway is that Reka is a signal, not an isolated case. A founder from ITB, trained at DeepMind and backed by global capital, is building an AI company from Indonesia for the world. That is exactly the sort of asymmetric story investors should watch early: the market is still focused on the loudest AI names, while the next generation of deep-tech winners is quietly being formed in places that are easy to overlook. For long-term capital, the actionable thesis is simple: stay overweight the AI infrastructure stack and the enablers of global model development, because the demand curve is still broadening.