South Africa's Refiant Raises $5M to Build Efficient AI Models

BREAKING NEWSApril 17, 2026 at 12:02 PM UTC

TLDR

South African startup Refiant AI secures $5 million funding round to develop tools for reducing AI model size and energy consumption.
Founded in 2025, Refiant AI compresses AI models to run on smaller devices, aiming to address energy consumption challenges in AI industry.
Refiant AI's technology focuses on reducing resource use for AI models, offering an alternative approach to traditional high-performance computing infrastructure.

South African startup Refiant AI has raised $5 million in a seed round to develop tools that reduce the size and energy use of artificial intelligence models. The funding will support product development, hiring and enterprise partnerships.

The company was founded in 2025 by Viroshan Naicker, Siddharth Gutta and Mathew Haswell. It builds software that compresses AI models by reducing computational weight and retraining them to maintain performance, allowing them to run on smaller or local machines.

The round was led by VoLo Earth Ventures. The investors said the approach could reduce the need for large-scale computing infrastructure as companies expand AI use.

Global demand for AI is driving investment in data centres with high-performance chips and cooling systems. This infrastructure requires large amounts of energy, increasing costs and pressure on power supply.

Refiant AI said its technology focuses on reducing resource use rather than expanding capacity. “AI’s growing energy footprint is one of the most urgent challenges,” Gutta said, adding that improving efficiency could change how models are deployed.

Key Takeaways

The funding reflects a shift in how the AI industry is addressing scale. Most large technology firms have focused on increasing computing power by building data centres and deploying more graphics processing units. This model depends on access to capital, energy and infrastructure, which are limited in many regions. Startups like Refiant AI are working on an alternative approach by reducing the size and energy needs of AI models. This could allow companies to run AI systems on local devices or smaller servers, lowering costs and improving access. For emerging markets, including Africa, this model is significant. It reduces reliance on foreign infrastructure and makes it easier to deploy AI in sectors such as finance, healthcare and logistics. It also aligns with global pressure to reduce energy consumption linked to data processing. Investors are starting to back companies that improve efficiency rather than scale alone. The main challenge will be maintaining model performance while reducing computational requirements. If successful, this approach could change how AI systems are built, distributed and monetised.

Venture Capital