Ora Computing Receives 3.5 Million Euros for More Efficient AI Models

The Austrian deep tech startup Ora Computing has closed a seed funding round totaling 3.5 million euros. The round is led by Constructor Capital and Greencode Ventures. In addition, existing investor XISTA Science Ventures is once again investing in the company, having already played a central role in the founding and development of Ora Computing.

With the fresh capital, the startup plans to further develop its technology for compressing AI models, tap into new areas of application, and launch its commercial platform for cloud providers and edge AI applications.

The Growing Cost Problem of AI

While the performance of modern AI systems is increasing rapidly, so are the costs of operating them. Inference, in particular, is increasingly becoming a significant cost factor. Large companies are now investing millions each month in computing power alone. At the same time, many applications in the automotive sector, in industrial plants, or on mobile devices are reaching technical limits, as modern language models are often too large and too resource-intensive for local hardware.

Reducing AI Models by Up to 80 Percent

Ora Computing’s technology makes it possible to significantly compress large AI models. According to the startup, models can be reduced in size by up to 80 percent while running up to four times faster. The loss in accuracy is said to be only between zero and five percent.

This not only reduces the need for computing power but also lowers energy consumption. According to the company’s calculations, even a one-percent market penetration could save more than 50,000 metric tons of CO₂ per year.

An Alternative to the AI Industry’s Arms Race

Ora Computing is deliberately positioning itself against the current trend toward ever-larger AI models. Instead of focusing on maximum scalability, the founders are betting on more efficient, specialized models optimized for specific use cases. The vision: In the future, AI should not run exclusively in massive data centers, but also directly in vehicles, machines, or other end devices.

Unlike many existing solutions, the compression works independently of the hardware used and can be integrated directly into existing inference frameworks. Companies need neither to overhaul their infrastructure nor to perform costly retraining of their models.

Research Meets Industrial Application

Ora Computing was founded by Stefan Sack and Raimel Medina. Both are alumni of Professor Maksym Serbyn’s renowned research group at the Institute of Science and Technology Austria and have backgrounds in quantum computing. After a period in stealth mode, the company made its first public appearance in late 2025. Since then, Ora Computing has already validated its technology with companies in the automotive and edge computing sectors.

The team was able to compress a language model with 70 billion parameters in just a few hours and for less than $1,000. Comparable projects in the industry often cost several hundred thousand dollars.

Investors See Efficiency as the Key to the Next Generation of AI

For investors, the company’s appeal lies primarily in the growing importance of efficient AI infrastructure. Rather than relying exclusively on ever-more-powerful hardware, they see intelligent model optimization as a crucial lever for the next phase of artificial intelligence development.

With the new capital, Ora Computing now plans to expand its team, support even larger AI models, and launch a commercial product for cloud providers and enterprises that want to run AI directly on end devices.