€3.5M Seed Backs Ora Computing’s AI Efficiency Layer

Ora Computing has secured €3.5 million in seed funding led by Constructor Capital and Greencode Ventures to address the increasing costs and energy demands of artificial intelligence. The startup’s software achieves up to 80% reduction in AI model size, enabling deployment on hardware previously unable to support them and potentially unlocking significant environmental benefits. According to Ora Computing, with just 1% market penetration, its technology could eliminate more than 50,000 tonnes of CO2 annually. CEO & co-founder Stefan Sack believes the next wave of AI will be driven by compact, efficient models optimized for specific applications.

80% Model Compression Achieves Efficiency for AI Inference

Ora Computing’s software reduces AI model size by up to 80%, enabling the deployment of artificial intelligence models on hardware previously considered insufficient. This has significant implications for localized AI processing and reduced computational demands. The startup has developed a method for compressing foundation models that fundamentally alters how efficiently they operate, rather than simply reducing file sizes; this allows companies to expand AI implementation without necessarily increasing infrastructure costs. This compression maintains a narrow accuracy loss, ranging from 0 to 5%, addressing concerns that optimization frequently compromises performance. This reduction stems from the decreased compute power required to run the compressed models, lessening the energy footprint of AI inference, a process that can currently reach tens of millions of euros per month for large-scale deployments.

The company demonstrated this capability by compressing a 70 billion parameter model in a matter of hours at a compute cost of under $ , a figure significantly lower than the hundreds of thousands of dollars typically associated with comparable compression tasks. Terhi Vapola, Founder and Managing Partner of Greencode Ventures, emphasizes this point, stating, “AI’s energy appetite is growing faster than the world can build the infrastructure to feed it. One key approach is to make AI itself more efficient, and that is exactly what Ora does.”

Ora Computing’s Algorithm Optimizes Tradeoffs Without Retraining

Ora Computing is addressing a critical bottleneck in the rapidly expanding field of artificial intelligence: the increasing computational cost and energy demands of running large models. Current AI deployments, particularly those utilizing massive foundation models, can incur tens of millions of euros monthly in compute expenses, a figure that only increases as model sizes grow.

This compression isn’t achieved through traditional methods requiring extensive and expensive retraining; instead, Ora’s algorithm dynamically maps the tradeoff between model size and accuracy, allowing for optimization tailored to specific hardware and budgetary constraints. Founded by Stefan Sack and Raimel Medina, both researchers from the Institute of Science and Technology Austria, Ora Computing emerged from stealth late last year and has already validated its solution with companies in the automotive and edge silicon sectors. “We founded Ora Computing to challenge the assumption that massive scale is needed to reach useful intelligence.” The recently secured €3.5 million seed funding will fuel team expansion, broaden compression capabilities to encompass the largest frontier models, and facilitate the launch of a commercial product aimed at cloud inference providers and edge AI deployments.