VCs Name 19 AI Efficiency Startups to Watch
European VCs identify startups working to reduce the cost and energy demands of AI models as compute infrastructure struggles to keep pace with demand.
Venture capitalists have pinpointed 19 European startups focused on making artificial intelligence more efficient. The move comes as the AI boom's compute demands outstrip the capacity to build new data center infrastructure.
Investors from firms like Plural and Notion Capital highlighted these companies in a survey. The goal is to reduce the massive cost and energy requirements of current AI systems.
Hardware & Chip Innovations
Several startups are designing new hardware to run AI models with less power. Salient Circuits is creating a "digital brain" based on neuromorphic principles. This approach could drastically cut energy use for on-device, always-on AI applications.
Another company, Axelera AI, is building a new class of energy-efficient AI chips. Its technology focuses on performing computations directly in memory. This reduces the data movement that consumes so much power in traditional systems.
Fabric8Labs uses a unique 3D printing technique for chip manufacturing. It aims to produce processors with superior thermal performance. Better heat management allows chips to run faster without overheating, improving overall efficiency.
Software & Model Optimization
Other startups are tackling efficiency through software. Numenta is developing algorithms inspired by the human brain's neocortex. Its method could make large language models far more efficient without sacrificing capability.
Luminous is building what it calls a "sparse expert" model. This architecture activates only the necessary parts of the network for a given task. The result is a significant reduction in the computational load per inference.
CentML focuses on compiler-level optimizations for AI workloads. Its software automatically optimizes how models run on existing hardware. This can yield immediate performance gains and cost savings without changing chips.
Data Center & Systems Efficiency
Improving data center operations is another key frontier. Zeta Alpha uses AI to optimize energy use in its own facilities. It applies predictive analytics to manage cooling and power distribution more effectively.
Nano Interactive has developed a liquid cooling system specifically for AI servers. This direct-to-chip technology allows for higher compute density. More processing can be packed into a smaller, cooler footprint.
Recurve takes a broader systems approach. It analyzes the entire AI stack, from chips to cooling to model design. The company then identifies the most impactful places to cut energy consumption and cost.
Specialized Applications & Tools
Some startups target efficiency for specific AI uses. Aikido Security focuses on making AI security tools less resource-intensive. Its goal is to enable continuous, real-time threat monitoring without prohibitive compute costs.
Another, Giskard, provides an open-source testing platform for AI models. It helps developers identify inefficiencies and biases before deployment. Catching issues early prevents wasteful runs of flawed models.
Finally, companies like Baseten and Qwak offer MLOps platforms designed for efficiency. They streamline the process of deploying and managing models at scale. This reduces the operational overhead and resource drift that often inflates costs. The full survey details are available on Sifted's website.
