A new AI model could play a key role in reversing our power-hungry needs, says Román Orús, Chief Scientific Officer, Multiverse Computing.
Artificial intelligence has become part of everyday business life. Manufacturers use it to monitor production lines, banks use it to run scenarios, and hospitals use it to support clinical decisions. But as AI grows in capability, it also grows in size. The most advanced AI systems, especially large language models (LLMs), require huge amounts of computer power. That means high energy use, expensive hardware, and rising environmental impact.
AI compression is rapidly emerging as one of the most effective ways to solve this problem. It makes AI models dramatically smaller, faster and less energy-hungry, while still keeping the accuracy that organisations need. For businesses looking to combine digital transformation with ambitious sustainability goals, it represents a practical and immediate opportunity.
Why traditional AI models are becoming unsustainable
Most organisations use AI through cloud services or large servers that consume significant amounts of electricity. As models grow, so does the power required to run them. This becomes a major problem when many businesses are also facing energy-efficiency targets, regulatory pressure, and rising energy costs.
The challenge is not only about environmental impact. Bigger models also need highly specialised computer chips that are in short supply and expensive. For smaller organisations, or any business trying to scale AI across operations, this creates a barrier to adoption.
AI compression changes this equation by reshaping the model so that it demands far less computing power. This allows companies to run AI sustainably and affordably, without sacrificing performance.
What is AI compression?
AI compression is a way of shrinking a large model into a smaller one without losing what makes it useful. Think of a detailed image being compressed into a much smaller file size while keeping the picture clear. The content stays the same, but the storage space drops dramatically.
Quantum-inspired compression techniques take this a step further. Instead of simply removing pieces of the model, they reorganise it. They break it down into smaller, more efficient building blocks and remove unnecessary repetition. This ensures the model still understands patterns and relationships in the data but no longer carries the “weight” of its original size.
The result is an AI system that can be up to 95% smaller, uses far fewer resources, and still performs almost the same tasks with only marginal differences in accuracy.
Why smaller AI models use far less energy
AI uses energy mainly when it processes data. Larger models require far more steps to produce an answer, and those steps need powerful processors. When the number of steps is reduced, the energy required drops with it.
Compressed models can reduce energy consumption by around 50% because:
They have fewer operations to perform per query.
They can run on standard hardware rather than high-end chips.
They eliminate the need for constant cloud access, reducing the energy associated with data transfer.
This means businesses can integrate AI without expanding their power usage or relying on additional infrastructure. It also supports national and regional sustainability policies by helping organisations meet energy-efficiency targets.
Running AI locally, without the cloud
One of the biggest benefits of compression is the ability to run AI models on local servers or even on smaller devices. Because they are lighter, they no longer need constant access to the cloud. This brings several advantages.
For one, it significantly increases security. Sensitive information stays within the organisation and doesn’t have to be sent externally for processing. It also improves reliability. Factories, vehicles, and hospitals can operate AI-powered tools even in locations with limited or no connectivity.
Proven benefits across industries
Compressed AI is already demonstrating meaningful improvements in several industries.
Manufacturing:
A European factory used AI compression to shrink its quality-control model. The smaller model ran twice as fast and used about 50% less energy. It integrated more easily into local systems and supported real-time decision-making on-site.
Healthcare:
Hospitals can now run advanced AI tools on local computers or private servers. Patient data stays securely inside the hospital, while doctors benefit from faster diagnostics and analysis.
Defence and remote operations:
Drones and field devices can run AI locally without internet. This provides reliable intelligence in remote or sensitive environments with limited connectivity.
These examples reflect a broader shift: AI is moving closer to the action rather than distant data centres.
A more sustainable future for business AI
For companies balancing innovation with sustainability, compressed AI cuts energy use and environmental impact. It enables digital transformation without traditional models’ heavy infrastructure. It widens AI access, as smaller models run on existing hardware.
By focusing on efficiency rather than size, compressed AI offers a new direction for the future of intelligent technology, one where powerful systems can also be sustainable, secure and accessible.
