With the explosion of artificial intelligence and growing concerns about energy consumption and associated costs, it’s natural to wonder how advances in AI can remain financially viable. This concern is particularly pressing for small and medium-sized businesses wondering whether Large Language Models (LLMs) will remain accessible or become exclusive to larger businesses.
Only two years ago, the industry faced significant challenges related to processing power and chip availability. NVIDIA, primarily known for its GPUs for the gaming industry, has found itself in the spotlight as these GPUs have become essential for scaling large language models. Experts predicted that the limitations of Moore’s Law and affordability issues would slow progress, suggesting there was no way to meet demand or build enough affordable chips to power the future of AI.
Just 24 months later, NVIDIA’s market value has grown from around $300 billion to $3.652 billion in November 2024, nearly double the combined valuation of the 40 DAX companies listed on the German stock exchange. This unprecedented growth is fueled by substantial investments in AI, next-generation chip development, and collaborations with quantum technology companies to incubate new business models. What seemed impossible just two years ago is now a reality, a testament to the rapid and exponential pace of technological advancement.
This transformation highlights a key point: technological and scientific advances often exceed our expectations, leading to solutions that make cutting-edge technologies more accessible and affordable. Just as NVIDIA overcame previous limitations, the AI industry is poised to develop models and systems that will be economically viable for businesses of all sizes.
The open question
In the short term, however, as the premium prices of advanced AI models rise, there is growing momentum toward open source alternatives. Open source models can offer cost-effective solutions, allowing businesses to implement and customize AI without incurring significant expenses. By disrupting the revenue models of large AI companies, open source approaches could foster a more robust and versatile AI ecosystem. Community-driven improvements could lead to models that are cost-effective, efficient and adaptable to various sectors. However, the risks of open source AI are clear: once powerful AI tools are freely available, they can be exploited in unintended ways. As Meta’s Llama open source model release highlights, open source AI is already influencing global developments, even in sensitive areas such as military technology. Recent articles even claim that China used Llama to bring AI to the battlefield, calling into question the very idea of whether AI might be too powerful to be free.
Brain-inspired AI models: efficiency and innovation
Another promising path is progress. Taking inspiration from the efficiency of the human brain, AI researchers are developing models with specialized regions for different tasks: architectures that enable fast, energy-efficient responses in certain areas, while other areas are dedicated to complex problems, a bit like how the brain works.
Our brain is remarkably energy efficient and performs complex calculations with minimal energy consumption. Technological replication, on the other hand, requires considerable resources. Yet scientific advances and a growing understanding of biology are driving AI models toward higher performance without requiring as much computing power. Over the past 80 years, progress has been exponential. While the Gartner Hype Cycle teaches valuable lessons about disappointment, it also shows how breakthroughs can lead to “quantum leaps” in development. These advances not only reduce costs; they open up new possibilities in physics and science as we replicate biological efficiency in technology.
Leading companies like Microsoft with its Co-Pilot and the continued progress of OpenAI indicate that by 2025, what I call “Industry AI” will be ready for deployment. These are specialized AI agents with domain-specific knowledge – more streamlined than fundamental models but capable of outperforming them in their respective domains. This is the most fundamental shift in the history of human work as automation and AI agents deliver 24/7 capabilities, challenging traditional workforce structures . Small businesses will soon have access to AI solutions tailored to their needs without a high price tag, thanks to these specialized models designed to reduce computational loads and associated costs.
The role of energy innovations in reducing AI costs
Energy consumption is an important factor in the operational costs of AI models. But advances in energy technologies are expected to change this dynamic. Advances in fusion energy, as well as a large-scale transition to renewable sources like solar and wind, mean that future energy costs may not be the biggest risk. Instead, the focus could be on competitiveness, with China currently at the forefront of battery innovations and transitions to renewable energy.
Additionally, advances in materials science are opening up new opportunities in energy storage and distribution. Improved storage solutions and efficient distribution networks enable energy to be used more efficiently, reducing waste and operating costs.
As energy becomes more abundant and affordable, the cost of running high-performance AI models is expected to decrease, making advanced AI solutions accessible to a wider range of businesses, regardless of size.
A new era of AI accessibility
While the increasing costs of high-performance models present challenges, emerging trends in specialization, open source alternatives, and innovative business models offer viable solutions. Brain-inspired architectures make AI models more efficient and less resource-intensive. This convergence of innovation in AI and energy efficiency portends a future in which AI is not just the domain of large companies but a tool accessible to all, driving growth and innovation across all the sectors.
So perhaps the biggest concern today should not be whether small and medium-sized businesses can afford AI, but rather whether – in a market increasingly dominated by tech giants – there will remain small and medium-sized businesses. worry.
Anders Indset is a 5x Spiegel bestselling author and deep tech investor. This Norwegian native has twenty years of experience working with multinational companies and is the owner of the Njordis Group, the Global Institute of Leadership & Technology (GILT), and initiator of the Quantum Economy Alliance. His latest book is The Viking Code – The Art and Science of Norwegian Success. For more information, please visit, https://www.andersindset.com.