Opinion of: Gaurav Sharma, CEO of io.net
Artificial intelligence may be in its early days, but it has already delivered significant scientific and technological breakthroughs around the world. Unfortunately, these developments come at a cost: the dangerous centralization of AI.
In Forbes’ 2025 list Of the top 50 private AI companies, all are based in the developed world, with 80% in the US.
AI remains skewed toward well-capitalized tech giants in the developed world.
For many of the emerging economies, the price of entry into the AI revolution is unaffordable. We need to ensure that AI innovation and development is accessible to the widest range of projects.
The imbalance in AI Access
At the heart of the problem lies access to compute. The training and elimination of large AI models requires extensive GPU power. The supply is not keeping pace, driving the price for NVIDIA’s H100 chips up to more than $30,000.
An ambitious AI research company can spend 80% or more of its computing funding – resources that might otherwise go to R&D or talent. Well-funded tech giants can raise billions to secure them. The rest of the world is not.
The consequences are far reaching. AI-driven risks are becoming a technology monopoly, controlled by a number of corporations and countries. The promising applications of AI in agriculture, education or health care in developing economies may never materialize – not because of a lack of talent, but because of limited access to compute.
Geopolitically, the undersupply of compute begins with oil or silicon glass. Countries without sovereign access to compute will be forced to import it, creating dependencies on countries that may not be aligned with their national goals and exposing importers to foreign energy and real estate markets. These dependencies threaten economic competitiveness and national security.
The dangers of centralized AI influence
If access to computing remains concentrated in developed countries, so will influence.
Frontier AI technology, from LLMs to diffusion models, will be developed by both perspectives, Diversification and embeddedness of systematic risks. Developing countries risk being locked out of contributing to or benefiting from the technology that defines the global economy.
Centralization ensures that returns do not go back to privileged access, leaving out smaller players, often those who build locally relevant tools. Over time, barriers to competition in the AI market could turn into a monstrous oligopoly, freezing the developing world outside of a major industrial shift. Concentrated control control always made distortions, and AI is no different.
Balancing the scales with decentralized compute
The solution to the challenges of access and centralization is surprisingly simple: compute marketplaces powered by blockchain. Like Uber’s locked idle cars and Airbnb’s locked spare rooms, decentralized compute marketplaces unlock underutilized hardware. The result is lower prices and a more diverse and resilient ecosystem of suppliers and buyers.
Related: The Bitcoin Network was used to secure local GOP Convention election results
Worldwide, millions of GPUs sit in data centers, businesses, universities and homes. By pooling these GPUs into on-demand clusters via a blockchain, underutilized hardware can be used at a fraction of centralized compute costs. Startups in lower-income countries are able to scale AI workloads, no longer shut out by the advantage of industry leaders.
Blockchain’s critical role
Without blockchain, this model would not be possible. Tokens are the coordination and trust layer, which aligns incentives across decentralized physical infrastructure networks (DEPINS). Top depins require compute suppliers to stake tokens to incentivize reliability, with penalties for downtime. Developers pay in tokens, enabling seamless settlement across borders.
For hardware providers, tokenized rewards create a fair economy: paying compute owners based on usage, providing previously unavailable revenue without sacrificing their core purpose. For developers, access to cheaper compute means participation and innovation in AI. This creates a virtuous feedback loop – as more participants join the market for decentralized compute, compute becomes more affordable and abundant.
Responding to challenges
Some critics have argued that decentralized compute is not as performant as hyperscalers, citing latency and quality concerns. The reality is different. Depins deliver competitive performance across latency, concurrency and throughput. Techniques such as Smart Workload Routing, Mesh Networking and Tokenized Incentives for High Availability help maintain performance and dynamically optimize it based on workload needs.
In addition, some depins have built transparent network explorers, enabling developers and investors to verify performance claims in real time. These mechanisms help make depins even more reliable and effective than traditional providers.
The depins are more diverse than the hyperscaler offerings. Over 13 million Devices are now online, allowing developers to tap a wide spectrum of hardware and find the right tool for their AI projects, from high-performance cloud-grade GPUs to specialized edge devices.
A field playing field for AI
We have a narrow window to define the technological landscape for future generations. Many US and Chinese corporations may lead the way, but decentralized computing markets offer a promising alternative. By lowering costs and expanding access, startups, scaleups, researchers and businesses around the world can compete on a level playing field. Emerging economies can develop models for their own languages, health care systems, cultural beliefs and financial needs.
The question is not whether decentralization is necessary, but how to get developers around the world on board with this opportunity, while simultaneously increasing the number of companies that list their excess compute on depins. Only through decentralized computing can it truly be accessible and serve as many people as possible, instead of just hidden oligopolies.
Opinion of: Gaurav Sharma, CEO of io.net.
This article is for general informational purposes and is not intended to be and should not be construed as legal or investment advice. The views, thoughts, and opinions expressed herein are those of the author alone and do not necessarily reflect or represent the views and opinions of Cointelegraph.
