Opinion: Gaurav Sharma, CEO of io.net
Artificial intelligence may be in its infancy, but it has already delivered significant scientific and technological breakthroughs in developed countries. Unfortunately, this development has come at a cost: the hazardous centralization of artificial intelligence.
ON Forbes list for 2025 of the 50 largest private AI companies, all are based in developed countries and 80% are based in the US.
Artificial intelligence continues to gravitate toward well-capitalized tech giants in developed countries.
For many emerging economies, the price of entry into the AI revolution is unaffordable. We need to ensure that AI innovation and development is accessible to as wide a range of projects as possible.
Imbalance in access to artificial intelligence
The crux of the problem is access to computation. Training and deploying huge AI models requires enormous GPU power. Supply has not kept pace, pushing the price of Nvidia H100 chips to over $30,000.
An ambitious artificial intelligence research company can spend 80% or more of their computing resources – resources that could otherwise be devoted to R&D or talent. Well-funded tech giants can raise billions to secure them. The rest of the world can’t.
The consequences are far-reaching. AI-based innovations may become a monopoly technology controlled by a handful of corporations and nations. Promising applications of artificial intelligence in agriculture, education and health care in developing economies may never be realized – not because of a lack of talent, but because of narrow access to computing.
From a geopolitical perspective, computing power shortages are beginning to mirror the situation in the oil and silicon markets. Nations without sovereign access to computing will be forced to import it, creating dependence on countries that may not be aligned with their domestic targets and exposing importers to foreign energy and real estate markets. These dependencies threaten economic competitiveness and national security.
The dangers of centralizing the influence of artificial intelligence
If access to computing remains concentrated in developed countries, this will also have an impact.
Frontier AI technology, from LLM to diffusion models, will be shaped by the same perspectives, narrowing diversity and introducing systemic risk. Developing countries risk being cut off from contributing to or benefiting from the technology that defines the global economy.
Centralization ensures that profits flow disproportionately to those with privileged access, leaving smaller players, often those who create locally relevant tools, behind. Over time, barriers to competing in the AI market could morph into a destabilizing oligopoly, preventing the developing world from making key industrial change. Concentrated control of infrastructure has always been disruptive, and artificial intelligence will be no exception.
Balancing weights using decentralized computation
The solution to the challenges of accessibility and centralization is surprisingly basic: computational markets powered by blockchain. Just like Uber unlocked unused cars and Airbnb unlocked spare rooms, decentralized computing marketplaces unlock unused hardware. The result is lower prices and a more diverse and resilient ecosystem of suppliers and consumers.
Related: Bitcoin network used to secure election results at local GOP convention
Globally, millions of GPUs remain idle in data centers, enterprises, universities, and homes. By clustering these GPUs on-demand via blockchain, underutilized hardware is made available at a fraction of centralized compute costs. Startups in lower-income countries can afford to scale AI workloads that are no longer precluded by the capital advantage of industry leaders.
The key role of Blockchain
Without blockchain, this model would not be possible. Tokens provide a layer of coordination and trust that aligns incentives across decentralized physical infrastructure networks (DePINs). Leading DePINs require compute service providers to stake tokens to augment reliability, which comes with downtime penalties. Developers pay with tokens, enabling seamless cross-border settlements.
For hardware providers, tokenized rewards provide a more equitable economics: they reward capacity owners based on usage, providing previously unavailable revenues without sacrificing their core purpose. For developers, access to cheaper computing encourages participation and innovation in AI. This creates a positive feedback loop – as more players enter the decentralized computing market, computing resources become cheaper and more common.
Meeting the challenges
Some critics argue that decentralized computing is not as effective as hyperscalers, citing latency and quality issues. The reality is completely different. DePINs deliver competitive performance in terms of latency, concurrency, and throughput. Techniques such as smart workload routing, mesh networking, and tokenized high availability incentives lend a hand maintain performance and dynamically optimize it based on workload needs.
Additionally, some DePINs have built see-through network explorers that allow developers and investors to verify performance claims in real time. These mechanisms lend a hand make DePIN even more reliable and cost-effective than established providers.
DePINs are also more diverse than hyperscalers’ offerings. Over 13 million devices are now available online, so developers can take advantage of a wide range of hardware and find the right tool for their AI projects, from high-performance cloud-class GPUs to specialized edge devices.
A level playing field for artificial intelligence
We have a narrow window to define the technology landscape for future generations. Many US and Chinese corporations could take the lead, but decentralized computing markets offer a promising alternative. By lowering costs and expanding access, startups, scalables, researchers and enterprises around the world can compete on a more level playing field. Emerging economies can build models for their own languages, healthcare systems, cultural beliefs and financial needs.
The question is not whether decentralization is necessary, but how to convince developers around the world to take advantage of this opportunity while increasing the number of companies that put their excess computing power into DePINs. Only by decentralizing computing can artificial intelligence become truly accessible and serve as many people as possible, not just entrenched oligopolies.
Opinion: Gaurav Sharma, CEO of io.net.
This article is for general information purposes and is not and should not be treated as legal or investment advice. The views, thoughts and opinions expressed here are solely those of the author and do not necessarily reflect or represent the views and opinions of Cointelegraph.