Opinion by: Gaurav Sharma, CEO of io.net
Artificial intelligence may be in its early stages, but it has already produced significant scientific and technological advancements across the developed world. However, these breakthroughs have resulted in the concerning centralization of AI.
According to Forbes’ 2025 list of the top 50 private AI companies, all are located in the developed world, with 80% situated in the US.
AI continues to be skewed towards well-funded tech giants in developed nations.
For many in emerging economies, the barriers to entry in the AI revolution are insurmountable. We must ensure that innovation and AI development are accessible to a wide array of projects.
The imbalance in AI access
At the core of the issue is access to compute resources. Training and deploying large AI models necessitates massive GPU power. Supply has not kept pace, resulting in Nvidia’s H100 chips reaching prices of over $30,000.
An ambitious AI research firm might allocate 80% or more of its funding on compute—resources that could otherwise be directed toward R&D or talent. Well-established tech giants can raise billions to secure them, while the rest of the world struggles.
The ramifications are extensive. AI-driven innovation risks becoming a monopolistic technology, dominated by a select few corporations and nations. Promising AI applications in agriculture, education, or healthcare within developing economies may never come to fruition—not due to a lack of talent, but because of restricted access to compute.
Geopolitically, the insufficient supply of compute resembles oil or silicon. Nations lacking sovereign access to compute will inevitably have to import it, creating dependencies on countries that may not share their national goals and exposing them to fluctuations in foreign energy and real estate markets. These dependencies pose threats to economic competitiveness and national security.
The dangers of centralizing AI influence
If compute access remains concentrated in developed countries, so too will influence.
Next-generation AI technologies, from LLMs to diffusion models, will be shaped by similar perspectives, narrowing diversity and embedding systemic risks. Developing nations risk being excluded from contributing to or benefitting from the technology steering the global economy.
Centralization ensures that disproportionate returns flow to those with privileged access, marginalizing smaller players, often those creating locally relevant tools. Over time, barriers to entry in the AI market could evolve into a destabilizing oligopoly, sidelining the developing world from a crucial industrial transformation. Concentrated control over infrastructure has always led to distortions, and AI will not be an exception.
Balancing the scales with decentralized compute
The remedy for the challenges of accessibility and centralization is surprisingly straightforward: compute marketplaces powered by blockchain. Just as Uber unlocked underutilized vehicles and Airbnb utilized spare rooms, decentralized compute marketplaces harness idle hardware. The outcome is reduced prices and a more diverse and robust ecosystem of suppliers and consumers.
Related: Bitcoin network used to secure local GOP convention election results
Worldwide, millions of GPUs lie idle in data centers, enterprises, universities, and residences. By pooling these GPUs into on-demand clusters through blockchain, previously underutilized hardware becomes available at a fraction of centralized compute costs. Startups in lower-income countries can scale AI workloads, no longer obstructed by the capital advantages of industry leaders.
Blockchain’s critical role
This model would not be feasible without blockchain. Tokens serve as the coordination and trust layer, aligning incentives across decentralized physical infrastructure networks (DePINs). Leading DePINs require compute suppliers to stake tokens to promote reliability, with penalties for downtime. Developers pay in tokens, facilitating seamless settlement across borders.
For hardware providers, tokenized rewards yield fairer economics: compensating compute owners based on usage, generating previously unavailable revenue without compromising their core function. For developers, accessing cheaper compute encourages participation and innovation in AI. This creates a virtuous circle—more participants entering the decentralized compute market leads to cheaper and more abundant compute.
Addressing the challenges
Some critics assert that decentralized compute lacks the performance of hyperscalers, citing latency and quality issues. The truth is quite the opposite. DePINs offer competitive performance in latency, concurrency, and throughput. Techniques like smart workload routing, mesh networking, and tokenized incentives for high availability help maintain performance and optimize it dynamically according to workload demands.
Moreover, certain DePINs have created transparent network explorers, allowing developers and investors to verify performance claims in real time. These mechanisms enhance the reliability and cost-effectiveness of DePINs compared to traditional providers.
DePINs also offer greater diversity than hyperscaler options. Over 13 million devices are currently online, enabling developers to access a wide variety of hardware and find the right tools for their AI projects, from high-performance cloud-grade GPUs to specialized edge devices.
A level playing field for AI
We have a fleeting opportunity to shape the technological landscape for future generations. While many US and Chinese corporations may have taken the lead, decentralized compute marketplaces present a promising alternative. By reducing costs and enhancing accessibility, startups, scaleups, researchers, and enterprises worldwide can compete more equitably. Emerging economies can develop models tailored to their languages, healthcare systems, cultural values, and financial requirements.
The question is not whether decentralization is essential, but how to onboard developers globally to this opportunity while simultaneously increasing the number of companies listing their excess compute on DePINs. Only through the decentralization of compute can AI become genuinely accessible and serve a broader audience, rather than just entrenched oligopolies.
Opinion by: Gaurav Sharma, CEO of io.net.
This article is for general information purposes and is not intended to be and should not be taken as legal or investment advice. The views, thoughts, and opinions expressed here are the author’s alone and do not necessarily reflect or represent the views and opinions of Cointelegraph.
