The future of cloud computing : hyperscale and DePIN in perspective: a comparative study of centralized and decentralized GPU infrastructure models
Valliovuo, Iiro (2025)
2025
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2025121637349
https://urn.fi/URN:NBN:fi:amk-2025121637349
Tiivistelmä
Cloud computing forms the backbone of modern digital infrastructure, with hyperscale data centres operated by providers such as AWS, Google, and Microsoft dominating the global market. These centralized infrastructures deliver high levels of performance, scalability, and reliability, yet they also raise challenges concerning cost transparency, vendor dependence, and governance concentration. Alongside these established models, a new approach is emerging through Decentralized Physical Infrastructure Networks (DePIN), specifically in the realm of GPU colocation, which distributes computing capacity across open ecosystems coordinated by blockchain technology and token-based incentives.
This thesis examines the evolving landscape of cloud infrastructure by systematically comparing centralized hyperscale systems with decentralized alternatives, focusing on Akash Network, Render Network, and Aethir. The study applies a hybrid infrastructure assessment methodology to evaluate four key dimensions: cost efficiency, performance reliability, governance structures, and environmental sustainability.
The findings indicate that centralized infrastructures remain advantageous for workloads requiring regulatory compliance, deterministic performance, and strict service-level guarantees. In contrast, decentralized networks demonstrate distinct advantages in cost optimization – offering pricing significantly lower than hyperscale rates – and sustainability through the reuse of underutilized hardware. While decentralized governance introduces transparency, it currently faces challenges in decision speed compared to corporate models. The thesis concludes that the two paradigms are complementary rather than mutually exclusive, suggesting that a hybrid architecture combining centralized stability with decentralized flexibility represents the most viable direction for future cloud ecosystems.
This thesis examines the evolving landscape of cloud infrastructure by systematically comparing centralized hyperscale systems with decentralized alternatives, focusing on Akash Network, Render Network, and Aethir. The study applies a hybrid infrastructure assessment methodology to evaluate four key dimensions: cost efficiency, performance reliability, governance structures, and environmental sustainability.
The findings indicate that centralized infrastructures remain advantageous for workloads requiring regulatory compliance, deterministic performance, and strict service-level guarantees. In contrast, decentralized networks demonstrate distinct advantages in cost optimization – offering pricing significantly lower than hyperscale rates – and sustainability through the reuse of underutilized hardware. While decentralized governance introduces transparency, it currently faces challenges in decision speed compared to corporate models. The thesis concludes that the two paradigms are complementary rather than mutually exclusive, suggesting that a hybrid architecture combining centralized stability with decentralized flexibility represents the most viable direction for future cloud ecosystems.