Author: Alex Goh, CoinDesk; Translated by Wuzhu, Golden Finance
Younger readers may not remember, but cloud computing was once the future. The advent of unlimited computing and storage resources represents one of the few technological "revolutions" that deserves the name. But the age of AI has rendered the centralized cloud model not only obsolete, but also a danger to those who build on it and to every user.
If that sounds a bit exaggerated, consider the recently discovered vulnerability affecting Hugging Face, a major AI-as-a-Service platform. This vulnerability could have allowed a tampered model uploaded by a user to execute arbitrary code through its inference API function to gain escalated control. Fortunately, the vulnerability was discovered in time and does not seem to have seriously affected users - although the researchers noted that such vulnerabilities are "far from unique."
The problem here is not AI at all; it is the outdated, centralized X-as-a-Service model, in which there is no incentive to ensure the security of the system or to develop the applications that the market and ordinary users need. The preferred future of AI — secure, reliable, and, most importantly, able to harness vast amounts of computing resources — can only be achieved by disrupting cloud computing and embracing the decentralized revolution.
‘Big Cloud’ and the Monopoly of AI
Giant companies like Microsoft, OpenAI, Google, and Amazon dominate the AI space because they have the vast financial, human, and computing resources needed to make it work at scale.
This is terrible for the development of AI and the exact opposite of its democratizing potential. When algorithms and applications are built by a small group of developers at a trillion-dollar California company, it lends AI agents narrow, one-dimensional, and incredibly subjective biases. Everything from financial services to creativity… to human interaction is affected.
The technical arguments against market monopolies in AI are equally compelling. Throughout the training process, AI must constantly absorb new data, including data from other AI applications. However, the current trend toward centralization in Big AI means that platforms and applications remain highly siloed, even with an open source model. This stymies innovation and leaves room for buggy or malicious applications that can multiply with dizzying and potentially catastrophic consequences.
More importantly, the centralized model presents huge and clear risks when it comes to protecting users’ personal data, privacy, and, in many cases, financial information. When one entity holds a large amount of sensitive and business-critical data, it represents a single point of failure for attackers and enables providers to censor or deny service to their users based on arbitrary and indisputable decisions.
Democratization through Decentralization
When it comes to AI, the cloud model is clearly a dangerous dead end. AI requires such staggering computing power that it stretches even the capabilities of hyperscale centralized cloud platforms and the microchip industry that services it. The chip shortage is so severe that the H-100 servers used by the industry’s most advanced AI applications now require a 52-week waiting list.
By decentralizing, we can create a network of nodes that tap into a vast reserve of unused CPU power, eliminating this problem once and for all. This modular approach to decentralized physical infrastructure (DePIN) is perfect for a variety of reasons: it is almost infinitely scalable, much cheaper than installing new servers with a cloud provider (costs are typically around 80% less), and it helps solve the silo problem of parallel computing and AI, allowing applications to more easily learn from each other. In addition, decentralized AI powered by blockchain technology offers innovative ways to reward creators of large language models (LLMs) through crypto tokens and smart contracts — providing a sustainable and fair model for rewarding innovation and contributions in the field of AI.
The rise of new economic models — especially those based on digital tokens — not only increases the need for more secure decentralized infrastructure; it also supports it. Building the AI ecosystem on a token economy incentivizes developers to create more secure AI agents and enables them to deliver these models to crypto wallets for ownership. This gives users complete peace of mind that their data is their own and cannot be shared without their knowledge or permission.
Perhaps most importantly, the token model means that AI projects will provide what the market really wants and needs, because compute and storage costs reflect the iron laws of supply and demand. Under the current monopoly, AI has no incentive to serve real-life needs. In a decentralized scenario, users themselves can reward developers based on the popularity of an AI agent or the benefit it brings to the world. This is a far cry from the big tech oligarchs that currently (but won’t for long) rule AI.
Decentralization also provides an answer to the vulnerabilities we’ve seen in platforms like Hugging Face. With the rapid development of blockchain technology, and specifically zero-knowledge (ZK) proofs, we now have a range of tools to ensure the security and provenance of AI applications. For those of us who understand the space, we often forget the speed and depth of this technological change. It’s not that traditional cloud providers are doing their best to preserve outdated models; it’s just that decentralization and ZK are recent inventions, and it will naturally take some time for industry players to realize how best to apply them to their (and their customers’) interests.
This is largely a question of education: showing that decentralized AI architectures, when built correctly, are private and secure by design, with all on-chain data encrypted, but still enabling interaction and collaboration between different projects, nodes, and parties.
For AI, centralization doesn’t work at any level: technical, philosophical, ethical, or market. More importantly, I suggest that as people grow increasingly tired of (and wary of) Big Tech’s outsized influence — from developers to technology providers to everyday users like you and me — the time for our own revolution has clearly arrived.
Kikyo
Catherine
Weiliang
Alex
Miyuki