In the development of artificial intelligence (AI) infrastructure, 'computing power, algorithms, data' are regarded as the three indispensable pillars. In the first two years since the advent of GPT, computing power deployment became the core narrative of AI infrastructure. Major enterprises have been laying out graphics cards and data centers, with various Web3 mining companies like Coreweave fully transforming into AI data centers, while distributed computing represented by IO and Aether has become the industry focus. With the advancement of the industry, the bottleneck of computing power is no longer the most critical issue at this stage, and the extension on the edge side will be the new focus in this field. The value of blockchain in the computing power field mainly lies in promoting user participation through token incentives.
The development of algorithms and models is the current focus, especially the release of DeepSeek V3, which has caused a huge stir in the industry. Can the development in this field significantly reduce the costs of AIGC? When running ultra-large models, can a distributed structure be adopted to realize the architecture of the model in the face of constraints such as electricity and resources? Additionally, can past frameworks for distributed computing that incorporate privacy protection, such as TEE, secure multi-party computation, federated learning, and homomorphic encryption, become practical? These are all key points for industry implementation at present. There are still not many valuable crypto projects in this area, but from the perspective of the entire AI industry, the road may be difficult but solutions can always be found.
The real challenge lies in the data layer—the lifeline of AI development. Do we really have enough clean data for training? Currently, all public domain data has basically been scraped clean by mainstream big companies, which has become the key point where Crypto can truly change the AI industry.
1. Ensure high-quality, clean training data.
2. Breaking down data silos, especially the barriers of private domain data, to achieve data flow while protecting privacy and authenticity.
This is precisely where the Web3 project Roam—a Depin project and decentralized wireless network—breaks through. Roam's global open wireless network not only promotes connectivity but also transforms private domain data into accessible and privacy-protected AI-usable datasets, turning them into AI foundational data that can be called upon.
How Roam disrupts data acquisition
Roam's innovation lies in addressing one of the most urgent data challenges for AI: the inaccessibility of operator-level data. Traditionally, telecom companies have rich time, location, and user-related data. However, due to structural barriers, even tech giants like Google and Meta find it difficult to access these datasets. Instead, these companies choose to create their own ecosystems, such as iOS and Android, exacerbating data monopoly and fragmentation. In the AI era, this monopoly hinders comprehensive model training and global data integration.
With the support of the Wireless Broadband Alliance, WiFi Alliance, and GSMA member units, Roam has developed the Global WiFi OpenRoaming + eSIM Top Up product. Starting with the construction of a new generation of WiFi OpenRoaming networks, it builds a new type of network that is free, secure, and roamable globally. Meanwhile, leveraging itself as a top global eSIM traffic supplier, it transforms previously paid, closed operator services into an open, globally unified new type of network. Users who participate in OpenRoaming construction receive free eSIM traffic rewards. When users arrive in any country, the eSIM Top Up product allows traffic to automatically switch to local traffic. Ultimately, this enables users to have a network that can automatically log in with on-chain DID without incurring expensive roaming fees, whether at home or on the road, whether in their residence country or while traveling. Compared to traditional operators and even some Web3 operators' business models, Roam represents a truly multinational, future operator model that transcends different connection methods. AI companies can hardly gain the support of all traditional operators and internet/mobile giants to train specific models. However, Roam has opened a channel to provide operator services through WEB3 means, and in return, the data generated in the process can become a data source for all AI companies under privacy protection.
Roam provides a mechanism for how Web3 can change data generation management, thereby breaking the private domain data barriers encountered by AI, and achieving real cases of the flow of traditional private domain data. The transformation of the business model often brings rapid changes. As of Q4 2024, Roam has become one of the largest decentralized physical infrastructure (DePIN) projects in the world. The Roam Discovery Program further expands the boundaries by integrating multiple DePIN ecosystem partners to enrich AI training datasets. Roam can not only provide operator-level data to AI but also collaborate with ecosystem partners to provide weather data, CDN data, energy data, authentication data, payment and lending data, and even data on the social, payment, and operational status of AI agents to AI models.
In addition to breaking the traditional business model and opening a path to utilize WEB3 to provide a large amount of previously confined data for AI society, Roam has some additional important contributions to the infrastructure data layer for AI.
Firstly, the security and privacy protection of data across devices and regions. The core technology of Roam is to upgrade public WiFi into a secure login mode combined with DID. At the same time, it achieves cost-free roaming of WiFi and eSIM integrated networks. The private key of the user’s blockchain, the private key of DID, the certificate of WiFi OpenRoaming, and the certificates behind eSIM are all unified and coordinated. All of this can be managed and encrypted through a TEE-based KMS (Key Management System) within Roam's existing devices. Roam's data interaction complies with W3C DID and verifiable credential standards, and combined with ZK technology, it can alleviate concerns about data security. The Roam technology team has also done some of the earliest work in the world to combine blockchain-based digital identity with eSIM, WiFi, and Bluetooth chips.
Secondly, leveraging these new multidimensional spatial data such as time, space, and identity to help analyze which data is generated by real society and which is produced by AI. A combination of Roam's 3W data (Who, When, Where) and payment data generated from WiFi fingerprint payment terminals, along with tags from different data collection devices, can ensure that this data is not generated by AI but from real life. A broad-spectrum, multidimensional spatial data evaluation system like this is key to ensuring the value of AI training data. The Roam development team has a decade of long-term accumulation in this field.
Thirdly, it facilitates AI agents to expand new living spaces and generate new data. They currently all live in Twitter and TG, but with Roam's help, they can exist in various smart home devices. The core support Roam receives from major industry players like Samsung is that the devices from these big companies are almost all WiFi devices. Currently, the only way these devices connect to the network is by manually entering usernames and passwords, which is very unsafe and not automated. In the world of Roam, AI agents, humans, and devices are all unified under DID tags. They all have corresponding WiFi OpenRoaming certificates. This constructs a network foundation for AI agents to automatically transition from the internet world to the IoT world. During the process of an AI agent moving from one device to another, all data generated is encrypted, protected, and stored with the help of its private key in TEE, and combined with OpenRoaming WiFi and blockchain, it allows for secure transmission and distribution. The entry of AI agents into the IoT world while achieving cross-domain data interoperability means they can truly become companions to humans. Human interactions with them no longer require humans to focus on a screen. Regardless of the scenarios humans are in, they can record in real-time, respond quickly, and simultaneously construct a real avatar of humans. This is also the first step in humanity's transition from a biological to a silicon-based civilization.
Roam redefines the data infrastructure of the AI era with its open and secure global network. From breaking the traditional operator model to supporting the construction of AI data layers, Roam is driving the deep integration of AI society and Web3 through dual innovation in technology and business models. In this transformation led by Roam, we not only see a significant enhancement in global network services but also the emergence of the prototype of a silicon-based civilization.