Dr Fazal Ali
Next is now. The New Silk Roads inscribe a geopolitical and technological realignment in which AI, not silk or spices, is the most prized commodity.
The screen economies of “Next” turn on hyperscale campuses; parallel processing; extreme power density; rapid scalability; specialised accelerators (GPUs/TPUs), ultra-high-bandwidth networks; liquid cooling systems; megawatt-to-gigawatt power capacity to handle intensive training and inference workloads; ultra-fast fabrics; disaggregated, scalable storage; and dynamic orchestration that allocates memory, storage, and compute resources in real-time to match workload demands.
Digital assets are the most prized commodities along the New Data Silk Roads. The trauma of trade along the Old Dirt Silk Roads worsened with assaults on caravans by looters and raiders. Today, along the New Digital Silk Roads, bad actors launch prompt-injection and AI-distillation attacks from behind liquid-crystal displays and exit nodes.
These AI-distillation attacks use third-party proxy services that act as grey-market middlemen. These networks manage sprawling “hydra clusters” of tens of thousands of fraudulent accounts to bypass regional blocks and rate limits.
Prompt-penetration perpetrators type malicious instructions into a chat interface or an API. They use persuasive personas, override commands such as “Ignore previous instructions”, or suppression techniques to bypass the AI’s safety guardrails. In an indirect attack, bad actors hide malicious prompts inside external content that the AI processes later.
When the AI summarises a web page, reads an uploaded document/PDF, or scans a database containing this hidden text, it gets tricked into following the attacker’s commands rather than the developer’s instructions. Attackers can also embed malicious text instructions in an image’s metadata or an audio file, which the AI then executes when processing the media.
Along the new Digital Silk Roads, Crime-as-a-Service (CaaS) is a commercialised, subscription-based business model where illicit actors sell or rent specialised tools, infrastructure, and expertise to other individuals.
This lowers the technical barrier to entry. Actors with little to no hacking skills can launch sophisticated attacks. This ecosystem mirrors the legitimate “Software-as-a-Service” (SaaS) industry and features modular services, IT-style help desks, and pay-per-use economics.
The CaaS market has a diverse bundle of services. The Ransomware-as-a-Service (RaaS) model involves developers who create ransomware payloads and partner with affiliates who deploy them. Profits are split between the developer and the affiliate.
Phishing-as-a-Service (PhaaS) offers pre-built kits containing fake login pages, email templates, and credential-harvesting tools. DDoS-as-a-Service (DDoSaaS) is a playbook that prescribes renting access to networks of compromised devices to flood and take down websites during distributed denial-of-service attacks.
Malware-as-a-Service (MaaS) uses a subscription model for malicious code. Beyond digital crime, the “as-a-service” model has expanded into the physical realm as Violence-as-a-Service (VaaS).
This involves a decentralised, four-stage supply chain where instigators order a violent act such as assaults or vandalism, and recruiters on social media or encrypted messaging apps hire inexperienced individuals, frequently minors, to execute the crime. The primary target remains the various layers of AI cluster architecture.
The compute layer will usually house core processors such as Nvidia H100 or H200 GPUs or Google TPUs that perform the matrix operations. Interconnection fabrics form another layer. These ultra-fast intra-node connections, such as Nvidia NVLink, and inter-node network fabrics, such as InfiniBand or RDMA, allow thousands of GPUs to communicate simultaneously without slowdown.
The storage layer consists of high-speed storage arrays that can instantly feed massive training datasets and save model states. At the cluster orchestration layer, workloads are managed, and jobs are scheduled, with resources allocated efficiently across thousands of nodes.
Because they coordinate tens of thousands of processors, AI clusters are resource-intensive. A single cluster can draw megawatts of power, which requires liquid or immersion cooling to dissipate the extreme heat generated during operation.
Along the “New AI Silk Roads”, emerging markets, particularly in Eurasia, are forging localised AI ecosystems, sovereign models, and independent data infrastructures.
Currently, the infosphere is defined by three major economic and technological corridors. China’s Digital Silk Road (DSR) is an infosphere layer of its Belt and Road Initiative. China is considering exporting dilution refrigerators and quantum cloud platforms to the Global South. This allows Beijing to scale its AI ecosystem globally while testing advanced algorithms in diverse overseas environments.
The Euro-Indian Convergence is an alternative tech ecosystem that links Europe’s heavy industry with India’s immense pool of semiconductor designers. This corridor focuses on sovereign supercomputing (the GANANA grid) and harmonised AI regulations that link European HPC Centres of Excellence and top Indian research institutions.
The Middle Eastern Tech Hubs are investing heavily in AI-native infrastructure, such as 6G networks. The technology is still in the experimental and developmental phase globally. The UAE is aiming for a commercial rollout between 2029 and 2030.
Next is fracturing into distinct regional hubs, each with its own cultural, linguistic, and regulatory AI standards. Next remains uneven and unequal.
Dr Fazal Ali completed his Master's in Philosophy at the University of the West Indies. He was a Commonwealth Scholar who attended the University of Cambridge, Hughes Hall; the Provost of the University of Trinidad and Tobago; the acting President of UTT; and the Chairman of the Teaching Service Commission. He is the President of NIHERST and an external services consultant with the IDB.
