Australia Gets First Blackwell-Powered AI Factory
Sharon AI and Cisco, in partnership with Nvidia, have launched Australia’s first Cisco Secure AI Factory. The facility is built around Nvidia's Blackwell architecture and is designed to support sovereign AI workloads with a focus on security and data residency, targeting highly regulated industries.
- The Nvidia Blackwell architecture is a significant leap from the previous Hopper generation, featuring a dual-die GPU design with 208 billion transistors. This design allows for up to 2.5 times the performance of the prior generation, and up to a 25x reduction in cost and energy consumption for some AI models. The architecture introduces a second-generation Transformer Engine and new Tensor Core technologies that support 4-bit floating-point (FP4) AI inference, doubling the effective model size and compute throughput. - The Cisco Secure AI Factory is a reference architecture, not just a hardware bundle, integrating compute, networking, storage, and security. It combines Cisco's UCS servers and Nexus networking with Nvidia's GPUs and AI Enterprise software, embedding security at every layer with tools like Cisco AI Defense and Hypershield. The modular design allows customers to build their own or opt for turnkey solutions. - Sovereign AI capability is a key driver for this facility, ensuring that Australian organizations can keep their data and AI processing within national borders, subject to Australian laws and privacy regulations. This addresses growing concerns around data residency and the legal risks associated with foreign jurisdictions, aligning with Australia's National AI Plan. - The facility will be equipped with 1,024 Nvidia Blackwell Ultra GPUs and utilize VAST Data's storage systems to handle high-throughput AI workloads. Hosted at NextDC's S3 data center in Sydney, it represents a significant investment in high-performance computing infrastructure for the Asia-Pacific region. - Sharon AI, founded in 2024, describes itself as a "neocloud" company focused on high-performance computing and GPU infrastructure. The company recently listed on the Nasdaq (SHAZ) and has managed investments in over 300 megawatts of HPC capacity in Australia and the US. - The competitive landscape for AI chips is intensifying, with hyperscalers like Google (TPU), AWS (Trainium), and Microsoft (Maia) developing custom ASICs to optimize performance and cost for their specific workloads. While Nvidia currently dominates the AI accelerator market, this "build vs. buy" trend is expected to lead to a more fragmented market by 2027. - Go-to-market (GTM) strategies are increasingly powered by AI, with a major trend for 2026 being the rise of "agentic AI" that can execute multi-step sales and marketing tasks autonomously. High-performing revenue teams are moving away from siloed point solutions towards unified platforms that consolidate CRM data, marketing signals, and customer insights to enable real-time, data-driven actions. - Strategic partnerships are crucial in the semiconductor industry to manage the immense costs and expertise required for research, design, and manufacturing. Collaborations between chip designers, manufacturers (foundries), software developers, and research institutions are essential for advancing new technologies like advanced packaging and AI-driven chip design.
