What’s up with… SoftBank, HKBN, Qualcomm

• SoftBank wants to build its own AI servers 
• HKBN builds a domestic AI factory
• Qualcomm joins OCUDU initiative

In today’s industry news roundup: Japan’s SoftBank is reportedly working with Nvidia and Foxconn on plans to become a domestic AI server manufacturer; Hong Kong operator is deploying domestic GPUs in its AI infrastructure-as-a-service platform; Qualcomm has taken a seat at the OCUDU Ecosystem Foundation’s top table; and more!

In what might herald a new twist in sovereign tech developments, Japan’s SoftBank Corp. is reportedly in discussions with AI chip giant Nvidia and Taiwanese contract technology manufacturer Foxconn about plans to produce home-grown GPU-based AI servers in Japan. According to the Nikkei news agency, as reported by Reuters, SoftBank plans to initially assemble externally sourced components before ultimately taking control of the whole production process itself. SoftBank Corp. is bullish about the near-term deployment of AI infrastructure in digital and communications networks, as it is not only investing in AI datacentres but is also one of the leading flag-wavers for AI-RAN deployments and is one of the founders of the AI-RAN Alliance.  

HKBN (Hong Kong Broadband Networks) is investing HKD100m (US$12.8m) in an AI infrastructure-as-a-service platform that “integrates leading domestic GPU computing power, advanced models, network connectivity and industry-specific application services”. The platform, dubbed the AI+ Domestic Computing Power Platform, has been designed to “provide businesses with a low-barrier and scalable AI deployment option, and to help address core pain points that enterprises face during AI implementation – including insufficient computing power, data security concerns and compliance risks – and marks a significant step towards building Hong Kong’s self-driven AI ecosystem, further cementing its status as a leading technology hub in Asia.” For once, this is an AI factory not built using Nvidia technology – instead, HKBN is deploying GPUs developed by domestic chip firms MetaX and Biren – and will be located at a datacentre in Cyberport. 

Wireless chip giant Qualcomm Technologies has joined the OCUDU (Open Centralized Unit/Distributed Unit) Ecosystem Foundation, a Linux Foundation-hosted initiative that aims to “advance open, secure and interoperable Open RAN centralised unit and distributed unit (CU/DU) implementations, as a Premier Member: Doug Knisely, senior director of technical standards at Qualcomm, is joining the OCUDU Ecosystem Foundation governing board. The OCUDU Ecosystem Foundation was founded in March this year with the support of an investment from the National Spectrum Consortium (NSC) and the FutureG Office, which awarded funds to DeepSig and Software Radio Systems (SRS) to build the initial software for the OCUDU. Other members include AMD, AT&T, DeepSig, Ericsson, Nokia, Nvidia, Samsung, SoftBank, Software Radio Systems and Verizon, along with 60 general and associate members. “Open industry collaboration is critical to enabling flexible, high-performance and interoperable RAN architectures that can scale across diverse cloud and edge environments,” stated Lorenzo Casaccia, VP of technical standards at Qualcomm. “By joining the OCUDU Ecosystem Foundation, Qualcomm Technologies looks forward to contributing its Open RAN and standards expertise to advance open-source CU/DU platforms and accelerate 5G and early 6G Open RAN network innovation and ecosystem,” he added. 

The European Parliament and the Council of the EU have agreed to introduce “simpler, innovation-friendly rules for artificial intelligence (AI),” the European Commission (EC) has announced. Five months ago, the EC proposed the Digital Omnibus on AI “as part of the EU’s simplification agenda to boost Europe’s competitiveness,” which will “make the implementation of the AI Act for EU businesses easier while maintaining its benefits for European society, safety and fundamental rights.” As part of the new agreement, the rules governing certain high-risk areas, including biometrics, critical infrastructure, education, employment, migration, asylum and border control, will apply from 2 December 2027 instead of the previous deadline of 2 August 2026. In addition, and following lobbying from the likes of German tech giant Siemens and Dutch AI chip-making infrastructure specialist ASML, machinery has been excluded from the AI Act as it is already subject to existing regulation, noted Reuters.   

Spanish industrial firm Amper, which specialises in dual-use technology (designed for use in civil as well as military applications) is acquiring Zaragoza-based defence and security communications specialist Teltronic from private equity firm Nazca (a significant shareholder in Amper) in a deal valued at up to €225m (including the assumption of debt). For Amper, the deal “represents a transformational and highly complementary acquisition from a technological, industrial and geographical perspective,” the company noted, as it is “expected to generate significant commercial, technological and operational synergies, while consolidating its positioning in key markets, such as Brazil and Mexico, and enabling access to others, such as the US and Canada.” Amper CEO Enrique López Pérez noted: “The incorporation of Teltronic is a transformational transaction for Amper, positioning us as a Tier 1 player in Europe in defence and security communications, while reinforcing our ambition to become the leading Spanish mid-cap company in dual-use technologies and strengthening our strategic relationship with Nazca.” The deal is worth noting as the provision of communications technology to the defence sector is becoming a major focus for companies such as Nokia, which is making headway with the US Department of War, and Ericsson, as well as many others.  

– The staff, TelecomTV