6G steps out of the lab with four times the bandwidth — and a network that can sense

The least consumer-friendly object on show when the ninth China International Import Expo opened in Shanghai on 5 November was a piece of test equipment considerably larger than a phone. Qualcomm's 6G terminal prototype, described as the first to match the latest 3GPP definition, had previously lived mainly in the laboratory; this was its first public physical appearance at a large exhibition. Prototypes and products are usually separated by years, so the useful reading is not that a launch is near, but that the verification phase has produced something a manufacturer is willing to put on a stand and let the supply chain inspect.

The number everyone quotes is bandwidth: the device has been tested at 400 MHz downlink, against 100 MHz in the 5G era. More important than the multiple is what the hardware has been doing. It is not a concept mock-up: it has been used in joint technical testing and interoperability verification with industrial partners including China Mobile and ZTE. In communications, getting one box to talk to another vendor's box is a harder and more informative milestone than a peak figure recorded in isolation, because it surfaces the ambiguities that a written specification leaves behind.

Yonggang Li, technical marketing director at Qualcomm China, offered a historically grounded reading: judging by the 5G experience, large test prototypes of this kind eventually become chipsets and products. First-generation equipment is almost always cabinets and cabling before it is compressed into baseband silicon and modules. The bulk on display is therefore evidence of an early engineering stage rather than a barrier to deployment — though it says nothing about a commercial timetable, which still depends on standards and spectrum decisions that have not been settled.

Reducing 6G to faster data would miss the more substantive change. At the Light of Internet expo in Wuzhen, an end-to-end 6G prototype system combined communication, sensing and computing in a single setup and demonstrated remote control of construction machinery nearly a thousand kilometres away. Transmission was one function among several: the system was simultaneously measuring its environment and allocating compute. A link that must deliver a command at a determinate moment, while continuously knowing the state of the channel and the machine at the other end, is a different engineering problem from one that must simply move a video stream quickly.

The GSMA chair puts the distinction plainly: the biggest difference between 6G and 5G is sensing. Future networks should not only carry information but track objects in physical space in real time and use artificial intelligence to diagnose, repair and optimise themselves. The base station stops being a passive pipe and becomes a measuring node. That reframing explains why the industry keeps describing 6G as designed for an AI era: inference is increasingly split across device, edge and cloud, and the network is asked to participate in scheduling rather than merely transport the resulting traffic.

Coverage is the other half of the story. Zhang Huiqiang, a 6G test director at the Zhongguancun research consortium, points to integrated space–air–ground networks that could keep deserts, remote mountain regions, oceans and the sky connected. Terrestrial base-station density is limited by economics and geography, so folding non-terrestrial layers into one architecture addresses whether a connection exists at all rather than how fast it is. For a globally distributed audience, that may matter more than any urban peak rate. Glasses-free real-time 3D video sits at the opposite end, pushing bandwidth, latency and compute coordination at once — an application that cannot be solved by making the handset alone more powerful.

Several other firsts on the same show floor fill in the physical side of that chain. An immersion-cooled cabinet made its public debut, pointing at the heat problem in AI data centres; SK hynix presented a memory matrix spanning server, data-centre and mobile endpoints against a backdrop of sustained memory price increases through 2026; a driverless electric vehicle without a steering wheel or pedals, and a high-speed turn-mill machine tool in its Asian debut, showed the same pull toward placing capability close to the user on both the consumer and industrial sides. Before a network can compute faster, it has to be powered, cooled and supplied with storage.

None of this means 6G has arrived. Standards are unfinished, spectrum and terminal form factors remain under discussion, 400 MHz is a test condition rather than a speed anyone will see soon, and interoperability work is one early link in a long chain. The value of a public prototype is that it moves the argument from slides to measurable hardware: specifications can be debated indefinitely, but eventually a device has to connect, produce numbers and interwork with someone else's equipment. Once sensing, compute and communication are validated together in the same machine, the definition of the next generation shifts from how many gigabits it carries to how much judgement it can take on for the person using it.