China's V2X plan to 2030: vehicle-road-cloud and C-V2X direct

In September the Chinese government published its five-year plan for connected and new energy vehicles. The document sets an architecture, names a radio, and fixes targets to 2030. The targets all describe the vehicle; the roadside and the network carry none.

The plan is the 15th Five-Year Plan for the intelligent connected new energy vehicle industry, circulated as Lian Gui Han [2026] No. 305 by the Ministry of Industry and Information Technology and eight other departments on 9 September 2026. The eleven-page annex is the plan itself. Read it next to a European roadmap and the contrast lies in which questions the document treats as settled.

The three tiers of vehicle-road-cloud

The organising concept is vehicle-road-cloud integration, che-lu-yun yitihua in the original. The plan uses it three times, always inside quotation marks, and never defines it. It is treated as a term the reader already knows, which tells you something about how long it has been circulating in Chinese policy.

What the plan does instead is enumerate the parts, once, in the box on application and promotion. Important intersections and road sections are to carry intelligent perception, edge computing and communication capability. Cloud-control base platforms are to be built with connections across administrative domains, which the plan calls cross-domain interconnection. Traffic-management and traffic-event information reaches the vehicle in tiers. And the functional target is large-scale application of cooperative perception and decision-making, both vehicle-to-vehicle and vehicle-to-road.

That last phrase is the one worth pausing on. Cooperative decision-making, not only cooperative awareness, puts part of the driving task outside the vehicle. The European collective perception model shares what sensors detect and leaves the decision in the car. These are different distributions of responsibility, and the plan picks one in a subordinate clause.

C-V2X direct communication, and four absent terms

On the radio the plan is explicit and oddly careful:

Build C-V2X direct-communication roadside equipment

That is the instruction, in the networked-infrastructure task. The plan glosses the acronym in its own text, as vehicle networking technology based on cellular mobile communication. A plan that stops to explain C-V2X to its reader is not written only for engineers.

Four terms never appear. PC5 is absent, though direct communication, the term the plan does use, is the same idea named functionally. LTE-V2X and NR-V2X are absent, so the plan sets no generation and leaves the sidelink evolution open. ITS-G5 and DSRC are absent too, and that silence is the loudest thing in the section. Europe spent a decade unable to choose between the two access layers ETSI standardised; China’s planning document does not record that there was a choice.

The vehicle side gets one sentence: raise the penetration of in-vehicle 5G/5G-A and C-V2X direct-communication capability. The verb is raise. There is no percentage, no target year beyond the plan’s own, and no baseline. The two technologies are also bundled into a single penetration rate, so no C-V2X-only figure can be derived from it even in principle.

Nationwide for signals, local for sensors

One sentence in the networked-infrastructure task does something a roadmap rarely does. It grades the ambition:

Promote nationwide digital upgrading of traffic signals, signs, road markings and other traffic facilities; carry out large-scale deployment of roadside perception facilities according to local conditions

Digitising signals, signs and markings is nationwide. Deploying roadside perception at scale is according to local conditions. Two clauses, one comma, two completely different commitments.

The grading is an engineering judgement. Digitising a signal head is cheap, the asset already exists, and the benefit is universal, since a SPATEM stream needs the controller to know its own phase and nothing more. A roadside perception installation is a sensor, a compute box, a power feed and a maintenance contract per intersection, and its value depends entirely on how much traffic passes and how blind the approach is. Mandating the first and permitting the second is what a ministry does when it knows the second does not pay off everywhere.

The box on application keeps the same discipline. Perception, edge and communication go to important intersections and road sections. Not all of them.

The quantified targets are vehicle targets

The plan’s 2030 targets sit in one unbroken paragraph, all governed by a single by-2030 clause at its head. Take any sentence out of that paragraph without the marker and a goal reads as a fact.

What the plan counts	What it leaves directional
70% of domestic new passenger-car sales are NEV, 40% for commercial vehicles	in-vehicle C-V2X penetration, raised by an unstated amount
3.3 L per 100 km average fuel consumption	roadside equipment, built in unstated quantity
around 11.5 kWh per 100 km for battery-electric cars	internationally led standards, reaching the global front ranks
labour productivity up 15% against 2025	autonomous driving, at scale application
the automotive industry peaks its carbon before 2030	vehicle-road-cloud, with no number of any kind

Nothing in the right column has a number. No count of roadside units, no kilometre of covered expressway, no named pilot city, no yuan of budget. The horizon is single, with no 2027 checkpoint and no 2035 vision.

The asymmetry is the finding. A document that commits to 3.3 litres per hundred kilometres is a document capable of precision when it wants to be. On the infrastructure its own architecture depends on, it declines. Read generously, that is a ministry refusing to invent a number it cannot yet defend. Read closely, it means the hard commitments land on carmakers, who can be measured at the factory gate, while the road side stays at the discretion of whoever pays for it locally.

One more absence is worth naming. In eleven pages about connected vehicles, the only international standards body mentioned is UN/WP.29, the United Nations forum for vehicle regulations, which China commits to participate in deeply. ISO, IEC, ITU and 3GPP do not appear. The declared venue for influence is type approval.

What to watch

The plan is a statement of architecture, so the thing to track is whether the architecture arrives. Three signals will show it before any announcement does. Whether the digitisation of signals and signs runs nationwide, because that is the only infrastructure promise made without a local escape clause. Whether cloud-control platforms interconnect across provincial boundaries, which is where cross-domain interconnection stops being a phrase. And whether the cooperative decision-making in that subordinate clause shows up in a Chinese standard with conformance tests attached, because that is the line between shared awareness and shared driving.

For what is already in the ground in China and the three other regions, see the deployment map. For the mechanism pulling V2X features into cars without waiting for a mandate, see the business model V2X was missing.