Kakao Mobility Signs NVIDIA MOU Targeting Level 4 Autonomous Driving

Kakao Mobility signed a memorandum of understanding with NVIDIA to develop Level 4 autonomous-driving technology, with Kia purpose-built vehicles serving as the first integration target, the company announced on November 9, 2026. The deal pairs Kakao Mobility's autonomous-driving AI and mobility platform with NVIDIA's computing, sensing and safety technology, and names the Kia PV5 as the initial model for the combined system.

Level 4 autonomy means the vehicle handles all driving tasks within defined conditions without human intervention. The stated goal is a Level 4 system aligned with global standards. Kakao Mobility grounds that ambition in urban driving it says has validated its technology in Seoul's Gangnam district — a claim the company makes about itself, and one that a memorandum of understanding does not independently verify.

The stakes are straightforward. A Korean mobility-platform operator gains a credible compute and safety partner; NVIDIA extends its automotive architecture into a new national market with a commercialization path already attached. The MOU is non-binding by nature, but it connects to a vehicle program that already exists on paper: Kakao Mobility's July 2026 partnership with Kia to develop purpose-built vehicles for autonomous-driving services.

What hardware is Kakao Mobility committing to?

The technical core of the agreement is a planned autonomous-driving computing unit containing two NVIDIA DRIVE AGX Thor chips. Kakao Mobility says the dual-chip unit is intended to process large-scale AI computation in real time. Thor is NVIDIA's automotive-grade processor for in-vehicle AI workloads.

That hardware plan matters because of the software it is meant to run. Kakao Mobility is progressively developing an end-to-end autonomous-driving model — a single model that connects perception through decision-making, rather than separate modules stitched together. End-to-end models place heavy demands on in-vehicle compute, and the company is explicitly linking its development roadmap to the computing capacity required to run it in a vehicle.

The caveat: a computing unit that a company "plans to introduce" is a plan, not a product. The announcement does not say when the dual-Thor unit will be built, tested, or qualified for deployment.

Where do Hyperion and Halos fit in?

Two NVIDIA platforms frame the system architecture. NVIDIA describes Hyperion as a standardized reference architecture integrating compute, sensors, networking and safety systems, with DRIVE AGX Thor forming part of the platform. Kakao Mobility intends to use the architecture to optimize connections among the vehicle, sensors and computing unit while pursuing system stability and compatibility.

The safety work extends into software development and validation through NVIDIA Halos. NVIDIA describes Halos OS as a three-layer safety architecture built on ASIL D-certified DriveOS foundations, integrating safety middleware and deployable safety applications. ASIL D (Automotive Safety Integrity Level D) is the highest functional-safety rating under the ISO 26262 automotive standard, reserved for systems whose failure carries the most severe risk.

Together, Hyperion and Halos supply the system architecture and safety framework around Kakao Mobility's own driving software. In plain words: NVIDIA is contributing the chassis of the computing and safety stack, and Kakao Mobility is expected to make its driving model run safely inside it.

Why the Kia PV5 is the vehicle that matters

The companies will prioritize integrated development on Kia PBVs — purpose-built vehicles, chassis designed for specific services such as robotaxi or delivery use rather than adapted consumer cars. The work builds on Kakao Mobility's July 2026 partnership with Kia, and the companies identify the Kia PV5 as the initial model for the combined technology.

The plan is to develop and supply vehicles for Level 4 services using Kakao Mobility's software alongside NVIDIA's AI computing and safety systems. That gives the agreement a specific integration path: Kakao Mobility contributes the driving software and mobility platform, NVIDIA contributes computing and safety technology, and the Kia PBV is the first vehicle target.

Commercialization remains an objective of that development work. Note the wording — an objective, not a launch date. The announcement contains no timeline for a commercial driverless service, no production schedule for the PV5 integration, and no disclosed financial terms for the NVIDIA agreement.

What did the executives say?

Kakao Mobility CEO Ryu Gung-seon framed the collaboration in ecosystem terms. He said the partnership is intended to accelerate end-to-end development and prepare for Level 4 commercialization through an ecosystem spanning software, components and finished vehicles.

NVIDIA automotive vice president Rishi Dhall framed the nearer objective more narrowly. He described the goal as accelerating safe and scalable autonomous-driving services in South Korea. The difference is telling: one CEO points at commercialization and a full stack; the NVIDIA executive points at services in a single national market.

What comes next — and what is not yet answered

The companies are also leaving open medium- to long-term global expansion, based on technology validated in South Korea. That expansion is aspirational language in an MOU, contingent on a system that does not yet exist as an integrated whole.

The immediate test is integration on the PV5. Turning separate software, compute, safety and vehicle contributions into a validated Level 4 system is the next step toward the commercial driverless service described in the announcement. Each element — the end-to-end driving model, the dual-Thor computing unit, the Hyperion architecture, the Halos safety layers, the Kia vehicle platform — has a stated role but no demonstrated record of operating together.

For the Korean autonomy market, the deal signals that domestic mobility operators intend to compete on the end-to-end model paradigm, using NVIDIA's silicon and safety frameworks rather than building full stacks alone. For NVIDIA, it adds another national ecosystem to an automotive platform push that now reaches from Korean PBVs outward. Whether the PV5 integration produces a validated Level 4 system, and on what schedule, will determine whether this memorandum becomes a product or stays a partnership on paper.