TIER IV, the Tokyo-based company behind the open-source autonomous driving software Autoware, has signed a memorandum of understanding with automotive supplier Astemo to jointly build a next-generation development platform for end-to-end autonomous driving AI, the companies announced on August 5, 2026. The platform is targeted for commercialization around 2030, with Astemo aiming to put end-to-end AI models into passenger vehicles in the early 2030s.
The deal centers on TIER IV’s Co-MLOps, a collaborative data-sharing system for autonomous driving development. Under the agreement, Astemo will license the Co-MLOps architecture and related technologies to build a platform covering the full loop of AI model development: collecting, processing, training, evaluating and continuously improving models that meet the quality and safety requirements for mass-production vehicles.
End-to-end (E2E) models fold perception, planning and control into a single learning process rather than chaining separately engineered modules, and they are emerging as a candidate foundation for next-generation driver-assistance and autonomous driving systems. TIER IV’s own framing is specific about where this fits its roadmap: the company is working toward what it calls Level 4+ autonomy, gradually expanding fully driverless operation into more complex environments, and it treats E2E models as a key technology for getting there.
“Bringing E2E autonomous driving AI models into real-world deployment requires advancing both mass-production readiness and AI development infrastructure in parallel,” said Shinpei Kato, founder and CEO of TIER IV, in the announcement. “We are delighted to deepen our collaboration with Astemo by providing licenses for TIER IV’s Co-MLOps solution and related technologies.”
Astemo’s executive vice president, CTO and CISO Hiroshi Tamagawa said the company aims to strengthen the internet-of-vehicles platform it is developing and to extend the technology beyond the automotive industry into broader mobility applications.
What the two companies agreed to build
The MoU sets out two tracks. The first is the platform itself: Astemo licenses the Co-MLOps architecture and, alongside TIER IV, builds a development environment that integrates data collection through model evaluation and continuous improvement, with commercialization around 2030. In parallel, Astemo will stand up an internal framework for continuously developing E2E models that satisfy mass-production quality and safety requirements.
The second track is a joint study of E2E models for Level 4+ autonomy. The companies will pool insights and technical assets from the platform work to define and align technical requirements for E2E autonomous driving AI, and to draft a medium- to long-term development roadmap. The stated endpoint: Astemo deploying Level 4+ E2E AI models in passenger vehicles in the early 2030s.
That division of labor plays to each side’s position. TIER IV leads development of Autoware and has been building AI development infrastructure around it; Astemo, the supplier formed from Hitachi’s and Honda’s automotive parts operations, brings what the announcement describes as functional safety compliance and vehicle integration experience across braking, suspension, powertrain and ADAS products.
How Co-MLOps works
Co-MLOps, short for Cooperative Machine Learning Operations, is TIER IV’s answer to a structural problem in autonomous driving AI: every company collects its own driving data and builds its own development pipeline, duplicating effort, while smaller players struggle to assemble enough data to train competitive models at all. TIER IV launched the project in January 2024 to let participating companies share managed sensor data — camera images and lidar point clouds — under privacy and security controls, and to run common MLOps functions on the pooled datasets. Participants keep their own proprietary technology, functional safety processes and quality controls.
An initial proof of concept in 2023 collected video data across eight regions, including Japan, Germany, Poland, Taiwan, Turkey and the United States, and used it to train multitask perception models optimized to run on under 10 watts of power. The platform is built on Amazon Web Services infrastructure and takes a cloud-native approach, with over-the-air updates to recognition models on the roadmap.
Astemo is already inside this ecosystem. Both companies are involved in the Autoware Foundation and the SOAFEE initiative, and previously collaborated on the reference implementation of Open AD Kit, a cloud-native development framework built on Autoware. Astemo has also been involved in the Co-MLOps project since its launch.
The road to this agreement
The MoU formalizes a partnership that has been building for nearly two years. In October 2024, TIER IV and what was then Hitachi Astemo launched a joint initiative on autonomous driving systems and software-defined vehicles, built around a SOAFEE-compliant architecture integrating Autoware as the core driving software. That effort also targeted mass-production vehicle models around 2030.
TIER IV’s technology line has been moving in the same direction. The company released an E2E architecture for Autoware in July 2025, then in March 2026 published data-centric AI software stacks for Level 4 autonomous driving, configurable either as a hybrid system (perception AI paired with diffusion-model planning) or as a full E2E pipeline drawing on world-model concepts. To validate them, TIER IV began roughly hour-long supervised test runs in Tokyo, Pittsburgh and Munich, in collaboration with the University of Tokyo, Carnegie Mellon University and the Technical University of Munich, using different vehicles, chips and sensor suites in each city. A safety driver rides in each vehicle as local regulations require.
The Astemo agreement extends that arc from software stacks to production plumbing. The MoU combines TIER IV’s AI development platform expertise with Astemo’s functional safety compliance and vehicle integration experience as a Tier 1 supplier. Astemo gets a licensed development platform for E2E models without building the MLOps machinery from scratch.
The first scheduled marker is commercialization of the joint development platform around 2030, followed by Astemo’s targeted deployment of Level 4+ E2E models in passenger vehicles in the early 2030s — dates the companies describe as targets, with the joint requirements study and roadmap work now underway.
