Google Cloud Next Tokyo:7/30、31 東京ビッグサイトにて開催!
TVer

How TVer built its in-house ad infrastructure on Google Cloud to power Japan’s largest video streaming platform

Google Cloud Results
  • Processed hundreds of millions of daily ad requests with millisecond-level latency using Cloud Bigtable and Memorystore

  • Secured flexible compute capacity to handle hundreds of thousands of requests per second with Google Kubernetes Engine

  • Eliminated massive cross-cloud outbound data costs by unifying the ad delivery infrastructure directly alongside BigQuery

  • Executed a zero-downtime migration of the live ad delivery system alongside Google Cloud Consulting

  • Powered an in-house ad platform supporting 650 million monthly video views entirely on Google Cloud

TVer scaled Japan's biggest streaming platform by building its ad server in-house on Google Cloud, hitting 650M monthly views with zero downtime.

Taking control of ad delivery to fuel hyper-growth

By its 10th anniversary in 2025, TVer had evolved into a massive video streaming platform. To sustain this rapid growth and provide the best possible experience for both viewers and advertisers, the commercial TV official video streaming service made a strategic decision to internalize its advertising server—the backbone of its business.

Mr. Yusuke Ono, Senior Executive Officer and Head of Ad Product Division, explains the critical business driver behind this move:

"For TVer, the ad delivery system is an essential business foundation that can truly be called the heart of our business. By bringing it in-house, we believed we could cycle through a series of tasks—such as experimenting with new ad formats, taking on the challenge of new features, and improving delivery speeds—much faster than before."

The top priorities for the project were high availability, low-latency performance, and scalability—the three pillars of any business-critical system. TVer needed a robust infrastructure capable of processing hundreds of millions of ad requests daily without delays, ensuring that sudden traffic spikes wouldn't hinder business growth.

Placing our delivery servers in the exact same environment as our powerful BigQuery data platform was the optimal choice for both cost and risk management. Furthermore, our engineers wanted to use Go as the development language, and its high compatibility with Google Cloud was a major deciding factor for the team.

Mr. Jundai Takashina

Product Development Department,TVer

When asked why TVer ultimately chose Google Cloud to meet these demanding infrastructure requirements, Jundai Takashina from the Product Development Department, who leads system development, emphasizes that "ease of data handling and cost efficiency" were the most important factors.

"Even before we brought the system in-house, we were already using BigQuery as our unified data platform to analyze and aggregate ad logs. If we had built the ad delivery infrastructure on another cloud, we would have had to transfer hundreds of millions of daily ad logs across clouds, resulting in massive outbound data costs."

"We also wanted to avoid the risk of losing highly valuable business data due to transfer issues. Placing our delivery servers in the exact same environment as our powerful BigQuery data platform was the optimal choice for both cost and risk management. Furthermore, our engineers wanted to use Go as the development language, and its high compatibility with Google Cloud was a major deciding factor for the team."

Building a high-throughput, zero-downtime foundation with GKE and Bigtable with support from PSO

The new system architecture uses GKE Standard for computing, paired with a strategic mix of databases—Memorystore for Redis Cluster (MRC) and Bigtable—depending on the specific requirements.

TVer Architecture Diagram

Mr. Jundai Takashina, who led the technical architecture, elaborates on the strategic technology choices:

"To achieve the millisecond-level latency required for the ad delivery API, we chose NoSQL over a relational database. We are strategically combining them: using Memorystore for Redis Cluster (MRC) for areas where data consistency is critical and an ultra-high-speed response is necessary, and Bigtable for areas where storage capacity and cost efficiency are prioritized. We adopted GKE Standard because we valued its flexibility, allowing us to secure capacity in advance to withstand real-time delivery where hundreds of thousands of requests per second are concentrated."

Without PSO's support, a launch on schedule would have been impossible. Their in-depth review, which went beyond just infrastructure configuration to encompass the application's algorithms and data structures, was the decisive factor in clearing rigorous load testing.

Mr. Yusuke Ono

Senior Executive Officer and Head of Ad Product Division at TVer

Executing this highly complex migration without interrupting live ad delivery was a critical, high-stakes mission. To achieve a zero-downtime transition, TVer partnered closely with Google Cloud Consulting (PSO).

Mr. Ono reflects on how this partnership secured the launch:

"Without PSO's support, a launch on schedule would have been impossible. Their in-depth review, which went beyond just infrastructure configuration to encompass the application's algorithms and data structures, was the decisive factor in clearing rigorous load testing. We are extremely grateful for their high level of expertise and speed, including their prompt provision of code-level improvement suggestions."

Improving the viewing experience with a foundation that balances stability and innovation

TVer has achieved the ultimate freedom of 'being in control' over their product roadmap by migrating to this newly centralized ad server. Rather than merely maintaining the status quo, the team is actively leveraging Google Cloud updates to push the boundaries of what their platform can achieve.

Looking ahead to the future of the platform, Mr. Takashina reveals:

“In addition to upgrading to the latest generation machines, we want to aim for an even stronger delivery foundation by actively incorporating the latest technologies, such as improving aggregation efficiency with Bigtable continuous materialized views and migrating to Memorystore for Valkey.”

Ultimately, this technological transformation is about serving the end-users—ensuring viewers get a flawless streaming experience and advertisers reach their audiences effectively. Mr. Ono concludes:

TVer founders

“Through this project, being able to turn the reliable and scalable technology that powers Google's own services into an organizational asset was a massive win for us. Having more options for public clouds and being able to compare them objectively has given us the confidence we need for future large-scale product development. Moving forward, TVer's ad business will continue to leverage the power of technology to deliver amazing viewing experiences to our users. And, of course, we are always looking for new team members to join us in this exciting, cutting-edge work.”

About TVer Inc

Established in 2006 as Presentcast Inc. through a joint investment by advertising agencies and key commercial broadcasters in Tokyo, the company launched the official commercial TV streaming service 'TVer' in 2015. This established a new viewing habit of "catch-up streaming" in Japan, allowing users to watch programs free from the constraints of time and location.

Following a corporate name change in 2020, TVer has grown into one of Japan's largest video streaming platforms, backed by 10 commercial broadcasters in Tokyo and Osaka, as well as major advertising agencies. By its 10th anniversary in 2025, the platform surpassed 90 million cumulative app downloads and 650 million monthly video views.

Driven by the mission "Opening up television for a more exciting future. Together with TVer, to a new world," the company is currently expanding its real-time streaming and Connected TV (CTV) capabilities, aiming to create a next-generation TV experience that seamlessly fuses broadcasting and telecommunications.

Industry: Media and Entertainment

Location: Japan

Products; Google Kubernetes Engine (GKE), Bigtable, Memorystore, BigQuery, Cloud Logging, Google Cloud Consulting (PSO)

  • Google Cloud プロダクト
  • 100 種類を超えるプロダクトをご用意しています。新規のお客様には、ワークロードの実行、テスト、デプロイができる無料クレジット $300 分を差し上げます。また、すべてのお客様に 25 以上のプロダクトを無料でご利用いただけます(毎月の使用量上限があります)。
Google Cloud