Rakuten Launches AI Commerce Platform With Custom Model

Rakuten, Japan's largest e-commerce conglomerate, has officially launched an AI-powered commerce platform built on a proprietary foundation model designed specifically for retail and online shopping. The move positions the $12 billion company as a direct competitor to Amazon's AI-driven commerce capabilities and signals a broader trend of major retailers developing in-house AI infrastructure rather than relying on third-party providers like OpenAI or Google.

The platform, integrated across Rakuten's sprawling ecosystem of over 70 services — including Rakuten Ichiba, its flagship marketplace — leverages the custom model to deliver hyper-personalized product recommendations, dynamic pricing optimization, and AI-generated product descriptions at scale.

Key Facts at a Glance

• Rakuten's custom foundation model was trained on over 15 years of proprietary commerce data, encompassing billions of transactions across its ecosystem
• The platform serves more than 100 million active users in Japan alone, with plans for international expansion
• Unlike OpenAI's GPT-4o or Google's Gemini, the model is purpose-built for commerce tasks rather than general-purpose reasoning
• Rakuten has invested an estimated $500 million in AI infrastructure over the past 2 years
• The company previously released Rakuten AI 2.0, a 7-billion-parameter open-source language model, in early 2024
• Initial benchmarks suggest the commerce model outperforms general-purpose LLMs on product understanding tasks by approximately 35%

Custom Foundation Model Targets Commerce-Specific Tasks

Rakuten's decision to build a domain-specific foundation model rather than fine-tuning an existing general-purpose LLM reflects a growing industry trend. Companies with massive proprietary datasets increasingly find that purpose-built models deliver superior performance on specialized tasks compared to even the most capable general models.

The commerce foundation model reportedly handles several critical functions. These include semantic product search that understands natural language queries like 'lightweight running shoes for rainy weather,' automated product categorization across millions of SKUs, and real-time demand forecasting that adjusts inventory recommendations.

Rakuten's AI research division, led by Ting Cai, has been building toward this moment since at least 2023. The team leveraged Rakuten's unique position as a company that operates across e-commerce, fintech, mobile telecommunications, and digital content — giving the model training data that spans the entire consumer journey.

How the Platform Changes the Shopping Experience

For consumers, the most visible change arrives in the form of a conversational shopping assistant embedded directly into the Rakuten Ichiba marketplace. Unlike basic chatbots that simply redirect users to search results, this assistant can engage in multi-turn conversations, compare products across merchants, and even factor in a user's purchase history and loyalty point balance.

The platform introduces several consumer-facing features:

• AI Shopping Concierge: A conversational interface that helps users discover products through natural dialogue rather than keyword search
• Smart Bundle Recommendations: The model identifies complementary products across different merchants, optimizing for both relevance and Rakuten Super Points earnings
• Dynamic Review Summaries: AI-generated summaries of thousands of product reviews, highlighting key pros, cons, and use-case-specific insights
• Visual Search Enhancement: Users can upload photos to find similar products, with the model understanding context, style, and functional attributes
• Personalized Deal Alerts: Predictive notifications based on browsing patterns and anticipated needs

Merchants on the platform also benefit significantly. The AI system generates optimized product listings, suggests pricing strategies based on competitive analysis, and provides demand forecasting tools that were previously available only to enterprise-scale retailers.

Rakuten's AI Strategy Diverges From Western Tech Giants

Rakuten's approach stands in stark contrast to how Western e-commerce players have deployed AI. Amazon has primarily focused on using AI to optimize its logistics and advertising operations, with consumer-facing AI features like Rufus arriving relatively late. Shopify has taken a platform-agnostic approach with Shopify Magic, offering AI tools to merchants regardless of where they sell.

Rakuten's strategy is more vertically integrated. By building the foundation model in-house and tightly coupling it with its ecosystem — which includes Rakuten Mobile (Japan's 4th-largest carrier), Rakuten Bank, and Rakuten Travel — the company can create cross-service AI experiences that no competitor can easily replicate.

This ecosystem advantage mirrors what Apple has done with Apple Intelligence, where tight hardware-software integration creates capabilities that pure software companies struggle to match. Rakuten's 'Rakuten Ecosystem' approach means the AI model can consider a user's mobile data usage patterns, banking transactions, and travel history when making commerce recommendations — a level of personalization that raises both excitement and privacy questions.

Privacy and Data Governance Raise Important Questions

The depth of data integration powering Rakuten's AI platform inevitably raises data privacy concerns, particularly as the company eyes international expansion into markets governed by the EU's GDPR and California's CCPA. Japan's own Act on Protection of Personal Information (APPI) was significantly strengthened in 2022, but it remains less restrictive than European regulations in several key areas.

Rakuten has stated that the foundation model was trained using anonymized and aggregated data, with individual user personalization happening through a separate inference layer that processes data locally where possible. The company claims no raw personal data is embedded in the model's weights.

However, privacy advocates have noted that even anonymized commerce data can be re-identified when combined with location, timing, and behavioral patterns — exactly the type of cross-ecosystem data Rakuten possesses. The company will need to navigate these concerns carefully as it scales internationally, particularly if it targets European or North American markets where regulatory scrutiny of AI systems is intensifying.

Industry Context: The Rise of Vertical AI Models

Rakuten's launch fits into a broader industry shift toward vertical AI models — purpose-built systems trained for specific industries rather than general-purpose reasoning. Bloomberg built BloombergGPT for finance. Harvey AI raised $100 million for legal AI. And now Rakuten is betting that commerce deserves its own foundation model.

The economics support this approach. General-purpose models like GPT-4o and Claude 3.5 Sonnet excel at breadth but often require extensive prompt engineering and fine-tuning to match the performance of domain-specific models on specialized tasks. For a company processing billions of commerce interactions annually, even a small accuracy improvement in product recommendations translates to hundreds of millions of dollars in additional revenue.

Industry analysts at Gartner have projected that by 2027, over 50% of enterprise AI deployments will use domain-specific models rather than general-purpose LLMs. Rakuten's investment appears to be an early and aggressive bet on this trajectory.

What This Means for Developers and Businesses

For the broader tech community, Rakuten's move carries several implications. First, it validates the build-versus-buy approach for companies with sufficient proprietary data. Organizations sitting on large domain-specific datasets may find that investing in custom models delivers better ROI than perpetual API fees to general-purpose model providers.

Second, the launch signals intensifying AI competition in e-commerce. Merchants operating across multiple platforms may soon face pressure to optimize for AI-driven discovery on each platform separately — much as they currently manage SEO for Google and advertising on Meta.

Third, Rakuten's open-source track record — including the release of Rakuten AI 2.0 — suggests that portions of this commerce model technology could eventually become publicly available. This would benefit smaller retailers and developers building commerce applications who lack the resources to train models from scratch.

Looking Ahead: International Expansion and Open Questions

Rakuten has confirmed that international rollout is planned for late 2025 and into 2026, with initial expansion targeting Southeast Asian markets where the company already has a presence through Rakuten Viki and other services. Entry into North American and European commerce markets would represent a more ambitious — and challenging — move.

Several open questions remain. Can the model, trained primarily on Japanese commerce data, transfer effectively to Western consumer behavior and preferences? Will Rakuten open-source any components of the commerce foundation model? And how will regulators respond to an AI system that integrates data across financial services, telecommunications, and retail?

What is clear is that the era of AI-native commerce platforms has arrived. Rakuten's launch puts pressure on every major e-commerce player — from Amazon to Alibaba to Shopify — to articulate their own AI commerce strategies. The companies that move fastest to integrate domain-specific AI into every layer of the shopping experience will likely capture disproportionate market share in the years ahead.

For now, Rakuten's 100 million Japanese users serve as the world's largest test market for AI-driven commerce. The results of this experiment will shape how the entire industry approaches AI integration for years to come.