Xunce Technology: Data Services Transition to Token-Based Billing, Growth Accelerates

Overview

This is a conference minutes note published by Goldman Sachs following the 2026 Asia Communications & Technology Conference, centered on management commentary from Shenzhen Xunce Technology (3317.HK, rated NC—Not Covered). The report focuses on Xunce’s strategic positioning within the AI value chain, its business model transformation from traditional project-based engagements to token-usage-based pricing, and its differentiated competitive advantages versus vertical model vendors and cloud service providers. Overall, the report believes the company has captured the wave of surging demand for high-quality data driven by large model adoption, and the rapid scaling of its token-based model could deliver dual benefits of accelerated revenue growth and margin improvement.

Core views

Strategic Positioning: Management positions the company as an incremental infrastructure layer in the AI value chain—connecting upstream GPU and cloud providers with downstream large model developers. Xunce’s core capability lies in transforming enterprises’ raw data into structured tokens ready for direct consumption by large models, thereby enhancing model output efficiency and accuracy. The report notes management’s observation that as large models increasingly become productivity tools, market demand for high-quality, standardized data has significantly intensified. Business Model Upgrade: The company offers end-to-end services—from data collection and cleansing to real-time computation and large model fine-tuning—with the ultimate goal of enabling measurable business decision improvements for clients. This outcome-driven service framework underpins the critical shift in pricing from traditional project-based fees to token-usage-based billing. The report cites impressive data: annual recurring revenue from token-based billing (by-token ARR) has rapidly climbed from RMB 60 million in January to RMB 200 million in April, with management expecting this momentum to continue through the year. Differentiated Moat: The report highlights Xunce’s unique advantage in directly integrating with clients’ private clouds and accumulating deep data governance expertise across specific industry datasets. This addresses the 'last-mile' bottleneck encountered when feeding raw data into general-purpose large models—direct ingestion not only consumes excessive compute resources but also risks model hallucinations. Additionally, the company’s KPI-driven delivery approach fosters trust with new clients and creates high post-deployment stickiness.

Analysis framework

Goldman Sachs employs a classic analytical framework: 'Strategic Positioning → Business Model → Competitive Moat.' It begins by interpreting management’s articulation of the company’s role in the AI value chain, clarifying its intent to serve as a 'data infrastructure layer.' It then zooms in on the most informative aspect—the business model shift from projects to token-based pricing—and uses concrete, dynamic ARR metrics (RMB 60M in Jan → RMB 200M in Apr) as core evidence that the transition has entered a phase of tangible commercial scale-up. Finally, it examines competitive dynamics, demonstrating how Xunce differentiates itself from vertical model players and cloud providers, thus forming a complete logical loop. Methodologically, this note reflects a standard industry research approach to emerging SaaS/platform companies: tracking the evolution of pricing models (from one-time project revenue to usage-driven recurring revenue), as such shifts often signal a step-change in revenue quality and valuation logic. The report identifies the explosive short-term growth in by-token ARR as a leading indicator of enhanced growth potential and profitability quality.

Methodology notes

• Industry/Value Chain Analysis FrameworkUpstream-Midstream-Downstream Value Chain Transmission

  Value allocation and strategic positioning logic across the AI value chain (upstream hardware/compute → midstream data & model services → downstream applications)

  The report situates Xunce Technology in the midstream 'data infrastructure layer' of the AI value chain. Its value isn’t about replacing upstream GPUs or downstream models, but about building a data-processing bridge between compute and large models. In simple terms: large models need clean, structured data to run efficiently, yet enterprise raw data is often chaotic—Xunce solves this intermediate processing challenge. This framework explains why the company holds independent commercial value in the AI wave.

• Company Fundamentals & Financial FrameworkProfitability Quality Analysis

  ARR (Annual Recurring Revenue) as a key metric for assessing revenue quality and growth trajectory of SaaS/platform companies

  The report closely tracks the rapid ramp-up of by-token ARR (from RMB 60M in January to RMB 200M in April)—a classic SaaS analysis paradigm. Shifting from one-time project recognition to usage-based recurring revenue significantly enhances revenue predictability and quality. Moreover, the report explicitly links this model upgrade to future gross margin expansion, signaling structural improvements in profitability quality.

• Competition & Strategy FrameworkMoat / competitive advantage

  Industry-specific data governance know-how from private cloud integration and KPI-driven delivery create customer stickiness and form a competitive moat

  By contrasting Xunce with vertical model vendors and generic cloud providers, the report pinpoints its differentiation: (1) direct private cloud access enables deep, industry-specific data governance expertise, creating an experience-based barrier; (2) delivering against clear client business KPIs builds trust and high switching costs post-deployment. This reflects two classic moat dimensions: 'switching costs' and 'intangible assets (data know-how).'

• Company Fundamentals & Financial FrameworkVolume-price decomposition

  Revenue drivers under token-usage pricing (volume growth × price effect)

  Although the report doesn’t provide detailed volume-price modeling, its core logic implies such decomposition: the explosion of large model applications drives significant 'volume' growth in enterprise data processing needs, and standardized tokens as the billing unit directly tie revenue to usage—thus enabling revenue acceleration. The highlighted ARR surge is essentially a direct manifestation of 'volume-driven revenue growth.'

Key data

• Token-based Annual Recurring Revenue (by-token ARR)Increased from RMB 60 million in January 2026 to RMB 200 million in April 2026More than tripled in a few months; management expects this growth momentum to continue through the year.

Impact & implications

The report argues that the rapid scaling of token-based pricing delivers dual positive impacts for Xunce Technology. On the revenue side, the shift from project-based to usage-driven billing strengthens growth momentum and improves predictability—already preliminarily validated by ARR data. On the profit side, the standardized product nature of the offering is expected to structurally expand gross margins. From a competitive standpoint, this model upgrade further solidifies the company’s differentiated positioning as an AI data infrastructure layer, making its role in the AI ecosystem more independent and irreplaceable.

What to watch

• Monthly or quarterly trends in token-based ARR growth, monitoring whether the current high-growth trajectory can be sustained
• Actual margin improvement magnitude as the by-token model scales broadly
• Continued growth in data demand from downstream large model clients and the pace of new client acquisition