Xunce's TokenONE: AI Tokenization Could Boost Margins 20-30%

Xunce (3317.HK) launched TokenONE, the world's first TokenOS operating system, on May 25, 2026, marking a pivotal shift in how enterprise data is monetized for artificial intelligence. TokenONE enables companies to package proprietary data into structured "data Tokens" that large language models can invoke directly, bypassing traditional API gateways and data-cleaning pipelines. Xunce is pricing these vertical-domain Tokens at USD10 to USD100 per million Tokens, a range that reflects the varying scarcity and quality of specialized datasets, from financial filings to medical imaging archives. The company projects this Token business will lift its gross margins from the current 10%??0% range to 20%??0% by the end of 2026. That margin expansion is not incremental; it represents a fundamental re-architecture of how data owners capture value in the AI stack. For enterprises sitting on terabytes of underutilized proprietary data, TokenONE offers a direct revenue channel that bypasses the hyperscaler-controlled model training pipeline. This matters now because tokenization is emerging as the next AI infrastructure play, and Xunce is first to market with a dedicated operating system, a land-grab moment that will reshape data economics across industries.

The 10??0 Percentage Point Margin Expansion: Revenue Composition Shift

Xunce's gross margin improvement from 10%??0% to 20%??0% is driven by a structural shift in revenue composition, not cost cutting. Historically, Xunce's core business of selling data-processing middleware and integration tools carried thin margins because it competed on implementation labor and cloud infrastructure resale. TokenONE changes that calculus entirely. The TokenOS abstracts away the expensive human-in-the-loop steps: data cleaning, schema mapping, and API endpoint maintenance. Instead, enterprise data is tokenized once and invoked repeatedly by AI models on a per-token basis. Each invocation generates a micropayment at the USD10??00 per million Tokens rate, and the marginal cost of serving that token is near zero once the tokenization pipeline is built. That creates a software-like margin profile on what was previously a services business. The 10-percentage-point midpoint of the margin lift, from 15% to 25%, implies that TokenONE's gross profit per dollar of revenue will roughly double. For a company of Xunce's scale, that delta translates into tens of millions of dollars in additional gross profit annually, assuming Token adoption reaches even a modest fraction of its enterprise customer base. The key lever is that Token pricing is decoupled from compute cost; a Token's value derives from the uniqueness of the underlying data, not from the GPU cycles required to serve it. Xunce's existing middleware customers, who already trust the company with their data pipelines, provide a natural installed base for the TokenOS. The company expects that converting just 10% of its current enterprise clients to TokenONE will generate enough transaction volume to hit the 20%??0% margin target.

How TokenONE Rewrites the AI Data Supply Chain

The TokenOS creates a new layer in the AI data stack that sits between enterprise data lakes and model inference endpoints. Traditionally, an enterprise wanting to feed proprietary data into an LLM had to build custom connectors, negotiate data licensing terms, and maintain compliance pipelines. TokenONE standardizes this process: data is tokenized into discrete, model-invokable units with built-in access controls and usage tracking. For model providers like OpenAI and Anthropic, this means they can access high-quality, domain-specific data without negotiating individual data-sharing agreements with every enterprise. For enterprises, it means they retain ownership and control while monetizing data that previously sat idle. The pricing tier of USD10??00 per million Tokens is deliberately set below the cost of custom data integration, which can run into the hundreds of thousands of dollars per dataset. This pricing creates a frictionless procurement path: a hedge fund can buy 10 million Tokens of earnings call transcripts for USD1,000 and feed them directly into a fine-tuning pipeline. The TokenOS also includes a "Token Factory," a tool that lets enterprises mint their own Tokens from raw data, effectively turning every company with proprietary data into a potential Token supplier. The minting process is designed to be self-service: an enterprise uploads raw datasets, the TokenOS applies schema normalization and access-control tagging, and the resulting Tokens are listed on Xunce's marketplace within hours. That contrasts with legacy data-licensing workflows, which can take weeks of legal negotiation and weeks more of technical integration. The speed of Token minting is itself a competitive weapon ??it means enterprises can list new datasets immediately after regulatory approval or a business event, capturing the premium pricing that early data availability commands. This supply-side expansion is what drives the margin story: as more enterprises mint Tokens, Xunce collects a platform fee on each transaction without bearing the cost of data creation.

Competitive Reshuffle: Who Gains and Who Loses

Xunce's move directly challenges the data-brokerage and API-management incumbents. Companies like Cisco Systems and Splunk, which sell data integration and observability platforms, face a world where data is tokenized at the source rather than piped through their middleware. If TokenONE gains traction, the need for complex data-pipeline engineering diminishes; enterprises will simply mint Tokens and let AI models pull what they need. That threatens the recurring license revenue of traditional data-platform vendors. On the hyperscaler side, Google and Meta have the most to lose from a decentralized tokenization standard. Both companies have invested heavily in proprietary data moats, Google's search index and Meta's social graph, that they monetize through walled-garden APIs. TokenONE's open tokenization model will commoditize data access, reducing the premium those companies charge for exclusive datasets. Conversely, Alibaba and other cloud providers with large enterprise customer bases are natural distribution partners for TokenONE, with the TokenOS designed to embed directly into their cloud marketplaces. The biggest winners are enterprises with unique, high-value data that is currently under-monetized. Think Ferrari's design files or Uber's mobility patterns. These companies can now become data suppliers without building a separate AI infrastructure team. The losers are the middlemen who previously aggregated and resold data; their margin gets compressed between Xunce's platform fee and the enterprise's willingness to sell Tokens directly.

Downstream Effects on Hyperscalers, Fabs, and Enterprise Buyers

TokenONE's emergence creates second-order effects across the AI hardware and infrastructure ecosystem. If tokenization reduces the cost and friction of accessing high-quality training data, the demand for compute, including GPUs, ASICs, and memory, will shift from data-processing workloads to model-training and inference workloads. That benefits chip designers like Astera Labs (ALAB.O), which makes connectivity chips for AI clusters, and Arm Holdings, whose architecture powers many AI accelerators. Cerebras Systems (CBRS.O), which went public in May 2026 and is up 52% from its IPO price but down 27% from its first intraday high, stands to gain renewed demand as tokenization unlocks new training runs requiring its wafer-scale chips. However, the shift also creates risk for hyperscalers that have built massive data-engineering teams to prepare proprietary data for training. If TokenONE makes data preparation a commodity, those teams become cost centers rather than competitive advantages. Enterprise buyers, including banks, insurers, and healthcare providers, will face a new procurement decision: whether to build their own tokenization infrastructure or buy Tokens from Xunce's marketplace. The pricing at USD10??00 per million Tokens is low enough that most enterprises will choose to buy rather than build, accelerating the platform's adoption. The regulatory angle is decisive: tokenized data with usage tracking creates an audit trail that satisfies emerging AI governance requirements, positioning TokenONE as a compliance tool and a revenue driver simultaneously.

Policy and Strategy Signal: Tokenization as the Next AI Infrastructure Standard

Xunce's TokenONE launch is a strategic bet that the AI industry will standardize around tokenized data exchange rather than API-based data sharing. This mirrors the shift from proprietary file formats to open protocols that defined the internet era. The TokenOS is essentially an operating system for data commerce. It defines how data is packaged, priced, and transferred between parties. If it gains critical mass, it will become the de facto standard for enterprise AI data transactions, much like TCP/IP became the standard for internet traffic. That gives Xunce a trajectory toward platform dominance comparable to what Microsoft achieved with Windows or Google with Android. The timing is deliberate: the AI IPO boom is attracting high valuations but many recent hot IPOs underperform the S&P 500, as Reuters reported. SpaceX (SPCX.O) is expected to debut at a $1.75 trillion valuation as early as June 11, followed by OpenAI and Anthropic. In this frothy environment, Xunce is positioning TokenONE as the infrastructure layer that all these AI companies will need to access enterprise data. The company is effectively saying that the real value in AI is not in the models themselves, which can be fine-tuned or bypassed, as Stanford's Trustworthy AI Research Lab showed, but in the proprietary data that makes those models useful. TokenONE is a bet that data, not compute, will be the scarce resource in the next phase of AI, and that Xunce will own the marketplace where that data is traded.

The success of TokenONE hinges on whether enterprises trust Xunce enough to tokenize their most valuable data assets; that trust must hold as the platform scales to millions of daily Token invocations across sectors. The company must prove that its TokenOS is secure, that access controls cannot be bypassed, and that the pricing model remains stable as adoption scales. If it delivers, Xunce transforms from a middling middleware vendor into the operator of a data exchange that will rival the hyperscalers in strategic importance. The margin expansion from 10%??0% to 20%??0% is just the first signal; if TokenONE captures even 5% of the enterprise AI data market, the revenue implications dwarf anything in Xunce's current business. The next 12 months will determine whether TokenONE becomes the standard or a footnote, and whether Xunce's bet on tokenization pays off as the defining infrastructure play of the AI era.