Liu Liehong Technology News: Inner Mongolia Faces the Hard Part of Turning Computing Power Into a Data Market

Liu Liehong inspected nine Inner Mongolia facilities over two days, but the central conflict was larger than another government tour of data centers. This technology news points to a difficult transition in China’s digital strategy: turning abundant computing infrastructure into data products that companies can legally trust, exchange, and use.

China’s National Data Administration said Liu, its director, visited the autonomous region on September 1 and 2, 2026. The agency published its account on September 5. His itinerary covered cloud campuses, a computing-resource scheduling platform, an industrial park, a dairy research center, and a vocational college.

The mix matters. Inner Mongolia already occupies a strategic position in China’s East Data, West Computing program, which directs suitable computing workloads toward western regions with available land and energy. The new pressure is to prove that infrastructure can support an active market, not merely host more servers.

That challenge pits infrastructure capacity against market utilization. A region can install processors, connect data centers, and procure renewable electricity without creating valuable datasets or dependable exchange mechanisms. The next phase requires enforceable rights, common standards, credible security controls, and buyers with specific business problems.

What Liu Liehong’s Inner Mongolia Visit Actually Changed

The visit connected Inner Mongolia’s computing buildout to a broader test of China’s market-oriented data reforms.

According to the National Data Administration’s official account, Liu’s delegation visited China Unicom’s Hohhot cloud data center and China Mobile’s Hohhot data center. It also inspected facilities operated by Huawei Cloud, UCloud, and ChinData.

The delegation examined the Horinger cluster’s multi-cloud computing monitoring and scheduling platform. Such a platform tracks available resources across providers and helps direct workloads toward suitable capacity. It addresses fragmentation at the infrastructure layer, where isolated facilities can make regional capacity harder to coordinate.

The itinerary extended beyond cloud operators. Liu also visited the Inner Mongolia Data Element Industrial Park, the National Dairy Technology Innovation Center, and a vocational institute focused on digital intelligence. These stops joined three parts of the commercialization chain: infrastructure, sector-specific data applications, and workforce development.

The official statement emphasized Inner Mongolia’s policy, location, and energy advantages. It described the region as an important hub in the East Data, West Computing project and highlighted work on an integrated national computing network.

However, the statement announced no new regulation, investment package, procurement award, or implementation deadline. It also disclosed no utilization rate, transaction value, customer count, or financial return from the inspected projects.

That absence defines the story. The tour was a policy signal and an operational review, not evidence that a mature regional data market already exists. It showed which pieces officials want connected and which results remain unreported.

One unfamiliar term in the statement was “token factory,” a model that uses computing resources and prepared data to produce tokens for artificial intelligence workloads. Tokens are the units that AI models process when interpreting or generating content.

This concept moves the policy discussion beyond renting server capacity. It treats energy, processors, datasets, and model workloads as a production system. Inner Mongolia would supply more than remote computing if that model works. It would participate in preparing and processing the data that supports commercial AI.

The dairy research stop offered a practical illustration. Agricultural and dairy operations generate information about breeding, feed, animal health, production, logistics, and quality control. Combining those records can support forecasting, research, and operational decisions.

Yet combining them requires more than storage. Participants need authority to use each dataset, controls governing access, and rules for distributing the resulting value. Those institutional questions are the real subject behind the infrastructure tour.

The vocational college visit reinforced another constraint. Data centers need technicians, but a functioning data economy also needs specialists in data governance, security, labeling, compliance, product design, and industry operations. Hardware deployment can move faster than this workforce develops.

Liu’s trip therefore changed the emphasis surrounding Inner Mongolia’s role. The region is still a computing hub, but officials are increasingly evaluating whether its capacity supports data circulation, industry applications, and AI production. That is a more demanding standard than counting installed machines.

Why This Technology News Is About Utilization, Not Server Capacity

Inner Mongolia’s advantage begins with energy and infrastructure, while its next test is whether organizations consistently buy and use the resulting services.

China created the East Data, West Computing framework to coordinate computing resources across regions. Many latency-tolerant jobs, including offline analysis, model training, backup, and rendering, can run far from coastal commercial centers.

Inner Mongolia has natural attractions for this strategy. It offers space for large campuses and access to substantial energy resources. A cooler climate can also reduce part of a data center’s cooling burden, although facility efficiency depends on design and operating conditions.

These benefits explain why major telecom and cloud companies operate in the region. They do not automatically establish a market for data itself. Computing capacity supplies processing, while a data market needs reusable resources, defined permissions, discoverable products, and repeat customers.

China’s broader numbers show why officials are focused on that distinction. The country generated 52.26 zettabytes of data in 2025, representing a reported 27.44 percent of global output. The total increased 27.28 percent from the previous year, according to an official data survey.

Producing information does not mean organizations can exchange it. Much of it remains distributed across government bodies, industrial systems, consumer platforms, research institutions, and private companies. Legal restrictions, incompatible formats, security concerns, and unclear economic incentives limit reuse.

A data element market attempts to address those barriers by treating usable data as an economic input. The phrase does not mean selling every raw record. It covers authorized access, processed datasets, analytical outputs, application interfaces, and services that let participants generate value without surrendering unrestricted control.

The distinction is important for AI. Models benefit from large amounts of relevant, well-governed data, not simply from the total volume stored in a region. Duplicate, poorly labeled, legally uncertain, or inaccessible records contribute little to a production workload.

High-quality datasets demand sustained work. Organizations must remove errors, document origins, manage permissions, standardize fields, and update records. They must also determine whether information can be used for training, inference, evaluation, or only a narrowly defined transaction.

Computing utilization creates a related problem. A data center can report installed capacity while processors remain underused. Buyers evaluate workload performance, network latency, reliability, software compatibility, security, and total operating cost. Location becomes only one part of that decision.

The Horinger scheduling platform addresses this problem at the resource layer. A shared view of capacity can help match workloads with available providers. However, monitoring infrastructure cannot create demand by itself. Customers still need applications that justify moving work to the cluster.

Inner Mongolia’s industrial economy supplies potential applications. Energy, mining, agriculture, dairy production, logistics, and environmental management all generate operational data. These sectors can support forecasting, equipment maintenance, safety monitoring, resource scheduling, and scientific research.

The strongest opportunities will involve specific exchanges rather than abstract declarations that data has value. An energy operator might share controlled information with a forecasting service. A dairy research organization might combine health and production records under agreed rules. A manufacturer might let suppliers access selected quality data.

Each example requires measurable gains. Participants will want lower downtime, improved yields, faster research, reduced risk, or another defined outcome. If projects cannot demonstrate such results, they risk remaining subsidized showcases.

This is why Liu’s technology news deserves attention beyond China policy circles. Regions worldwide are competing to host AI and cloud infrastructure. Many face the same question after construction: how does physical capacity become recurring economic activity?

Infrastructure Capacity Versus a Working Data Market

China’s central policy tension is whether government-led infrastructure can produce decentralized, repeatable transactions among organizations that remain cautious about sharing information.

China has spent several years building the institutional framework for data circulation. A foundational national policy established four broad areas: data rights, circulation and transactions, income distribution, and security governance.

The framework recognizes a basic economic problem. Data can be copied, combined, and reused, so conventional ownership concepts do not always fit. Multiple parties can contribute to a dataset or product, while personal information and trade secrets impose separate limits.

The government’s response has emphasized structured usage rights and controlled circulation. Under this approach, an organization does not always transfer a dataset outright. It can authorize a defined use within a protected environment and retain restrictions on copying or redistribution.

Trusted data spaces are one mechanism for implementing that idea. A trusted data space is a governed environment where participants share or use data under common identity, access, security, and accountability rules.

China launched its first national pilots in 2025. By August of that year, the projects were exploring more than 900 use cases across 32 major industry categories. They had produced over 570 high-quality datasets in 14 fields and involved nearly 70,000 market entities, according to the National Data Administration’s pilot assessment.

Those figures suggest significant participation, but they do not settle the commercial question. A pilot use case can demonstrate technical feasibility without showing durable demand. A participating organization can join an initiative without becoming a recurring buyer.

The administration acknowledged that trusted data spaces remained at an early stage. It identified sustainable operating models, resource supply, and interoperability as continuing challenges. That caution is more informative than headline project counts.

Interoperability presents a particularly difficult issue. A regional platform becomes less useful if its identities, contracts, interfaces, and security controls do not work with systems elsewhere. Companies operating nationally need predictable rules across provinces and sectors.

Fragmented systems can also raise costs. A business may need separate integrations, reviews, and agreements for each data space. Smaller companies are especially sensitive to that burden because they have fewer compliance and engineering resources.

Government-backed exchanges face an incentive problem as well. Suppliers often hold the most commercially valuable information, but sharing it can expose competitive knowledge or legal risk. Buyers may not know whether an unfamiliar product is accurate enough for an important decision.

Pricing adds another layer. The value of data depends on context, quality, freshness, exclusivity, and the buyer’s intended use. One fixed valuation method cannot easily capture these differences.

China’s policy framework therefore combines markets with substantial public coordination. Officials want private and public participants to discover useful transactions, but they are also building common infrastructure, rules, registries, and pilots.

This hybrid model can lower coordination barriers. A trusted public framework can give cautious institutions a starting point for cooperation. It can also prevent every participant from building incompatible systems independently.

However, public direction can distort incentives if project completion becomes more important than actual use. Platforms may optimize for registered products, signed agreements, or demonstration cases because those indicators are easier to report than customer value.

Inner Mongolia sits directly inside this tension. Its computing resources make it a logical test site, and its industries provide distinctive data. The question is whether these advantages produce applications that survive after pilot support and official attention move elsewhere.

The region’s policy documents call for data applications in energy, mining, transport, tourism, agriculture, and livestock operations. They also support trusted data spaces, privacy-preserving computation, public-data authorization, and improved computing coordination.

Privacy-preserving computation refers to techniques that let parties calculate with protected information while limiting exposure of the underlying records. These techniques can reduce risk, but they cannot resolve unclear rights, poor quality, or weak demand.

A working market needs all these elements to reinforce one another. Infrastructure reduces processing constraints. Governance establishes permission. Standards reduce integration costs. Industry applications create demand. Revenue then gives suppliers a reason to maintain quality.

If one component remains weak, the system stalls. Inner Mongolia can offer inexpensive capacity, yet customers will not use unsuitable data. It can create data products, yet buyers will hesitate if legal responsibility remains uncertain. It can establish rules, yet activity will fade without commercially useful applications.

AI Raises the Value of Data and the Cost of Getting Governance Wrong

AI gives Inner Mongolia a clearer demand engine, but it also magnifies quality, security, accountability, and energy risks.

The National Data Administration explicitly linked Inner Mongolia’s data resources and computing foundation to AI innovation. The token-factory concept captures this connection by framing processed AI work as an output of coordinated energy, computing, and data inputs.

For developers, the attraction is straightforward. Training and operating AI models requires processors, storage, networking, and extensive data preparation. A region that coordinates these services can reduce operational friction for suitable workloads.

Enterprises have a different concern. They hold valuable internal data, but much of it includes trade secrets, customer information, employee records, or regulated material. They need confidence that participation will not create uncontrolled access.

Security rules therefore shape market supply. In 2025, six Chinese government bodies issued an implementation plan covering enterprise, public, and personal data circulation. It also addressed responsibility, security services, technical controls, and misuse risks throughout the data lifecycle.

The policy called for security optimization with minimized cost. That objective captures a real tradeoff. Controls that are too weak discourage suppliers and create legal exposure. Controls that are too expensive make smaller transactions uneconomic.

AI complicates the balance because models can retain patterns from training material. A dataset may be protected during transfer but still create risk through model outputs, logs, or derived artifacts. Governance must cover more than the original file.

Data provenance becomes essential. Provenance documents where information came from, how it changed, and which permissions govern its use. Without it, developers may struggle to remove problematic records or explain a model’s behavior.

Quality is equally important. A large dataset can contain outdated measurements, inconsistent labels, sampling bias, or missing context. Processing more low-quality information does not reliably improve a system.

Sector-specific expertise can reduce these problems. The National Dairy Technology Innovation Center, for example, can provide scientific and operational knowledge that a general-purpose exchange lacks. Domain specialists understand which measurements are meaningful and where errors enter the process.

That creates a plausible role for Inner Mongolia beyond hosting processors. The region can build governed datasets around sectors where local institutions already possess expertise. These products could support specialized models, research tools, and operational analytics.

Still, official reporting from Liu’s visit did not identify a completed commercial data product or disclose a customer outcome. It described directions and facilities rather than verified performance. Readers should treat the trip as evidence of policy focus, not proof of business success.

Energy claims deserve similar scrutiny. Inner Mongolia can connect computing growth with abundant renewable resources, and the itinerary included a zero-carbon computing base. Yet a facility’s label does not disclose its complete emissions profile.

Analysts need consistent information about electricity consumption, renewable matching, utilization, cooling, backup power, and embodied emissions. Otherwise, growing AI workloads can increase total energy demand even when individual facilities improve efficiency.

The geographic separation between western computing centers and eastern users creates technical constraints too. Network latency matters for interactive applications, while data-transfer costs can affect very large workloads. Model training and batch processing fit remote capacity better than every real-time service.

Providers must therefore match workloads carefully. A coordinated scheduling system can help, but it needs transparent performance data and dependable software support. Customers will judge the service through results, not regional capacity targets.

Foreign companies face another set of considerations. China’s domestic data-market reforms do not automatically simplify cross-border transfers. Multinational businesses must still evaluate applicable security, privacy, localization, and sector-specific requirements.

The practical lesson is that AI demand strengthens the case for a regional data economy while raising the standard for execution. More processors increase the value of accessible datasets. They also increase the consequences of unreliable governance.

The Evidence Gap Is the Biggest Risk in the Liu Liehong Technology News

The central uncertainty is not whether Inner Mongolia can build infrastructure, but whether officials will publish evidence of repeatable demand and measurable outcomes.

The National Data Administration’s announcement offered a detailed itinerary and a clear policy direction. It did not provide performance indicators for the sites Liu inspected.

There was no disclosed computing utilization rate. The announcement did not quantify how much work the multi-cloud platform had scheduled, how many organizations used it, or whether coordination reduced costs.

It also gave no data-market revenue for the region. Readers cannot determine how many listed data products generated repeat purchases or how much activity occurred without government-directed pilots.

National figures provide context but not a regional answer. China’s data-market transaction volume was estimated to have exceeded 160 billion yuan in 2024, increasing more than 30 percent year over year. That estimate accompanied the launch of a national public-data platform.

Transaction volume can measure several types of activity, depending on the reporting framework. It may include exchange transactions, services, processing, or other data-related business. Without consistent definitions, comparisons across regions remain difficult.

Pilot statistics create a similar issue. Counts of datasets, participants, or use cases show activity, but they do not reveal profitability, renewal, accuracy, or adoption depth. A market becomes credible when customers keep paying because a product solves a problem.

Independent evaluation is therefore essential. Universities, auditors, industry groups, and customers can test whether reported outcomes reflect operational improvements. Their findings would strengthen the evidence beyond government and operator statements.

The National Data Administration itself has recognized the early state of market development. Its work plan focused on obstacles summarized as supplying data, enabling circulation, supporting effective use, and maintaining security.

Those four problems remain tightly linked. Suppliers hesitate when rules are unclear. Buyers hesitate when quality is uncertain. Platforms struggle when both sides wait for the other to participate. Security failures can damage trust across the entire system.

There is also a risk of infrastructure duplication. Provincial and municipal governments may build overlapping platforms to satisfy local development goals. If standards diverge, participants face higher integration costs and smaller pools of buyers.

National coordination can reduce that danger. China had reportedly connected data-infrastructure nodes across multiple cities and provinces by 2025, while its trusted-data-space pilots covered city, industry, and enterprise projects. The long-term goal is an integrated national market.

Yet technical connection does not guarantee institutional compatibility. Participants still need aligned identities, usage rules, dispute procedures, and responsibility for failures. These questions become harder when data crosses industry or regional boundaries.

Small businesses present another test. Large telecom operators, cloud companies, and state-backed institutions can absorb compliance and integration costs. Smaller firms need standardized contracts, accessible tools, and clear value before they will participate at scale.

The government’s 2025 data-space pilot notice explicitly called for convenient products serving smaller companies. Evidence that these companies have become recurring suppliers or buyers would be an important measure of progress.

Talent is a related risk. Inner Mongolia can train technicians and data specialists, but experienced governance and industry professionals remain difficult to develop quickly. The vocational institute on Liu’s itinerary shows that officials recognize this constraint.

Market incentives will determine whether trained workers stay. Sustainable private demand, career development, and competitive projects matter alongside education capacity. Otherwise, regional training programs may supply talent to larger coastal markets.

The evidence gap does not make the strategy empty. It defines the threshold that future announcements must cross. Specific utilization, revenue, renewal, application, and efficiency results would turn policy intent into a testable economic claim.

What Businesses and Developers Should Watch Next

Three signals will show whether Inner Mongolia is building a functioning data economy or extending an infrastructure program.

The first signal is transparent utilization and customer data from the Horinger cluster. Relevant disclosures include scheduled workloads, processor utilization, repeat customers, service reliability, and the share of demand originating outside the region.

Rising repeat use would strengthen the case that Inner Mongolia’s capacity has found a market. Large installed capacity with limited sustained demand would weaken it. Aggregate figures should also separate general cloud workloads from AI training and inference.

The second signal is evidence from sector-specific data products. Energy, mining, dairy production, agriculture, and logistics offer logical starting points because local organizations already generate relevant operational information.

The strongest proof would include a defined dataset, participating suppliers, usage permissions, paying customers, and a measurable operational result. A long catalog of products without recurring transactions would provide weaker evidence.

China’s wider trusted-data-space program offers a benchmark. National pilots are already exploring hundreds of scenarios, but the official review identified operating sustainability and interoperability as unresolved challenges. Inner Mongolia needs to show progress on those exact issues.

The third signal is greater standardization across regional and industry platforms. Businesses should watch national rules for identity, contracts, data valuation, security responsibility, and interoperability.

A 2026 standards project covering data-value evaluation illustrates the ongoing effort to create shared methods. Standardization can lower transaction costs, but overly rigid valuation models risk ignoring context. The useful outcome is comparable documentation and process, not a universal price for every dataset.

Companies considering participation should ask practical questions before committing information or workloads. Who can access the data, and for which purpose? Can permission be withdrawn? How are derived products governed? Who bears responsibility after misuse or leakage?

Developers should examine provenance and licensing before training a model. They also need to understand whether access covers development, evaluation, production inference, and model improvement. A general right to analyze data may not authorize every AI use.

Infrastructure buyers should request workload-specific performance and energy information. Installed processor counts say little about real throughput if networking, storage, orchestration, or software availability becomes the bottleneck.

Knowledge workers also have a stake in the policy direction. Better authorized access can improve research and organizational decision-making. Poorly defined systems can instead create new silos behind complicated interfaces.

Teams evaluating large collections of internal documents can begin with disciplined knowledge management. Clear provenance, permissions, and context matter inside one company for the same reasons they matter across a regional data market.

Liu Liehong’s visit placed Inner Mongolia at a revealing point in China’s data strategy. The region has major infrastructure, policy support, industry data, and national attention. Those advantages make it a credible test environment, but they do not predetermine commercial success.

The next meaningful technology news will not be another list of facilities. It will be evidence that organizations exchange governed data, run sustained workloads, and obtain results worth repeating.

Watch the utilization figures, sector-level customer outcomes, and interoperability rules. If those signals improve together, Inner Mongolia will be converting energy and computing capacity into an active data economy. If they remain undisclosed, the gap between infrastructure and market demand will remain the defining story.