Entive Smart Home Plans a RMB 2 Billion AI Server Bet, but the Financing Risk Is Just as Large

Entive Smart Home plans to authorize up to RMB 2 billion for servers and related equipment, despite its early-stage position in AI computing services. The proposed purchase is unusually large for a company whose main business remains integrated stoves and other kitchen appliances.

Its wholly owned subsidiary, Gansu Yisuan Intelligent Technology, intends to buy the equipment from several suppliers. It would then use those assets to provide computing capacity to customers. The transaction does not qualify as a major asset restructuring or a related-party deal under the rules cited by Entive.

The central issue is not whether demand for AI infrastructure exists. It is whether a consumer appliance company can finance, deploy, sell, and maintain expensive computing assets before borrowing costs and depreciation overwhelm the resulting service revenue.

That tension became sharper because this is not a distant strategic concept. Entive had already approved a smaller server budget earlier in July, signed purchase contracts, and begun receiving equipment. The new proposal expands the ceiling before investors have seen clear evidence that the initial deployment can earn an acceptable return.

The Server Plan Expanded From RMB 550 Million to RMB 2 Billion

Entive is moving from an initial server deployment toward a capital commitment that requires shareholder approval and much closer scrutiny.

The board approved the new proposal on July 24, 2026, according to the company’s purchase announcement dated July 27. Gansu Yisuan would be allowed to sign server and supporting-equipment contracts with a combined value of no more than RMB 2 billion.

That amount represents a ceiling, not proof that every purchase has already occurred. Final payments, delivery schedules, and equipment configurations depend on the individual agreements signed with suppliers.

Entive has not publicly identified those suppliers. The company said disclosure could expose commercial secrets, cause contractual problems, or create unfair competition. It therefore completed an internal exemption process and referred to the vendors collectively rather than naming them.

The undisclosed specifications matter. AI server economics vary widely according to accelerator type, memory capacity, networking, cooling, software compatibility, and expected utilization. A purchase amount alone cannot show how much usable capacity Gansu Yisuan will receive.

The company describes the assets as new servers and related equipment intended for its computing business. Their prices will reference market conditions and be settled through negotiations with each vendor.

The proposal must still pass a shareholder vote. Entive said the amount exceeds 50 percent of its latest audited net assets. Purchases reviewed during the preceding 12 months also surpass 30 percent of its latest audited total assets.

Those thresholds explain why the latest authorization cannot stop at board approval. It requires support from at least two-thirds of the voting rights represented at the shareholder meeting.

The proposal follows a much smaller authorization approved on July 7. That earlier decision allowed Gansu Yisuan to purchase up to RMB 550 million of servers and related equipment.

A subsequent purchase update said cumulative contracts had reached RMB 529 million. Part of the equipment had arrived, while the remaining units had not yet been fully delivered.

One disclosed contract in that first round carried a value of RMB 280.602 million. Entive also said future purchasing would proceed according to customer-order conditions.

The new RMB 2 billion ceiling includes the earlier RMB 550 million authorization. It should not be read as RMB 2 billion added on top of an entirely separate RMB 550 million commitment.

Still, the increase is substantial. The maximum budget is approximately 3.6 times the earlier ceiling and almost 3.8 times the contracts already disclosed under that program.

Entive says the larger pool will help it build its own AI computing resources and establish a stable service platform. Customers would rent access to that capacity instead of buying and operating the physical infrastructure themselves.

That model converts servers into revenue-producing fixed assets. It also transfers purchasing, financing, installation, maintenance, and utilization risk to the service provider.

The announcement does not constitute a completed customer-order disclosure. It authorizes asset purchases and explains their intended use, but it does not identify the customers supporting the full RMB 2 billion capacity plan.

That distinction creates the article’s main tension. Entive is scaling the asset side of the business faster than public disclosures currently establish the corresponding revenue side.

Why Entive Is Building AI Capacity in Gansu

Entive’s computing strategy relies on turning western China’s data-center resources into rentable capacity for AI inference and other demanding workloads.

Gansu Yisuan sits at the center of that plan. Entive holds the subsidiary outright, which gives the listed company direct control over its purchasing and service operations.

The subsidiary has been developing computing projects in Qingyang, Gansu province. The region is part of China’s broader effort to locate data-intensive infrastructure where land, energy, and large-scale facilities are more available.

Entive’s 2025 annual report said its Qingyang computing center entered operation in December 2024. The first phase offered 2,500 petaflops of green computing capacity, according to the company.

A petaflop represents one quadrillion floating-point calculations per second. Actual AI performance still depends on hardware architecture, numerical precision, software efficiency, networking, and workload design.

The company says the center has supported inference work for customers including Meitu and Infinigence AI. Inference is the process of running a trained AI model to produce outputs, such as generated images or model responses.

That workload differs from model training. Training often demands large clusters for concentrated periods, while inference demand grows with continuous product use and the number of user requests.

Entive also reported a five-year computing-service relationship with Infinigence AI. A 2026 bond-rating report said that customer represented more than 80 percent of the computing unit’s sales and carried total service fees of RMB 106 million.

This existing contract gives the strategy more substance than a purely speculative market entry. It shows that Gansu Yisuan has deployed equipment, connected at least one major customer, and established a multiyear service arrangement.

However, the same figure reveals concentration risk. A computing operation dependent on one customer can suffer if usage falls, payments slow, technical requirements change, or the customer moves workloads elsewhere.

Entive has outlined a much larger regional ambition. In April 2025, Gansu Yisuan signed a framework agreement with Enflame’s Qingyang unit and the Qingyang municipal government.

The parties described a domestic computing cluster with at least 100,000 accelerator cards and 25,000 petaflops of capacity. Entive’s annual report associated that broader plan with expected investment of RMB 5.5 billion through 2028.

A framework agreement does not guarantee that every phase will be financed, built, or occupied. It establishes a direction for cooperation, while later contracts and funding decisions determine the project’s actual scale.

The RMB 2 billion server proposal fits inside this longer expansion path. It provides a potential mechanism for adding physical capacity more quickly than building the whole cluster at once.

Timing also favors inference-focused projects. Chinese companies are deploying generative AI across image editing, enterprise software, customer service, industrial systems, and autonomous machines.

Each active application consumes computing resources after deployment. That creates recurring demand, provided customers keep using the application and the service provider maintains competitive performance.

Industry investment has therefore shifted beyond the largest cloud companies. Telecom operators, specialized data-center businesses, hardware distributors, and companies from unrelated sectors are all seeking positions in computing services.

A June report from Xinhua Finance documented several large server, financing, and computing-service agreements among Chinese listed companies. Some entrants came from businesses far removed from traditional cloud infrastructure.

Entive is part of that cross-industry movement. Its difference is especially visible because consumers know it, when they know it at all, as a kitchen appliance manufacturer.

The company’s Gansu location and existing customer contracts offer a plausible operating foundation. They do not erase the difficulty of scaling from a limited commercial deployment into a much larger infrastructure provider.

The Core Tradeoff Is Capacity Growth Versus Balance-Sheet Pressure

The same financing that lets Entive accelerate its computing business can increase debt, interest expense, and depreciation before the new servers produce steady revenue.

Entive plans to fund the purchases with internal resources and externally raised money, including finance leases. A finance lease lets a company obtain equipment while paying for it over time, although the arrangement still creates financial obligations.

This structure can reduce the immediate cash required at delivery. It does not remove the economic cost of the servers or the need for sufficient customer income to cover scheduled payments.

Entive explicitly warned that the purchases would probably raise its debt-to-asset ratio. It also expects financial expenses to increase considerably, with potential effects on current and future results.

That warning deserves more attention than the headline purchase figure. An authorized investment can appear growth-oriented, while the related interest and lease expenses begin affecting earnings well before utilization reaches an efficient level.

The starting financial position offers little room for casual optimism. Entive reported 2025 revenue of RMB 337.7341 million, down 51.94 percent from the preceding year.

Net loss attributable to shareholders reached RMB 171.8822 million. A year earlier, the company had recorded an attributable profit of RMB 26.5414 million.

The maximum server budget is therefore almost six times Entive’s entire 2025 revenue. It is also more than 11 times the attributable loss reported for that year.

Those comparisons do not prove that the proposal is unaffordable. Purchases may occur in stages, finance leases can spread payments, and computing contracts can generate multiyear revenue.

They do show that the investment is transformative in financial scale. It cannot be treated as an ordinary equipment refresh inside an established technology business.

Entive attributed its 2025 revenue decline partly to slower demand in the property-linked kitchen appliance market. It also said its computing operation remained at an early stage and had not achieved sufficient scale benefits.

The old business and the new business therefore present different challenges. Kitchen appliances face weak orders, while computing services require heavy upfront investment and technical execution.

Entive is trying to use the second operation as a new growth engine before the first has recovered. That can diversify revenue, but it also places financing demands on a group already reporting losses.

Server depreciation adds another layer of pressure. Companies allocate the cost of equipment across its useful accounting life, reducing reported profit during that period.

Economic obsolescence can move faster than accounting schedules. New accelerators, memory systems, networking standards, and software stacks can reduce the relative value of older machines.

A server remains useful only if customers can run the workloads they want at a competitive total cost. Ownership alone does not create an advantage.

Utilization is therefore the key operating variable. A heavily used server can distribute financing, power, maintenance, and depreciation costs across more billable work.

An underused server continues generating many of those costs without matching revenue. The difference between those outcomes can determine whether a computing provider reports healthy margins or recurring losses.

Entive’s customer-order language suggests it recognizes that issue. The company previously said additional purchases would follow order conditions, rather than proceeding independently of demand.

Yet the latest announcement does not disclose contracted utilization for the expanded capacity. It also does not state whether customers have provided deposits, minimum-use commitments, or credit support.

Those commercial terms would help investors judge the balance between secured demand and speculative capacity. Their absence does not mean the orders do not exist, but it limits outside verification.

Contract performance creates another risk. Entive noted that changes in law, policy, technology, markets, macroeconomic conditions, or unexpected events could prevent complete execution.

Funding itself remains conditional. If Entive cannot raise enough money on time, purchase payments could fail and transactions could collapse.

This produces a clear tradeoff. Moving early can secure equipment and customers during a period of strong demand, while moving too far ahead can burden the company with debt and idle assets.

A Kitchen Appliance Maker Still Has to Prove It Can Operate an AI Cloud

Entive has established a computing foothold, but the proposed scale requires capabilities that extend far beyond buying servers and placing them in a data center.

Computing services depend on reliable power, cooling, networking, cluster scheduling, security, maintenance, and customer support. Failures in any part of that chain can make nominal capacity unavailable.

Customers also evaluate software compatibility. AI teams need supported frameworks, drivers, orchestration tools, monitoring, storage, and methods for moving data into and out of the cluster.

Large cloud providers combine infrastructure with mature software platforms and extensive engineering teams. Telecom operators add network resources, regional facilities, and established enterprise relationships.

Specialized computing providers can compete through focused hardware, local supply, lower operating costs, or closer customization. However, they still need enough scale and expertise to maintain service levels.

Entive enters this field without the operating history of Alibaba Cloud, Tencent Cloud, Huawei Cloud, or China’s major telecom carriers. Its established manufacturing and distribution experience does not automatically translate into cloud operations.

The company recognizes the gap. Its latest announcement says the computing business remains in an early expansion stage and has relatively limited industry operating experience.

It identifies market development, technical operations, and project management as potential risks. Those are not minor caveats because they cover the central functions required to monetize the equipment.

Entive’s partnerships can help close some capability gaps. Its work with Enflame gives the project a connection to domestic AI accelerator technology.

Its service agreement with Infinigence AI offers experience supporting a model-infrastructure company. Workloads linked to Meitu provide a concrete consumer-facing inference case.

These relationships create an initial operating network, but they do not establish independent evidence that Entive can fill a much larger fleet at attractive margins.

Customer concentration remains one of the clearest warning signs. If more than 80 percent of computing sales depend on Infinigence AI, diversification has not yet caught up with the infrastructure ambition.

The company needs additional customers whose workloads, contract periods, and credit profiles reduce dependence on one counterparty. A long customer list matters less than the amount of committed and paid usage.

Competition may also compress returns. A wave of server purchasing can relieve capacity shortages, giving customers more bargaining power and lowering rental rates.

An earlier market assessment distinguished between relatively abundant general computing and tighter intelligent computing capacity. That distinction matters because not every server can serve the same AI workload.

Even within intelligent computing, demand does not distribute evenly. Customers often prefer specific accelerator architectures, software environments, geographic locations, or network connections.

Export controls and supply policies can change the equipment available to Chinese operators. Domestic accelerators reduce some exposure, but they introduce their own compatibility and adoption considerations.

Entive must therefore match purchasing decisions with credible customer requirements. Buying a large number of accelerators before workloads are secured can create stranded capacity.

The opposite risk also exists. Waiting for perfect certainty can leave a provider without enough equipment when customers request immediate service.

This makes staged deployment more credible than a single jump to the full ceiling. Investors should distinguish between shareholder authorization, signed contracts, delivered assets, installed capacity, and billable operation.

Each step reduces a different uncertainty. Authorization addresses governance, signing creates obligations, delivery transfers equipment, installation enables service, and customer acceptance starts the revenue process.

The July progress update showed that even the first purchasing round had not completed delivery. That makes execution speed a measurable test rather than an abstract promise.

Entive also needs to demonstrate financial reporting transparency around this new segment. Computing revenue, gross margin, utilization, customer concentration, finance costs, and capital commitments should become increasingly visible.

Without those disclosures, investors would see the purchase value more clearly than the underlying economics. That imbalance tends to encourage speculation rather than informed evaluation.

The company’s kitchen appliance identity should neither disqualify the strategy nor excuse weak evidence. Businesses can enter new markets, especially through subsidiaries and technical partners.

The burden is simply higher when the proposed investment dwarfs existing revenue and introduces a very different operating model. Entive must prove the transition through contracts, utilization, cash collection, and margins.

Three Signals Will Show Whether the Bet Is Working

Shareholder approval, contracted utilization, and financing costs will determine whether the server plan becomes a durable business or an expensive balance-sheet experiment.

The first signal is the shareholder vote. Approval would authorize the expanded program, while rejection or delay would slow the planned buildout.

Approval alone would not validate the economics. It would show that represented shareholders accepted the governance and financing framework presented by the board.

The more informative detail will be any additional disclosure before or around the meeting. Investors should look for purchasing phases, customer commitments, funding arrangements, or limits tied to actual orders.

If Entive provides clearer demand coverage before approving more contracts, confidence in the order-led strategy would strengthen. If disclosure remains limited to a maximum purchase amount, uncertainty would persist.

The second signal is the conversion of purchases into billable capacity. Entive should report how much equipment has been delivered, installed, tested, accepted, and placed into customer service.

The July update established a useful baseline. Cumulative contracts stood at RMB 529 million, and only part of the equipment had arrived.

The next update should show progress beyond delivery. Installed capacity matters more because boxed or partially configured equipment cannot earn service revenue.

Utilization matters most. Entive has not disclosed a utilization rate for the proposed expansion, making future customer contracts and segment revenue essential evidence.

A significant increase in computing revenue, supported by multiple customers, would strengthen the case that capacity is following demand. Slow revenue growth beside rapid asset growth would weaken it.

Customer diversification should be examined at the same time. Reducing the share attributed to Infinigence AI would lower dependence on a single relationship.

The third signal is the financing burden in Entive’s interim and annual reports. Watch debt, lease liabilities, interest expense, operating cash flow, and depreciation.

Rising debt is not automatically negative when financed assets produce dependable cash. The concern appears when financial costs grow faster than gross profit from the related services.

Entive’s 2025 loss makes this test particularly important. The company needs computing revenue that improves group economics, not merely a larger asset base.

Investors should also compare the pace of funding with customer payment terms. Long receivable periods can create cash strain even when reported revenue grows.

The company’s credit assessment already highlights high customer concentration in the computing operation. Future rating commentary may provide another view of leverage and repayment capacity.

Beyond those three signals, equipment details will shape the commercial outlook. Accelerator models, cluster design, software support, and expected service life influence which customers the platform can serve.

Entive should avoid relying on broad claims about AI demand. The more useful evidence will be named deployments, service periods, committed workload volumes, and cash received.

For enterprise technology buyers, the story illustrates why capacity sourcing requires more than comparing headline processing numbers. Provider stability, software support, network performance, and long-term maintenance can be equally important.

For developers, the expansion may add access to domestic inference capacity if Gansu Yisuan supports the required frameworks and accelerators. The practical value will depend on availability, reliability, and workload portability.

For knowledge workers and AI product users, the infrastructure remains mostly invisible. However, its cost and reliability ultimately influence response times, feature availability, and the economics of AI-enabled services.

Entive’s move is therefore worth watching beyond one Chinese listed company. It represents a wider experiment in which manufacturers and other non-cloud businesses use financing and regional partnerships to enter AI infrastructure.

Some entrants will secure anchor customers and build repeatable service operations. Others will discover that server ownership is easier than maintaining utilization and positive cash flow.

Entive has already moved beyond a press-release-only strategy. It operates a computing center, has disclosed customer work, and has signed hundreds of millions of renminbi in equipment contracts.

The unresolved question is whether those early projects justify a ceiling of RMB 2 billion. The answer will not come from the authorization itself.

It will come from delivered servers, diversified contracts, steady utilization, manageable borrowing costs, and improving operating results. Until those indicators appear together, the proposal remains both an expansion plan and a significant financial stress test.