As foundation models become commodities, a new industry is emerging to convert them into usable intelligence. In China's AI infrastructure sector, it has picked up a name: the Token Factory.
A Token Factory doesn’t sell compute capacity. It sells inference capacity: manufactured intelligence, priced and delivered like a utility.
In Wuxi, China, a new type of AI infrastructure facility has appeared. It does not train frontier models. It does not manufacture chips. It manufactures something else: tokens.
That may sound like a strange distinction. After all, most of the AI race today is still measured by a familiar scoreboard: which model is smarter, which benchmark it wins, which company reaches the next generation of reasoning capability.
But inside the enterprises actually trying to deploy AI, the questions are becoming much less glamorous.
What happens if the network goes down?
Who catches it when the model is wrong?
Who is responsible when an AI system leaks information or makes a decision that cannot be explained?
China’s enterprise AI market is already moving beyond the model race. In 2025, industry tracking data from China Large Model Procurement Monitoring Report 2025, recorded more than 7,500 large-model-related procurement contracts, with disclosed contract values approaching 30 billion yuan. But the surprising part was what those contracts actually demanded: they rarely focused on which model was smartest. Instead, they focused on deployment, security, reliability, and control.
Read enough of these contracts and a pattern emerges. The recurring requirements are central deployment, edge execution, offline caching, resume after disconnect, tiered permissions, and audit trails. Acceptance criteria alone can run for dozens of pages.
These are not the questions people ask when they open ChatGPT. They are the questions organizations ask when AI moves from a demo into a production system.
That gap — between whether a model can answer well and whether an organization can actually run it — is where a new layer of the AI industry is emerging.
China is not the only place where this layer could exist.
NVIDIA CEO Jensen Huang has described the next generation of AI infrastructure as “AI factories” systems that transform compute and electricity into tokens. In the United States, however, this capability is largely absorbed inside vertically integrated platforms such as AWS, Azure, Google Cloud, and the major AI labs.
What makes China different is not that it invented the concept, but that it is among the first markets where this idea has begun to evolve into an independent industry with local governments, infrastructure providers, and startups building what China calls “Token Factories.”
What a Token Factory Actually Is #
Three misconceptions are worth clearing up first, because each one understates what’s being built.
It is not an API relay. A basic API proxy just forwards a request to a model provider’s endpoint; it doesn’t touch the underlying efficiency of inference. A Token Factory does something structurally different: it runs its own inference stack, splitting the process into stages that can be optimized independently, and pooling chips from multiple vendors under a single scheduler. The goal is one number: how many tokens this hardware can produce per dollar, per watt, per second.
It is not simple compute rental. A cloud sells compute capacity: GPU hours, a fixed unit regardless of what you do with it. A Token Factory sells inference capacity instead, the output of that compute, metered in tokens. Its margin depends entirely on how efficiently it converts hardware and electricity into that output. This is a genuinely different business, where the core metric isn’t how many GPUs a company owns but how many tokens each GPU can produce, and at what quality. Push that efficiency up, and the same hardware generates more revenue without buying a single additional chip.
It is not just a data center. A useful way to see the stack: at the bottom sit GPUs, electricity, and networking. In the middle sits the software layer that actually does the work: inference frameworks, model scheduling, cache optimization, cross-chip compatibility. On top sits the delivery layer: APIs, agents, industry-specific applications that enterprises actually touch. A data center is the bottom layer alone. A Token Factory is all three, sold as one product.
Put simply: the product isn’t compute, and it isn’t an interface. It’s manufactured intelligence, priced and delivered like a utility.
Why This Layer Is Emerging in China #
No single explanation carries the whole story. Five conditions are converging at the same time, and it’s the combination, not any one of them, that matters.
Open-weight models supply the raw material.
DeepSeek and Alibaba’s Qwen release full-size open weights. Moonshot’s Kimi and Zhipu’s GLM have released open weights too, though open-sourcing plays a smaller role in Moonshot’s own business than it does for DeepSeek. The exact mix varies company to company, but the pattern holds: several competitive Chinese models are simultaneously available for any team to download, deploy, and optimize, something that doesn’t exist among the closed U.S. frontier of OpenAI, Anthropic, and Google. If a model’s weights are public, the model builder trains, and someone else can manufacture and distribute.
Meta’s Llama is the obvious counterexample, and it’s worth addressing directly. An open U.S. model exists, but it has not produced the same independent inference ecosystem that is emerging around China’s model ecosystem. Most of the serious inference infrastructure built around U.S. models sits inside AWS Bedrock, Azure OpenAI Service, and Google Vertex AI, each of which already does real multi-model inference and routing across providers like Anthropic, Meta, and Mistral. That’s genuine infrastructure, just not an independent layer: an integrated version of the same function, embedded inside a hyperscaler ecosystem rather than sold as a public utility that works across any model or chip.
Unlike the U.S. frontier AI ecosystem, where companies like OpenAI are tightly integrated with hyperscale cloud platforms such as Microsoft Azure, China’s AI landscape is more fragmented. Some major technology companies, including Alibaba, ByteDance, Tencent, and Baidu, operate large cloud platforms of their own, but no single AI stack dominates the market from frontier models to global-scale cloud infrastructure.
That fragmentation, combined with open-weight models, creates room for a coordination layer to emerge.
A fragmented ecosystem needs someone to hold it together.
China’s AI stack is fragmented in a way the U.S. stack isn’t. On the model side: DeepSeek, Qwen, Kimi, GLM, and more, each with different architectures, strengths, and pricing. On the chip side: a shrinking pool of legacy Nvidia hardware as export controls tighten, alongside a fast-growing lineup of domestic accelerators such as Huawei Ascend, Cambricon, and Moore Threads, each running its own software stack, none of them interchangeable without real engineering work. On the infrastructure side: local data centers, telecom operators, and third-party clouds, each built for a different region and a different regulatory environment.
None of these pieces talk to each other natively. Someone has to translate a request into which model, on which chip, deployed where, and return a clean token stream on the other end. That’s the role SiliconFlow founder Jeffrey Yuan (Yuan Jinhui) has described his company playing: connecting models, chips, and applications. SiliconFlow has built exactly that. Its self-developed inference engine has adapted to more than 150 models, serving close to 10 million users and over 10,000 enterprise customers. Third-party data puts it among the top four providers in China’s public-cloud model-serving market, the only startup in that group, and on OpenRouter, an independent U.S. model-routing platform. It has emerged as one of the most-used Chinese model-serving platforms, frequently ranking among the top providers by daily token volume. A smaller, more technical example is Vectron, an optimization platform built by the Chinese startup Shishi Technology, which advertises compatibility with more than ten domestic chip types and twenty mainstream models: a narrower operator making the same underlying bet, that the money is in squeezing more usable tokens out of the same hardware, not in owning more of it.
Enterprise demand requires the model to leave the cloud.
Ask any enterprise buyer what stopped their AI rollout, and the answer is rarely that the model wasn’t smart enough. It’s usually that the data can’t leave the building. This isn’t unique to China. Samsung banned external generative AI tools company-wide in 2023 after employees pasted proprietary source code into ChatGPT, and similar bans have rippled through banks, chipmakers, and law firms worldwide since. What started as an emergency reaction has hardened into standard compliance practice across those industries. In China’s government and enterprise market, this requirement has its own name: data never leaves the domain.
That constraint reframes the entire buying decision. It’s no longer which API is cheapest per million tokens. It’s who can install this behind our firewall, with permission systems that track which employee can see which document, with audit logs that hold up if something goes wrong, and with a service contract that says who’s accountable when it doesn’t. That’s a systems-integration problem, not a subscription. A Token Factory is built to sell exactly that: not intelligence as a feature, but intelligence with a chain of custody attached.
Energy and compute infrastructure make scale possible.
Inference cost isn’t just about GPUs. Electricity can become one of the largest operating costs of running an inference operation. China has been experimenting with infrastructure models designed to reduce that cost, most visibly the East Data, West Computing program, which routes compute-heavy workloads toward the west, where land and renewable power are cheap, while demand stays concentrated in the east. One project in Zhongwei, Ningxia, sits next to a region with more than 18 million kilowatts of installed renewable capacity generating over 16 billion kilowatt-hours a year, pairing data centers directly with wind and solar generation to cut the transmission losses of pulling power off a conventional grid.
The metric that matters here isn’t how much electricity a facility uses. It’s how many tokens each kilowatt-hour produces, a number a Token Factory can actually move through better cooling, better power delivery, and tighter software optimization. That reframes energy from a fixed cost into an engineering variable, exactly the kind of problem an industrial-scale operator is built to solve and a small deployment team is not.
The state is building supply before demand fully exists.
The clearest way China’s Token Factory buildout differs from the U.S. market isn’t technical. It’s who’s paying for it, and when. Local governments and state-owned platforms aren’t just subsidizing this layer; some are building and operating it directly. Beijing Data Group’s Jingsuan Token Factory, the city’s first state-owned entrant, is positioned around trust rather than price: a compute foundation safe and compliant enough for state-owned enterprises that can’t take a chance on an unproven vendor, billing customers for what they actually consume rather than selling them hardware upfront.
China’s telecom operators have moved the same direction, explicitly reframing their core business as a pivot from traffic operations to token operations, treating tokens the way they once treated mobile data plans: a metered, mass-market utility sold at consumer prices. The difference is not that U.S. carriers lack AI infrastructure ambitions. Companies like Verizon and AT&T have invested in cloud, edge computing, and enterprise AI services. But no major U.S. telecom operator has yet repositioned itself around token-based AI services in the same way Chinese carriers are attempting to do.
None of this activity waited for enterprise demand to prove itself first. In many cases, the logic has been the reverse: build the capacity, lower the cost of access, and let demand catch up. American hyperscalers are also spending heavily on AI infrastructure, but much of that investment remains a corporate bet on future demand, with the payoff expected to come through their own cloud, model, and enterprise businesses. In China, a portion of that same infrastructure buildout is being driven by a broader mix of players, including state-backed entities, telecom operators, cloud providers, and startups, with a different tolerance for how long the market may take to mature.
Put together, these five conditions describe two different ways of organizing the same underlying challenge: how to turn raw model capability into something a business or government agency can actually deploy.
The difference is not simply vertical integration versus horizontal specialization. Both ecosystems contain elements of both. The U.S. has vertically integrated stacks such as OpenAI-Microsoft and Google, but it also has independent infrastructure providers emerging around AI compute and inference. China has vertically integrated players such as Huawei, Alibaba, and others, but its broader AI ecosystem remains more fragmented across model developers, cloud providers, chip suppliers, telecom operators, and application companies.
That fragmentation creates room for a distinct coordination layer: companies that connect models, compute, and applications, and turn distributed AI capabilities into standardized, metered services. This is the role China’s Token Factories are beginning to play.
None of this guarantees the model will succeed. Hardware depreciates quickly enough that a facility’s advantage can disappear within a product cycle. If token prices continue falling while hardware costs and infrastructure commitments remain fixed, margins could be squeezed from both sides. And there is nothing preventing AWS, Azure, or Google Cloud from building more sophisticated inference-optimization services of their own, potentially absorbing parts of this layer over time, just as cloud providers have absorbed other infrastructure functions in the past.
The Token Factory model is therefore not a guaranteed winner. It is one possible organizational response to the same question every AI ecosystem now faces: once intelligence becomes widely available, who will make it usable, affordable, and operational at scale?
But the underlying wager isn’t really about whether China can produce tokens more cheaply than the U.S. can this quarter or next. It’s about what happens once tokens become a standardized commodity, priced and metered like electricity: whoever controls the production infrastructure controls the entry point to everything built on top of it. China has run this playbook before, building telecom base stations years ahead of the mobile data traffic that would eventually fill them, on the bet that supply creates its own demand. The Token Factory looks like the same bet, updated for a different kind of network. The first era of AI was about building intelligence. The next may be about building the infrastructure that manufactures it.