A Thousand Billion Yuan Influx into Green-Powered Computing: China’s “East-to-West” Strategy Accelerates Toward a “Global Token Factory”

This week, the most striking news in the global AI computing sector did not come from a Silicon Valley product launch, but from a national-level strategic conference held on the grasslands of northern China. The 2026 Green Computing (Artificial Intelligence) Conference concluded on August 22 in Hohhot, Inner Mongolia, delivering results that stunned the industry: 13 key projects were signed on site, with a total investment of 136.1 billion yuan (approximately $19 billion). This figure not only sets a new record for a single signing event in China’s computing industry, but also sends a powerful policy and market signal. As the global AI race enters the deep waters of computing power and energy competition, China—leveraging its institutional strengths and massive market—is pushing the “East-to-West Computing Resource Transfer” project from blueprints to full-scale, cluster-based, and whole-industry-chain implementation.

I. The Energy Dilemma in the Global Computing Race and China’s Strategic Answer

Today, global demand for AI computing is growing exponentially. The training costs of large language models and multimodal applications remain staggeringly high, with electricity expenses and carbon emissions becoming hard constraints on industry growth. Recently, regions such as Pennsylvania in the United States have introduced new regulations on power access for hyperscale data centers, underscoring the reality that “computing is electricity.” Against this backdrop, China’s decision to build ultra-large-scale computing clusters in Inner Mongolia—a region rich in renewable energy—carries profound strategic intent. With abundant wind and solar resources, Inner Mongolia can supply sufficient low‑carbon power to energy‑intensive AI training facilities. The conference marks a turning point: China is no longer relying solely on spillover from developed coastal areas, but proactively constructing new computing growth poles in western energy hubs, aiming to secure cost and energy advantages in global competition. This not only demonstrates a firm commitment to the country’s “dual carbon” goals, but also pioneers a sustainable development path for the AI industry.

II. The Capital Call: Decoding the 136.1 Billion Yuan Investment Landscape

The 136.1 billion yuan in signed projects is far from a scattered set of commercial deals; it is a meticulously orchestrated industrial capital mobilization, with participants spanning the entire value chain—from primary energy to end‑user AI applications. Geographically, the investments are highly concentrated in two core data hubs: Hohhot and Ulanqab.

Hohhot secured eight heavyweight projects, including the much‑watched Volcano Engine GW‑class park, a chip laboratory established by the domestic AI chip leader Cambricon, and the Haiwu Computing Equipment Industrial Park. Ulanqab took five projects, equally impressive, including a second Volcano Engine GW‑class computing center, the computing‑electricity synergy project led by GCL Energy Technology, and a massive data center expansion by China Telecom.

A closer look at this investment matrix reveals a highly integrated structure. GCL Energy’s participation addresses the fundamental question of “where the power comes from”; its computing‑electricity synergy project will explore integrated source‑grid‑load‑storage operations, ensuring stable and low‑cost green power for computing centers. Infrastructure providers such as China Data Group will build the “computing factories.” The Cambricon chip laboratory means that high‑end domestic AI chips will be deployed, validated, and iterated in real‑world, massive‑scale commercial environments. Meanwhile, the simultaneous launch of two Volcano Engine GW‑class parks represents strong endorsement from a major computing demand‑side player of Inner Mongolia’s quality and scale. This closed‑loop investment structure—combining energy suppliers, infrastructure builders, chip designers, and application giants—signals that the Hohhot‑Ulanqab cluster will move beyond the primitive model of renting out server racks, evolving into a self‑reinforcing digital ecosystem.

III. From “China’s Cloud Valley” to “Global Token Factory”: A Strategic Leap

The core theme of the conference was to drive a fundamental strategic upgrade for the computing industry—transforming it from merely providing basic storage and computing resources to becoming a super‑factory that produces high‑value “tokens” at massive scale and low cost. In the generative AI era, tokens are the fundamental units of intelligence; whoever can produce tokens cheaper and faster gains the lifeblood of AI industrialization.

Inner Mongolia enjoys unique geographical advantages for this transformation. First, its low annual average temperatures facilitate natural cooling for data centers; combined with abundant wind and solar power, it can achieve a power usage effectiveness (PUE) of below 1.2—a world‑class level—dramatically reducing per‑unit computing energy costs. Second, Hohhot has already built a low‑latency computing service circle of “2 milliseconds within the Hohhot‑Baotou‑Ordos‑Ulanqab area and 5 milliseconds to the Beijing‑Tianjin‑Hebei region.” This means that data from AI‑intensive enterprises in Beijing and Tianjin can be transmitted to Inner Mongolia for remote computing and returned in the blink of an eye; the disadvantage of physical distance has been completely overcome by ultra‑fast fiber optic networks.

In terms of existing capacity, Inner Mongolia’s total computing power has reached 345,000 P (petaflops), of which 326,000 P are intelligent computing, ranking among the top nationwide. The Helingeer New Area has already gathered more than 62 large‑scale data centers and built the country’s first green‑computing AI full‑stack token trading platform. The projects signed this week will massively expand that capacity. Once the Volcano Engine GW‑class parks are completed, their total computing power will be measured in tens of thousands of P—enough to support the parallel training of dozens of billion‑parameter models simultaneously. This will fundamentally change the current industrial landscape, where many domestic AI startups are priced out by exorbitant computing costs.

IV. “Computing‑Electricity Synergy” Becomes the Core Paradigm for New Infrastructure

One strong signal from the conference is that “computing‑electricity synergy” has been elevated from an industry proposal to a mandatory requirement for national computing hubs. The conference theme placed special emphasis on this, mirroring international trends. As AI chip power consumption rises year by year, a single hyperscale data center now consumes electricity equivalent to a medium‑sized industrial city. GCL Energy’s project is not merely about power supply; through virtual power plants, energy storage scheduling, and smart microgrids, it will enable the massive computing cluster to flexibly respond to grid load changes—ensuring both stability and greater absorption of renewable energy.

This new paradigm means that future data center construction will no longer follow the old “build the power plant first, then the server room” model; instead, it will adopt an integrated approach of “planning, building, and operating energy and computing together.” It is foreseeable that all new large‑scale domestic computing hubs will adopt the “Helingeer standard” set by this conference, making green‑power usage ratios, PUE metrics, and computing‑electricity coordination capabilities hard entry requirements. This will not only reduce carbon emissions but also reshape the cost curve of AI computing from the ground up. Over the next three years, the marginal cost of training AI models in China could drop significantly, giving rise to a more vibrant ecosystem of AI‑native applications.

V. The Comprehensive Rise of Domestic Computing Power

Beyond the sheer scale of the 136.1 billion yuan signings, the conference also showcased an impressive array of technological breakthroughs from the domestic computing power supply chain. Multiple domestic chip and computing companies made major announcements.

China‑based Zhongcheng Hualong released its new‑generation high‑performance inference chip, optimized for large‑model deployment, achieving notable gains in energy efficiency. Alibaba’s T‑Head announced that its second‑generation AI chip will tape out in the second half of the year, targeting the cloud AI training market and competing directly with mainstream Nvidia accelerators. Sugon (Dawning Information Industry) unveiled a new super‑node architecture for large‑scale distributed computing. Together with the already large‑scale commercial deployment of Huawei’s Ascend super‑nodes, a comprehensive picture of domestic computing power—comprising national champions, internet giants, and specialized chip designers—is becoming increasingly clear.

Notably, Cambricon’s chip laboratory will be directly located in Hohhot, meaning that domestic chips will no longer remain in lab‑benchmark phases; they will undergo rigorous long‑term stability testing in real‑world 10,000‑card cluster environments. This application‑driven, scenario‑based development model will greatly accelerate the maturation of indigenous AI chips. As these projects come online, Helingeer is poised to become the world’s largest commercial deployment and validation base for domestic AI chips—a strategic asset of immeasurable value for securing China’s AI supply chain.

VI. From National Strategy to Industrial Ecosystem: The Far‑Reaching Impact Toward 2027

Looking at the big picture, the 136.1 billion yuan investment is not merely a capital rally; it marks a critical juncture in deepening the “East‑to‑West Computing Resource Transfer” national strategy. It sends a clear message to global markets: China is resolutely turning AI computing into a ubiquitous national infrastructure, as accessible as water and electricity. Unlike Silicon Valley’s largely market‑driven evolution, China leverages top‑down design, policy guidance, and a combination of state‑owned and private capital to dramatically shorten the cycle from scarce resource to affordable utility.

Looking ahead to 2027, as these billion‑dollar projects are fully or partially commissioned, the underlying computing supply structure of China’s AI industry will undergo a qualitative leap. Reduced computing costs will directly ignite intelligent transformation across vertical industries: medical imaging AI training cycles could shrink from months to days, autonomous driving simulation costs will plummet, and scientific research in drug discovery and weather forecasting will gain unprecedented computational support. At the same time, the operation of large‑scale green‑powered computing centers will offer a valuable “China solution” to the global AI industry’s energy challenges.

In this digital race that will define the next two decades, Inner Mongolia’s Helingeer is transforming from the former “China’s Cloud Valley” (a passive data backup site) into a “Global Token Supply Base” that actively exports intelligence. When clean electricity and intelligent computation intertwine across this vast grassland, a greener, more affordable, and more self‑reliant AI era is accelerating into reality. This is not only an opportunity for Inner Mongolia, but also China’s solid contribution to global AI governance and development. With the ink dry on these contracts, the real challenges—and the most exciting phase of industrialization—have only just begun.