The global AI race has expanded beyond model capabilities into infrastructure and capital organization. Alibaba announced a HK$80 billion placement this week to fund its full-stack AI ambitions, while CoreWeave's second-quarter revenue backlog stood at roughly US$104 billion, before adding over US$25 billion in new commitments early in the third quarter. Clients secure compute capacity first, companies front-load capital expenditure, and land, power, data centers, networks, and financing have become the foundational layer of AI competition. Against this backdrop, ITC Properties (00199) hosted an investor day in Shenzhen on August 26. The company has passed a resolution to rename itself "ITC New Cloud Computing Group Limited," signaling a shift from a traditional property platform to a global AI infrastructure operator connecting energy, AIDC, GPU, and Token services. The day before the event, shares closed at HK$2.875 with turnover of HK$35.62 million and a market capitalization around HK$3.1 billion.

From "property discount" to "infrastructure pricing," the transformation logic begins with a view on infrastructure cycles. Co-chairman He Xuechu compares AI to the highways and ports of the industrial era, or the communication networks of the internet age. AIDC similarly depends on land, energy, approvals, engineering, and long-term operations, layered with GPU clusters, compute scheduling, and model capabilities. "Model companies will keep iterating, some will disappear, but the infrastructure that carries productive capacity will persist. What we truly aim to manage going forward is a global AI productivity network," says He. He translates the strategy into operable assets, compute revenue, technical capability, long-term clients, and sustained cash flow, converting conceptual transformation into verifiable metrics.

Executive Director and General Manager of ITC Zhisuan, Cao Xinwei, disclosed that the company's global planned capacity is approximately 2.4GW across seven countries and 13 nodes. This represents planning and reserve scale, not operational capacity; valuation depends on how many projects clear approval, construction, delivery, and leasing. The Rudong project in Nantong has a total planned capacity of 1GW, with the first phase at 200MW, intended as a Yangtze River Delta inference compute base powered by wind, solar, and geothermal energy. The company's scenario assumes electricity prices entering the "5-yuan range" per kWh and PUE below 1.2; Cao states that first-phase capacity has already secured pre-commitments from clients. Beijing plans 50MW with key conditions such as land, power, energy consumption, and water already secured; Zhangbei plans over 100MW; Hong Kong will handle international GPU deployment and compute exports. This creates a division of labor: Nantong for inference, the Beijing area for coordination, and Hong Kong for global connectivity. Value lies not in summing capacity but in whether nodes can sit close to clients, secure low-cost energy, and build replicable delivery capabilities. "We want green power, not for the sake of being green. The core is to genuinely lower electricity prices and operating costs through multi-energy integration," says Cao.

From MW, GPU to Token: if resources are merely turned into data centers, the company would still be priced like a traditional IDC. The plan is to convert MW into GPU compute first, then through unified scheduling, inference optimization, and model services, generate Tokens, contracts, and cash flow. Four revenue streams include AIDC operations, GPU leasing and compute services, enterprise model deployment, and usage-based MaaS services: the underlying business provides scalable cash flow, while upper-layer services aim to raise unit compute revenue and margins. "If an excellent AI Infra engineer can improve GPU efficiency by 20%, the economic effect is close to adding 20% more GPU capacity without the same level of capital expenditure," notes Yang Sen, Strategy Director at ITC Properties. This point marks the valuation divide between Neocloud and traditional data centers: the former sells dispatchable, metered, and continuously optimizable compute capacity. The longer the platform operates, the more data it accumulates on models, chips, loads, and nodes, and the more likely technical efficiency translates into gross profit.

Capital constraints and valuation gap: management estimates that near-term projects including the Nantong first phase, Beijing, Suzhou, and Hong Kong Science Park total roughly 310MW, with overall investment close to HK$8.5 billion. The company plans to rely primarily on project financing, with banks and financial leasing institutions providing about 70% to 80% of funds at the project level; equity will be sourced from industrial capital, RMB funds, and strategic partners, while the listed company retains tools such as placements and convertible bonds. Financing costs, project equity ratios, and delivery pace will determine the value shareholders ultimately capture. The valuation logic presented by the company is straightforward: underlying AIDC is priced on deliverable capacity, utilization rates, and stable cash flow; GPU services depend on equipment utilization, contract terms, and capital costs; MaaS and Token platforms hinge on scheduling efficiency, unit Token cost, and revenue scale. As each project advances, the valuation anchor shifts from resource reserves to assets under construction, operating cash flow, and platform revenue. Within this framework, management believes that if near-term projects are delivered on schedule and achieve stable leasing, the company's value could be several times higher than the current market cap. This is not a consensus target price, nor does it imply the value has been realized; more accurately, the current share price still largely reflects early-stage capital expenditure and execution risk, with limited pricing of the earnings structure after AI infrastructure platformization. The discount offers potential upside but also leaves the burden of proof on the company.

Rerating begins with the first cash flow. The path management outlines is to first establish recurring compute revenue, then push GPU long-term leasing, platform scaling, and more project deliveries, eventually entering flagship project operations at scale. The valuation triggers are therefore clear: project delivery, GPU leasing and renewals, Token revenue growth, and financing milestones. Once these nodes appear consecutively, the market's valuation methodology is likely to shift. Capital markets will not wait for all projects to mature before rerating, nor will they rerate on global planning maps alone. They often start from the first megawatt of power delivered, the first GPU lease, the first Token invoice, and the first positive cash flow. Whether the stock is cheap depends on how much planned capacity the company can convert into contracts, revenue, and cash flow; based on the disclosed progress, client pre-commitments, and commercial path, this delivery curve is no longer starting from zero. For investors, the key question is no longer how big the story is, but whether the pace of delivery can outpace market expectations.