Third-quarter growth slowed. Why does Goldman Sachs still expect stronger contributions ahead?

  • Revenue in the third quarter of 2026 was 1.92 billion yuan, up 30% year on year, slowing from 50% growth in the first half and coming in 7% below Goldman Sachs' expectations. Adjusted net profit was 593 million yuan, up 25% year on year and 9% below its expectations. The latter excludes the 3.8 billion yuan investment gain from the C-REIT listing in the third quarter of 2025, making this a comparison of adjusted earnings.
  • Goldman Sachs attributes the shortfall mainly to the timing gap between AI computing capacity investment and revenue recognition. Management says the two can be separated by approximately 4 months. Following increased investment in the third quarter, Goldman Sachs expects the associated revenue and profit contributions to emerge gradually from the fourth quarter.

Why did AI computing capacity investment suddenly accelerate?

  • The company made approximately 8 billion yuan of new AI server investments in the third quarter, close to the 10 billion yuan invested in all of 2025. Investment activity was lower in the first half amid high GPU costs and the company's focus on establishing supply-chain partnerships; procurement accelerated markedly after supply improved.
  • On a capital expenditure basis, third-quarter spending was 6.8 billion yuan, up 170% year on year and 286% quarter on quarter. Fixed assets and construction in progress increased by 8.5 billion yuan quarter on quarter. The investment is already reflected in assets, while revenue recognition lags.
  • The shift toward AI data centers is driven by differences in returns: management targets a gross margin of approximately 45% and a payback period of 3 years. Competition in traditional data centers is intense, so the company may slow capacity expansion in that business in 2027 and direct more resources toward AI data centers.

How will the investment plan of up to 57 billion yuan be funded?

  • The company plans to invest up to approximately 57 billion yuan in AI computing capacity from the fourth quarter of 2026 to the first quarter of 2027. Related financing includes 50 billion yuan in new credit facilities and 6.8 billion yuan in equity financing at its indirectly wholly owned subsidiary Hebei Runlan, with funds mainly earmarked for AI server purchases.
  • Of the equity financing, 4 billion yuan provided by a wholly owned subsidiary of the Export-Import Bank of China has already been received, with up to another 2.8 billion yuan in subsequent financing. The credit facilities, funds received and subsequent financing are at different stages; the subsidiary's financing does not dilute shares at the listed-company level.
  • Funding pressure remains during expansion. Goldman Sachs forecasts net debt to EBITDA of 4.8 times and 4.9 times in 2026 and 2027, respectively, and free cash flow yields of -14.6% and -10.6%, respectively. Financing supports purchases, while operating payback must await subsequent revenue and profit contributions.

When will new capacity begin generating revenue?

  • Domestic capacity delivered in the third quarter totaled 290 MW, comprising 200 MW at the Langfang AI data center and 90 MW at traditional data centers. The focus of capacity expansion has clearly shifted toward AI.
  • At its first overseas campus on Batam Island, Indonesia, the company has delivered 20 MW ahead of schedule. It expects to deliver the remaining 40 MW of phase one in the fourth quarter of 2026 and begin receiving US dollar revenue from Singaporean customers in the first quarter of 2027. Delivery and the start of revenue are separate milestones.
  • The company also plans to deliver 100–200 MW on Batam in 2027. Delivery of the Hong Kong project is expected to begin in 2027 and finish in 2028. Overseas expansion is another avenue of growth, but revenue contributions depend on construction and delivery progress.

What conditions underpin Goldman Sachs' buy rating?

  • Goldman Sachs maintains its buy rating and 12-month price target of 109 yuan. It forecasts compound annual growth rates of 27% for utilized IT capacity and 33% for revenue over 2025–2030. Its thesis is that capacity reserves, customer relationships and low-cost financing support expansion, while GPU computing services add profit under favorable demand and pricing conditions.
  • The price target applies 20 times estimated 2030 EV/EBITDA, discounted to the end of 2026 at a cost of equity of 10%. Data center operations and GPU computing services are assigned multiples of 25 and 10 times, respectively. Order wins, utilization ramp-up and pricing determine investment returns; chip supply, financing and overseas execution also affect the pace of expansion.