What is behind the upward revision?

  • Goldman Sachs maintains its forecast for Alibaba’s combined capital expenditure in FY27–29 at RMB 712 billion yuan, but raises its revenue growth forecasts for AI cloud and computing from 53%/38%/33% to 53%/45%/39%. The main change is not an increase in expansion budgets: tight supplies of high-end chips and higher rental prices have improved the near- and medium-term revenue outlook for cloud computing.
  • On this basis, Goldman Sachs raises its revenue and earnings-per-share forecasts for FY27–29 by about 0%/2%/4%, respectively. The higher AI revenue mix and cloud-business margin contribution are partly offset by increased investment in AI labs and applications.

What does the 20 GW computing-power target require?

  • In its base case, Goldman Sachs expects Alibaba’s AI IT computing capacity to rise from about 5 GW at the end of 2026 to 12.8 GW at the end of 2032, below the company’s 20 GW target. The upside case reaches 19.6 GW, but requires an additional 37 billion dollars of investment over FY27–32 compared with the base case, and assumes that about 30% of computing capacity is obtained through leasing.
  • This means the 20 GW target cannot be reached through expansion of self-built capacity alone: leased computing capacity can support scale, but changes how capital expenditure and revenue are matched. Goldman Sachs also notes that domestic chips have lower unit prices, but the need for more accelerator cards, networking, power-supply systems and cooling narrows the cost advantage at the system level.

When can cloud take over e-commerce’s funding role?

  • Goldman Sachs’ base-case model expects cloud EBITDA to match e-commerce in the quarter ending March 2028, and to exceed e-commerce in the quarter ending September of the same year. The group’s funding gap is expected to narrow from the end of 2027 and turn into a surplus around the middle of 2028; group free cash flow is expected to turn positive from FY29. The company says it has no plans for equity financing for the remainder of FY27. Goldman Sachs believes cash, bank borrowing, bonds or convertible bonds could also be financing channels, but these are not committed funding sources.
  • This projection still depends on a stable contribution from e-commerce. Goldman Sachs forecasts domestic GMV growth of 1%/2%/1% in FY27–29, and a 5% decline in customer management revenue in FY27. Meanwhile, EBITA losses in the AI labs and applications businesses are expected to be about RMB 51.4 billion yuan and 57.6 billion yuan in FY28 and FY29, respectively, reducing the funding buffer.

Can returns on AI computing capacity be sustained, and what is the investment view?

  • Goldman Sachs’ model assumes that the EBITDA margin on AI computing remains above 50% over the next year or two, then normalizes as rental prices revert. If assets with an approximately 6-year life recover their costs in less than 3 years, and cash flows are maintained in years 4–6, the pre-tax unlevered IRR is about 24%; even if subsequent cash flows are halved, it is still about 15%. Accordingly, customer contract terms and revenue visibility affect the return assessment.
  • Goldman Sachs expects MaaS to rise from 34% of AI computing revenue in FY27 to 70% in FY32. As a result, subsequent monetization warrants more attention to Token pricing, not just the hourly rental price of GPUs. It believes the full-stack chips, Qwen models and applications can sustain monetization after rental prices fall. Goldman Sachs maintains its Buy rating and 12-month price targets: US ADS at $ 177 and Hong Kong shares at HK$ 172, emphasizing that resilience in core e-commerce remains critical.