Alibaba Group Holding (BABA): Nomura sees Alibaba's 20GW cloud-capacity ambition as reinforcing long-term AI-cloud growth

Overview

This conference note examines how Alibaba's Cloud Conference announcements support its strategy to build an integrated AI stack spanning proprietary models, chips and cloud capacity. Nomura views the 20GW operated data-center target for 2032 as the most important disclosure and maintains a constructive long-term view of Alibaba Cloud.

Core views

Alibaba presented AI as the foundation of a future “machine intelligence era,” with CEO Eddie Wu arguing that infrastructure demand remains at an early stage. The company expects more capable AI agents to undertake longer and more complex tasks, driving substantial growth in compute and token consumption. Nomura interprets Alibaba's response as a coordinated investment program across three layers: foundation models, proprietary chips and cloud infrastructure. At the model layer, Alibaba is developing Qwen toward longer-horizon agentic capabilities, recursive self-improvement and native multimodality. Management plans a new 5–10T-scale frontier model, compared with 2.4T parameters for Qwen3.8-Max. In discussions with management, the company rejected the idea that it should reduce proprietary-model investment to free compute for cloud monetization. Instead, management regards frontier foundation models as strategically important for future AI systems and applications. At the hardware layer, T-Head is building a vertically integrated data-center stack that includes AI accelerators, CPUs, smart NICs, storage and scale-up interconnects. Zhenwu M890 already supports Alibaba Cloud supernodes running models with more than 2tn parameters. Its successor, Zhenwu V900, is scheduled for 1Q27 and is designed to deliver roughly three times M890's single-chip compute performance. Together with Yitian CPUs, Panmai smart NICs, Zhenyue storage chips and the proprietary ICN interconnect, this stack is intended to support agentic inference and large-scale model deployment. Nomura estimates T-Head chips currently represent about 10% of Alibaba's AI compute resources and expects internally developed compute to become more meaningful from 2027F, improving self-sufficiency and potentially supporting cloud margins through lower costs than externally sourced accelerators. Nomura considers Alibaba's target to operate 20GW of global data-center capacity by 2032 the key announcement. This compares with the institution's estimate of about 4GW currently, making the target about five times larger, and it is also about ten times Alibaba's 2022 capacity. The target is for capacity operated by Alibaba rather than necessarily owned and financed entirely on its balance sheet. Nomura therefore expects a mix of self-owned facilities, leased capacity and third-party-funded structures, potentially including a managed-data-center model in which external investors own and fund physical infrastructure while Alibaba supplies the cloud platform, cluster technology, operating capabilities and enterprise access. Using an assumed infrastructure investment of USD30bn per GW, Nomura estimates that moving from 4GW to 20GW could imply roughly USD480bn of incremental ecosystem infrastructure investment through 2032F. This is not presented as Alibaba's own incremental capital expenditure because leasing and third-party capital may fund part of the build-out. The larger capacity target reinforces Nomura's constructive view of Alibaba Cloud's medium- and long-term revenue potential: Alibaba has previously indicated an ambition for USD100bn of external cloud revenue around 2030, while Nomura estimates external AI-cloud revenue could potentially reach USD168bn by 2032F if operated capacity reaches 20GW. Nomura maintains its Buy rating and USD178 SOTP-based target price. Its valuation assigns USD87bn to China Ecommerce Group using 5x FY27F P/E, USD278bn to AliCloud using 7x FY28F P/S, and USD40bn to the net value of non-core assets, including international ecommerce. The USD178 target implies 22x CY27F (FY28F) P/E, compared with 15x currently.

Analysis framework

Nomura links management's conference announcements to a three-layer AI infrastructure strategy: models, chips and cloud capacity. It assesses the scale and timing of planned model and chip upgrades, estimates the current and future role of proprietary compute, then translates the 20GW capacity target into potential ecosystem investment and external AI-cloud revenue. The target price is derived using sum-of-the-parts valuation for China ecommerce, AliCloud and non-core assets.

Methodology notes

• Valuation methodsSOTP (Sum-of-the-Parts) Valuation

  Sum-of-the-parts valuation

  Nomura values Alibaba by separately assigning values to China Ecommerce Group, AliCloud and non-core assets, then combines those values into a USD178 target price.

• Valuation methodsP/E and PEG Valuation

  P/E multiple valuation

  Nomura values China Ecommerce Group at 5x FY27F P/E and notes that its target price implies 22x CY27F (FY28F) P/E.

• Valuation methodsPS valuation

  Price-to-sales multiple valuation

  Nomura values AliCloud at 7x FY28F P/S, reflecting a revenue-based valuation approach for the cloud business.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Alibaba Group Holding (BABA.US)

  Primary covered company; its integrated AI model, chip and cloud strategy underpins Nomura's maintained Buy rating.

  Strengths

  Planned expansion across Qwen models, T-Head compute hardware and global operated cloud capacity; potential for rising proprietary-compute contribution and external AI-cloud revenue.

  Comparison

  The 20GW 2032 target is approximately five times Nomura's current estimate of about 4GW and roughly ten times Alibaba's 2022 capacity.

  Risks

  Margin downside from ramp-up investments and regulatory risk related to payments and internet finance, which could affect Alibaba's main business and its value in Ant Group.

Key data

• Operated data-center capacity target20GW by 2032About five times Nomura's approximately 4GW current estimate and about ten times Alibaba's 2022 capacity.
• Estimated incremental ecosystem infrastructure investmentUSD480bnBased on USD30bn per GW to expand operated capacity from 4GW to 20GW; not equivalent to Alibaba's own capital expenditure.
• Planned frontier model scale5–10T parametersCompared with 2.4T parameters for Qwen3.8-Max.
• Zhenwu V900 performanceAround 3x M890 single-chip compute performanceV900 is scheduled for 1Q27.
• T-Head share of Alibaba AI computeApproximately 10%Nomura expects the proprietary-compute share to increase meaningfully from 2027F.
• Potential external AI-cloud revenueUSD168bn by 2032FNomura's estimate if Alibaba reaches 20GW of operated capacity.
• Target priceUSD178Maintained Buy rating; target implies 22x CY27F (FY28F) P/E versus 15x currently.

Impact & implications

Nomura believes the 20GW capacity goal materially strengthens the case for Alibaba Cloud's long-term external AI-cloud revenue opportunity. A mix of owned, leased and third-party-funded infrastructure could allow capacity to expand faster than balance-sheet spending alone, while greater use of T-Head chips could improve compute self-sufficiency and support cloud profitability over time.

Risks

• Margin downside could result from the ramp-up in investments.
• Regulatory risks related to the payments and internet-finance industry could hurt Alibaba's main business and its value in Ant Group.