1. [para. 1][para. 2] A Moody's Ratings report published Thursday says Chinese hyperscale cloud providers will keep spending heavily on AI over the next two years, but capital expenditures will remain well below those of U.S. rivals, widening the global computing infrastructure gap. Chinese firms are stepping up AI investment yet have smaller revenue bases and weaker financial flexibility, making them more reliant on debt.
2. [para. 3] Moody's expects total capex by China's leading technology companies to rise from about $65 billion in 2025 to $140–165 billion annually in 2026 and 2027. The six largest U.S. hyperscalers are projected to spend over $785 billion in 2026, roughly six times Chinese peers, potentially approaching $1 trillion in 2027.
3. [para. 4] Company disclosures show the scale difference: Alibaba reported capex of 67.7 billion yuan ($10 billion) for the quarter ended June 2026; Tencent's rose 176% year-on-year to 52.8 billion yuan; Baidu's nearly tripled to 11.4 billion yuan. In the U.S., Alphabet, Meta, and Microsoft reported $44.9 billion, $31.1 billion, and $31.9 billion, respectively.
4. [para. 5][para. 6] The spending gap will likely translate into divergence in deployed capacity. Per the IEA, U.S. data-center capacity was 52 GW at end-2025 versus 28 GW in China; U.S. should reach 100 GW by 2030 (~14% annual growth), versus China's 67 GW (~19%), so the absolute gap widens. Chinese per-megawatt build-out costs are lower, aided by cheaper land and construction.
5. [para. 7][para. 8] Policy support could help: on Aug. 19, China's National Development and Reform Commission discussed integrating computing capacity, new power systems, and next-generation communications networks. Still, financing conditions remain much more favorable for U.S. providers, which have higher equity valuations and deeper investment-grade debt markets.
6. [para. 9][para. 10] U.S. providers also use leases: the six majors have about $1.2 trillion in existing and uncommenced data center lease commitments, including over $820 billion uncommenced. Chinese providers trade at lower valuation multiples, raising equity costs. On Aug. 23, 2026, Alibaba planned an HK$80 billion ($10.2 billion) share placement for its AI strategy.
7. [para. 11] Rising AI investment is being funded from a smaller profit base. Alibaba's core e-commerce faces slower gross merchandise volume growth and competition from PDD Holdings, Douyin, and Kuaishou; on-demand retail and food delivery competition weighs on margins at Alibaba and Meituan, though reduced food-delivery subsidies should improve unit economics.
8. [para. 12][para. 13][para. 14][para. 15] Heavier AI spending will pressure free cash flow and leverage in both countries. As capex growth outpaces operating cash flow, FCF should decline sharply in 2026–27, and several Chinese companies could turn negative. In the U.S., capex should reach or slightly exceed operating cash flow; Meta, Alphabet, and Oracle issued large investment-grade bonds to pre-fund multiyear plans. Chinese providers face similar pressure from a smaller base, with capex-to-operating-cash-flow ratios rising sharply; AI spending also lifts R&D, model/application costs, and upfront payments. Debt financing has been disciplined, with much new spending absorbed by internal cash flow and net reserves; selective offshore dollar and yuan bond issuance could increase.
9. [para. 16][para. 17] Large cash balances remain a key credit strength. Chinese cloud providers hold sizable net cash reserves; U.S. peers hold hundreds of billions in cash and securities, allowing elevated spending to be absorbed over 12–24 months without significant credit deterioration. Strains are visible: Meta's Q2 free cash flow fell 91% to $784 million, and it raised its 2026 capex guidance lower end to $130–145 billion.
AI generated, for reference only
Sign in with Google
Sign in with Facebook
Sign in with 财新