ARTIFICIAL intelligence (AI) spending is set to remain one of the biggest investment themes through the rest of the decade, as technology giants race to build the computing infrastructure needed to support the next wave of AI adoption.
The opportunity remains substantial, but investors are increasingly having to look beyond headline AI growth and ask which companies can turn massive spending into sustainable cash flow and earnings.
Moody’s Ratings says heavy AI investment is putting pressure on the financial strength of major technology companies over the next 12 to 24 months, although large cash buffers and strong underlying businesses provide a cushion.
For investors, that creates a market where the winners may not simply be the companies spending the most, but those with the strongest ability to finance that spending and eventually monetise it.
The divide is particularly clear between United States and Chinese hyperscalers.
Spending gap
While China’s technology giants are accelerating their AI investments and narrowing the gap in computing capacity through lower infrastructure costs, government support and cheaper open-source models, US peers continue to hold a sizeable advantage in both funding capacity and access to leading-edge chips.
Moody’s says US hyperscalers retain a “durable lead in compute and monetisation”, while China peers are narrowing the gap by combining lower costs with policy support and open-source technology.
That gap is reflected in spending. The six US hyperscalers are projected to spend more than US$785bil on capital expenditure (capex) in 2026, roughly six times the combined spending expected from their leading Chinese counterparts. US spending is then expected to approach US$1 trillion in 2027.
China’s technology companies are no slouches, however.
Their combined capital spending is expected to rise to about US$140bil in 2026 and US$165bil in 2027, compared with US$65bil in 2025.
Alibaba Group Holding Ltd, for example, announces plans to invest 380 billion yuan, or about US$57bil, over three years.
ByteDance is planning capital spending of as much as US$70bil in 2026, according to a May 2026 Bloomberg report.
The spending race matters because it determines the scale of the infrastructure available to run AI models.
US data centre (DC) capacity stands at 52GW at end-2025, compared with 28GW in China, according to the International Energy Agency (IEA).
By 2030, US capacity is forecast to reach 100GW, while China’s rises to 67GW.
Key battleground
China is growing faster, with a projected 19% compound annual growth rate, against 14% for the United States. Yet the absolute capacity gap is still expected to widen.
For investors, this suggests the United States retains an important structural advantage as AI moves from experimentation towards widespread commercial deployment.
Access to Nvidia Corp’s leading-edge chips remains a particular strength, with restrictions limiting China’s ability to obtain the same hardware.
China is attempting to compensate elsewhere. Lower land, construction and operating costs, cheaper electricity and government support reduce the cost of building AI infrastructure.
Beijing’s AI+ initiative and Six Networks framework are also encouraging investment, while state-owned enterprises are becoming increasingly involved in developing AI DCs.
That makes China an interesting investment market in its own right. Rather than matching US spending dollar for dollar, China companies are pursuing a lower-cost and more efficient approach, helped by domestic AI models such as Qwen, ERNIE, Kimi and DeepSeek.
Moody’s notes that open-source models are helping to lower deployment costs, while China’s models are increasingly capable of handling commercial applications.
This could support faster adoption even if China remains behind the United States in overall computing capacity.
Cloud computing remains the key battleground for monetisation. US hyperscalers have an advantage because their cloud businesses are larger, more mature and deeply embedded with enterprise customers.
Contracted cloud spending also provides visibility as businesses move AI workloads from pilot projects into production.
Amazon.com Inc’s Amazon Web Services, Microsoft Corp’s Azure, Alphabet Inc’s Google Cloud and Meta Platforms Inc are already seeing meaningful growth from AI-related services and applications.
The strength of these businesses gives US technology giants another investment advantage: they can use profits from existing operations to finance the AI build-out.
Alphabet and Meta continue to benefit from digital advertising, Microsoft from software and enterprise services, while Amazon combines retail with cloud and advertising.
Moody’s says these businesses generate strong operating cash flow that helps fund AI capital expenditure, keeping leverage relatively modest even as spending rises sharply.
China’s companies face a tougher equation. Alibaba’s eCommerce operation is dealing with slower gross merchandise value growth and intense competition, while Baidu Inc’s online marketing revenue is declining.
Tencent Holdings Ltd has a more resilient core business, supported by gaming, advertising and finance technology, but rising AI research and infrastructure costs could weigh on margins. That makes the ability to finance AI investment an increasingly important part of the investment case.
Moody’s expects free cash flow to weaken materially across both markets in 2026 and 2027 as capex grows faster than operating cash flow. Some companies are likely to turn free cash flow negative and increasingly rely on debt markets to finance AI infrastructure.
The pressure is particularly notable in the United States, where DC lease commitments for the six hyperscalers total about US$1.2 trillion, including more than US$820bil in leases that have yet to commence.
Leasing reduces upfront capital requirements but creates longer-term obligations. Moody’s therefore expects a “material increase in adjusted debt and lease-related cash outflows” in the coming years.
For investors, that does not necessarily signal a deterioration in credit quality.
The major US hyperscalers continue to benefit from substantial liquidity, strong earnings and access to deep investment-grade debt and equity markets.
China’s companies retain funding options too, including offshore dollar bonds, offshore yuan bonds and increasingly supportive domestic bank financing. Alibaba announced on Aug 23 a proposed HK$80bil, or US$10.2bil, share placement, with all proceeds earmarked for its full-stack AI capabilities.
The broader investment opportunity therefore extends beyond the companies developing AI models.
DCs, cloud infrastructure, semiconductors, power supply and related equipment all stand to benefit from the build-out.
But the next phase is likely to reward scale, funding flexibility and monetisation rather than spending alone. As Moody’s puts it, cloud remains the primary AI monetisation channel, but scale and pricing diverge.
That divergence could become increasingly important as investors assess whether today’s enormous AI capital expenditure translates into tomorrow’s recurring revenue.
For now, the financial cushion remains considerable.
US hyperscalers collectively hold several hundred billion dollars in cash and marketable securities, while their counterparts in China also maintain sizeable net cash positions.
Those buffers give both groups room to keep investing through the next 12 to 24 months.
The bigger question for investors is how efficiently that capital is converted into computing capacity, customer demand and ultimately cash flow as the AI race enters a more expensive and competitive phase.