Revenue growth from cloud computing has emerged as the main measure of whether major US technology companies are turning heavy artificial intelligence (AI) investments into returns, as investors increasingly scrutinize cash flow and the cost of expanding infrastructure, BMI Country Risk and Industry Research said on last Friday.

Microsoft’s Azure, Google Cloud and Amazon Web Services (AWS) recorded year-on-year revenue growth of 48 percent, 103 percent and 39.3 percent, respectively, in the third quarter of 2026, highlighting strong demand for computing capacity and AI services, the research house said in a note.

The results suggest that investors are placing greater emphasis on revenue generated from AI infrastructure rather than capital expenditure (capex) commitments alone.

BMI expects this trend to continue into 2027, as computing capacity remains a greater constraint on growth than customer demand. However, it cautioned that the benefits of the AI investment cycle are becoming unevenly distributed, with investors likely to favor companies that can convert spending into cloud revenue and positive free cash flow (FCF).

The divergence is already evident in the financial performance of the major technology groups. While cloud growth has strengthened at Microsoft, Alphabet and Amazon, the cost of building data centers and procuring components has put pressure on cash generation at several companies.

Rising memory prices and large-scale reservations for computing capacity also suggest that supply constraints are shifting from advanced processors to the physical infrastructure and components needed to support them. BMI expects access to electricity and component procurement to become increasingly important in determining which AI expansion plans can be delivered.

Cloud businesses show stronger monetization

Google Cloud delivered the strongest growth among the three major cloud providers, with revenue rising 103 percent year on year to $30.7 billion in the third quarter. Its operating margin also improved to 36.2 percent.

Google Cloud Platform accounted for most of the revenue, while sales of the company’s tensor processing unit (TPU) systems made a meaningful contribution. These systems combine hardware, computing, storage, networking and software capabilities rather than selling chips alone.

Alphabet’s customer base has also expanded beyond AI developers and research laboratories to include financial services companies such as Citadel Securities and Jump Trading. Its cloud infrastructure also serves AI companies, including Anthropic.

Microsoft Azure revenue grew 48% year on year, while operating margins were about 50 percent. The company expects Azure growth of 49 percent to 50 percent in constant-currency terms towards the end of the year, supported by improved efficiency across its central processing unit and graphics processing unit fleets.

Microsoft’s AI monetization is also extending beyond cloud infrastructure. Its Foundry revenue more than tripled year on year, while daily users of its Copilot products increased 50 percent quarter on quarter and more than fivefold year on year.

The company is also shifting towards a hybrid pricing model that combines per-seat software licenses with consumption-based charges for AI and agentic services.

AWS revenue increased 39.3% year on year to $46 billion, marking its sixth consecutive quarter of accelerating growth. Amazon chief executive Andy Jassy indicated that AWS could achieve growth rates above 40 percent in both 2027 and 2028.

AI inference, which involves using trained models to generate responses and perform tasks, now accounts for well over half of AWS’s AI revenue. Its Bedrock platform, which allows customers to build generative AI applications using different foundation models, recorded nearly 17 times growth in token volume year on year.

Amazon also said tens of thousands of customers were using its AgentCore platform to deploy AI agents, with the number of agents deployed in the third quarter exceeding the combined total for the previous three quarters.

Heavy investment weighs on cash flow

Despite the strong cloud growth, the scale of infrastructure spending remains a key concern.

Alphabet spent $51.4 billion on capex in the third quarter, equivalent to about 40 percent of quarterly revenue. The company expects full-year 2026 capex of $195 billion to $205 billion and has indicated that spending will increase significantly in 2027.

BMI estimates Alphabet’s capex could reach $300 billion next year. About 65 percent of its third-quarter investment went towards servers and networking equipment, which generally have shorter useful lives than data center buildings. This could result in depreciation costs rising more quickly as the new infrastructure enters service.

Alphabet reported negative free cash flow (FCF) of $9.1 billion in the quarter, despite operating cash flow of $42.3 billion. Its cash outflow was also affected by the accumulation of inventory for TPU systems being prepared for customers.

The company had more than $561 billion in remaining performance obligations (RPOs), or contracted revenue yet to be recognized. More than 55 percent of this backlog is expected to be recognized within 24 months, providing visibility into future demand. However, BMI noted that the timing of deliveries and payments could still weigh on near-term cash flow.

Amazon’s cash capex reached $75.9 billion in the third quarter, up 122 percent year on year. BMI estimates its full-year 2026 capex at $220 billion, rising to about $275 billion in 2027.

Its FCF deficit widened to $34.9 billion in the quarter, reflecting the acceleration in infrastructure spending. Nevertheless, operating income rose 61 percent year on year, while its operating margin reached a record 13.9 percent.

Meta Platforms also recorded negative FCF for the first time since 2012, at $2.1 billion. The company spent $34.9 billion on capex in the quarter and raised the lower end of its 2026 capex guidance to $135 billion, from $130 billion previously.

Meta’s reported operating income fell 37 percent year on year to $13 billion, largely due to a $9.9 billion legal settlement charge. Excluding the charge, operating income would have increased 11 percent.

Microsoft, in contrast, generated FCF of $18.6 billion. Its capex of US$45.2 billion was below market expectations, partly because of delays in capacity deliveries from third-party cloud infrastructure providers. BMI expects spending to pick up towards the end of the year and into early 2027 as those projects come on stream.

Demand remains strong, but capacity is tight

The major cloud providers continue to report substantial contracted demand, although the conversion of backlogs into revenue is taking longer than some investors had anticipated.

Microsoft’s RPO reached $677 billion, up 73% year on year, while Amazon’s backlog increased by more than 150 percent. Amazon said customers were reserving large blocks of computing capacity, sometimes reaching 1 gigawatt, well in advance, providing visibility into demand through 2027 and 2028.

Meta does not have a comparable contracted cloud backlog, but is seeking to generate more revenue from its computing infrastructure through AI services and its Muse personal agent.

More than 10 million users in the US and Canada had set up Muse within about six weeks of its launch. Shopping was its most common use case, with more than a quarter of weekly active users engaging in shopping-related activities. The average purchase size exceeded $200.

However, BMI said Muse’s monetization remained at an early stage. Meta currently relies on subscriptions, while potential future revenue streams include transaction fees, business subscriptions, enterprise services and advertising.

The report also highlighted growing uncertainty over the financing of AI infrastructure. CoreWeave’s 2032 senior unsecured bond yield rose above 15.5 percent, with interest payments equivalent to about a quarter of its revenue. The company plans to increase installed capacity from 1.5 gigawatts to 8 gigawatts by 2030, supported by cloud backlogs exceeding $104 billion.

BMI views the development as evidence that bond investors are becoming more selective about companies with heavy borrowing requirements and less transparent cash flows.

Looking ahead, the report expects the market to increasingly distinguish between technology companies that can demonstrate AI-driven revenue growth and those whose capex, depreciation and financing costs are rising faster than monetization.

It said a recovery in FCF would be a positive signal for the technology sector if driven by stronger revenue and operating efficiency rather than cuts to infrastructure investment. This would indicate that the AI investment cycle remains intact, while allowing cloud providers and infrastructure suppliers to capture continued demand.

However, access to power, memory and other critical components could determine how quickly companies can expand capacity and translate their AI ambitions into financial returns.

EV supply chains dominate APAC automotive investment as focus shifts to components