Big Tech’s Multi-Billion AI Infrastructure Bet Has an Achilles’ Heel

The debate over Big Tech’s AI spending may finally be over.

Microsoft, Google, and Amazon all reported stronger-than-expected cloud growth, driven by enterprise demand for AI infrastructure, models, and software. More importantly, executives across the three companies indicated that demand still exceeds available capacity, suggesting the AI investment cycle is far from over.

Collectively, Microsoft, Google, and Amazon now expect to spend around $550–615 billion in capital expenditure this year, with much of that investment directed towards expanding AI infrastructure and cloud capacity.

The latest results also suggest that AI is becoming a meaningful contributor to cloud revenue. While investors have expressed concerns that AI spending would take years to generate returns, the latest quarter indicates those investments are beginning to pay off.

That optimism was also reflected in the stock market. Amazon crossed the $3 trillion market capitalisation mark for the first time, becoming only the fifth publicly traded company to reach the milestone after NVIDIA, Microsoft, Apple, and Alphabet.

But building AI infrastructure is only one part of the challenge.

“The build-out is easy. The question is utilisation. There is not enough discussion about how much of the workload will be in-house, or organic, versus external customer demand, which can be more volatile,” Holger Mueller, VP and Principal Analyst at Constellation Research, tells AIM.

He believes that we are in a gold rush. “The big potential of enterprise AI has not even started to move to cloud at scale. That will be the ultimate workload to win,” he says.

While hyperscalers continue to argue that AI demand is outpacing available capacity, Ashish Banerjee, Senior Principal Analyst at Gartner, believes the industry is still in the early stages of a prolonged infrastructure buildout.

“The AI infrastructure cycle remains in an aggressive expansion phase, where physical supply is scrambling to keep pace with demand,” Banerjee says. “Crucially, demand is no longer just driven by foundation model training; it is increasingly powered by production inference as enterprises embed AI into daily workflows.”

However, he cautions that strong demand today does not automatically guarantee long-term returns. “Capacity constraints should not be confused with guaranteed long-term ROI,” he says, noting that data centres and power grid connections typically take two to four years to deploy. 

“Hyperscalers must commit capital well ahead of full multi-year demand visibility. Under-investing risks losing market share, so hyperscalers are building ahead of the curve.”

Microsoft Strengthens Lead

Microsoft delivered one of its strongest quarters in recent years. Azure stood out, growing 43%, comfortably ahead of market expectations and its annual revenue surpassing $100 billion. 

Microsoft continues to ramp up its AI infrastructure investments. It reported $41 billion in capital expenditure during the quarter ended June, with roughly two-thirds directed towards short-lived assets such as CPUs and GPUs, reflecting growing customer demand for both AI and traditional cloud workloads.

“We continue to invest to meet customer demand,” CFO Amy Hood said during the earnings call. “Roughly two-thirds of our capex was for short-lived assets, primarily CPUs and GPUs, as customers increasingly build solutions that leverage both AI and non-AI infrastructure.”

The remaining capital was allocated to long-lived assets, including large data centre sites.

During the earnings call, Microsoft CEO Satya Nadella shared that the company has added 31 new data centres globally this quarter, bringing the total to 88 this year, as it aims to roughly double its overall capacity within two years.

“We are also bringing capacity online faster than ever. Over the last fiscal year, we have reduced dock-to-live times for new GPUs in our largest regions by nearly 50%,” he said.

Microsoft expects its AI investments to accelerate further. The company forecast more than $50 billion in capital expenditure for the upcoming quarter and projected Azure to grow around 45% in constant currency.

Microsoft now expects to invest up to $190 billion in capital expenditures for calendar year 2026, including in its own AI chips. Nadella said Maia 200 now delivers 30% better performance per dollar than the latest-generation hardware in Microsoft's fleet and supports both OpenAI and MAI models. 

Google Cloud’s Breakout Moment

If Microsoft’s quarter demonstrated consistency, Google’s showed acceleration.

Google Cloud revenue surged 82% YoY to $24.8 billion.

Demand for Gemini-powered services, AI infrastructure, and cloud platforms pushed the company’s cloud backlog to $514 billion, reflecting strong future commitments from customers. The company has raised its 2026 capital expenditure guidance to $195–205 billion while also reporting one of its strongest quarters ever.

In a statement, Google CEO Sundar Pichai said Cloud’s growth was driven by an integrated AI stack spanning chips, models, security, data platforms, and enterprise agents. He revealed nearly 500 Cloud customers have each processed more than 1 trillion AI tokens over the past year, while more than 2,000 enterprises consumed over 100 billion tokens. 

Moreover, Google is also monetising its custom AI chips beyond its own cloud business. In May, the company partnered with Blackstone to create a standalone TPU cloud, allowing enterprises to rent Google’s Tensor Processing Units as a compute-as-a-service offering.  

Mueller attributes Google Cloud’s strong momentum partly to its differentiated AI infrastructure. “Google Cloud offers the only real alternative to NVIDIA with its TPUs, which are faster and cheaper, and CxOs are noticing,” he notes.

Google’s smaller cloud business also amplifies the impact of large customer wins. “Google Cloud operates at a lower revenue level. So, for example, a $100 million win has a much bigger impact there than it would at AWS or Microsoft,” Mueller says.

However, Biswajeet Mahapatra, Principal Analyst at advisory firm Forrester, offers caution, noting that a single earnings cycle should not be interpreted as a fundamental shift in the competitive landscape.

“A faster quarterly growth rate is evidence of momentum, not proof of a structural reset,” he comments. “The challenger can narrow the relevance gap in AI-heavy workloads, but leadership will be decided by sustained enterprise adoption, ecosystem execution, governance fit, and capacity availability, not a single quarter's growth differential.”

AWS Delivers Fastest Growth in Years

AWS revenue increased 37% YoY to $42.2 billion during the quarter, its fastest growth in more than four years and significantly ahead of analyst expectations. 

“AWS is now a $169 billion annualised revenue run rate business…Our chips business now has an annual revenue run rate of over $25 billion, growing triple-digit percentages year-over-year,” CEO Andy Jassy said during the earnings call.

That confidence is reflected in Amazon’s capital spending plans. The company raised its projected 2026 capital expenditure to $220 billion. Much of that spending will go towards expanding AI data centres and cloud capacity.

The company reported $53.1 billion in cash capital expenditure during the second quarter.

“The higher cost of memory is pushing this number up from our prior estimate of about $200 billion. Even at that amount, we will still not have enough capacity to meet all the demand we have in 2026, and I believe this dynamic will also be true in 2027, too. In fact, the demand we already have for 2028 is striking,” said Jassy.

AWS also reported a contract backlog of $496 billion, suggesting enterprises continue committing billions of dollars to long-term cloud deployments.

“Customers choose AWS because we offer the broadest capabilities. They want their AI inference to reside near their other applications and data,” Jassy said. 

AI Revenue as Important as Infra

Balancing AI infrastructure build with actual financial return is becoming the defining challenge for Big Tech.

“As this massive buildout materialises over the next 24–36 months, the primary risk could shift from capacity shortages to localised overcapacity, hardware obsolescence, and margin pressure,” Banerjee observes.

Banerjee advises investors to focus on whether AI investments are translating into sustainable business outcomes rather than simply tracking capital expenditure. 

“I would watch cloud revenue growth, AI revenue, backlog conversion, capacity utilisation, and returns on invested capital,” he says. “Customer adoption matters, but usage is more important than the number of pilots or licences sold. Ultimately, durable growth requires AI workloads to move into production, generate recurring consumption and support revenue growth without permanently weakening margins or free cash flow.”

Mahapatra believes the bigger shift is happening in how enterprises are approaching cloud infrastructure. “The bigger signal is not simply that AI is lifting cloud growth, but that cloud demand is being redefined around capacity, control, and workload placement,” he says. 

“Public cloud remains central for modernisation, access to specialised services and AI workloads, but buyers are becoming more selective as cost predictability, resilience, data control, and vendor concentration become board-level concerns,” he adds.

Moreover, he notes that enterprises are increasingly rethinking their cloud strategies around “sustained consumption, financial accountability, and operational control.”

On the industry’s continued capacity constraints, Mahapatra believes hyperscalers remain in a “supply-constrained buildout phase, not a normalised consumption phase.”

“AI workloads are now becoming first-order architectural drivers,” he says. “Capacity planning has moved beyond conventional cloud elasticity into long-lived commitments around training, inference, data pipelines, model governance, power, cooling, and resilience.”

Big Tech’s massive AI infrastructure buildout is driven heavily by inference workloads. However, a significant gap remains between deployment and profitable enterprise ROI. That’s what investors will be watching next.

“The most important indicators are not only cloud revenue growth or capital expenditure guidance, but whether AI capacity converts into repeatable, governed, margin-accretive enterprise usage,” Mahapatra says.