Cloud giants deliver, Meta's spending bill lands differently

The results, published on Wednesday, were broadly strong across the board, as first reported by City AM. Alphabet posted revenue of $109.9bn (£81.6bn), a 22 per cent rise year on year, according to the company's quarterly filing. Google Cloud was the standout, with revenues climbing 63 per cent to $20bn, comfortably clearing analyst forecasts.

Shares in Alphabet (NASDAQ: GOOGL) jumped more than six per cent in after-hours trading.

Amazon (NASDAQ: AMZN) reported overall revenues of $181.5bn, with AWS, its cloud arm, recording its strongest quarterly growth in nearly four years at 28 per cent, reaching $37.6bn. Microsoft (NASDAQ: MSFT) posted $82.9bn in sales; its Azure platform grew 40 per cent, and its AI products now generate an annualised $37bn in revenue, more than double the figure from a year ago, according to the company's earnings release.

Meta (NASDAQ: META) was the outlier. Revenue grew 33 per cent to $56.3bn, ahead of market expectations, yet shares fell more than five per cent after hours. The trigger was a raised capital expenditure guidance: Meta now expects to spend between $125bn and $145bn this year, roughly $10bn more than previously indicated, as chief executive Mark Zuckerberg outlined ambitions to build what he described as "personal superintelligence for billions of people."

Matt Britzman, senior equity analyst at Hargreaves Lansdown, noted the contrast between the headline numbers and the market reaction.

"You can't tell from the market reaction, but Meta delivered another very strong quarter, with advertising momentum clearly accelerating."

Britzman added that the spending increase was "modest relative to Meta's existing investment plans" and came against a backdrop of strong revenue growth and healthy margins, according to his published commentary.

Alphabet also raised its full-year spending outlook to as much as $190bn without triggering a comparable sell-off. The difference, according to City AM's reporting, was that Google Cloud's profitability is already moving sharply in the right direction. Meta, with no cloud business generating AI-related revenue from third parties, faces a credibility gap on capex that dates back to the Reality Labs losses of 2022 and 2023.

What $650bn in AI capex means for enterprise buyers

The combined spending figure is staggering. Roughly $650bn flowing into data centres, chips, and networking equipment in a single year will expand global AI compute capacity at a pace that has few historical parallels in enterprise technology.

For operators at UK firms procuring cloud and AI services, the immediate consequence is likely to be sustained capacity growth. Synergy Research estimated that the European cloud market expanded roughly 20 per cent over the past year, and the infrastructure being built now will underpin the next wave of that expansion.

In theory, a supply glut should favour buyers. More capacity means more competition for enterprise workloads, which should translate into competitive pricing, broader service-level commitments, and greater willingness from hyperscalers to negotiate bespoke terms with mid-market customers.

But the picture is more nuanced than simple supply-and-demand dynamics suggest. Each of the four companies is building infrastructure optimised for its own stack: Google's Tensor Processing Units, Amazon's Trainium and Graviton chips, Microsoft's tight integration with OpenAI models, and Meta's in-house Llama ecosystem. The result is a landscape where switching costs could rise even as headline prices fall.

Procurement teams at UK scale-ups should be alert to this dynamic. A generous introductory pricing tier on one platform may look attractive in the short term, but deep integration with proprietary AI tooling can create dependencies that are expensive to unwind. The capex arms race is, in part, a land-grab for long-term enterprise relationships.

Pricing pressure versus lock-in

The tension between lower unit costs and higher switching costs is not new in cloud computing, but AI workloads amplify it. Training and fine-tuning models on a specific platform's hardware and software frameworks creates technical debt that compounds over time. For a mid-market firm running inference workloads at scale, migrating from one provider to another can involve re-engineering data pipelines, retraining models, and renegotiating data residency arrangements.

Finance directors weighing multi-year cloud commitments would do well to stress-test their contracts against a scenario in which their chosen provider's AI strategy diverges from their own needs. The spending boom makes capacity abundant; it does not automatically make it interchangeable.

The payback question: who has a path to returns

Wednesday's results drew a clear line between two business models. On one side sit Alphabet, Amazon, and Microsoft, each of which can point to fast-growing cloud revenue streams that are directly monetising AI demand. Google Cloud's 63 per cent growth rate, AWS's 28 per cent expansion, and Azure's 40 per cent gain all provide visible evidence that enterprise customers are paying for AI services at scale.

On the other side sits Meta, whose AI spending must ultimately be justified through advertising revenue and consumer engagement on its family of apps. The company's 33 per cent revenue growth suggests the advertising model remains robust, but the market is clearly applying a higher burden of proof to capex that lacks a direct enterprise revenue offset.

Gil Luria, head of technology research at DA Davidson, described the simultaneous reporting as the "biggest earnings day ever," according to City AM's reporting, and the divergent reactions bore that out.

The distinction matters beyond share prices. Cloud providers whose AI infrastructure is paid for, in part, by enterprise customers have a self-reinforcing loop: more spending attracts more workloads, which fund more spending. Meta's loop runs through consumer attention and advertiser willingness to pay for AI-enhanced targeting, a path that is plausible but less directly observable in quarterly filings.

For enterprise buyers, the structural difference has a practical implication. The cloud platforms subsidising their AI buildout with enterprise revenue have strong incentives to keep expanding services, improving performance, and competing on price to retain and grow their customer base. That competitive dynamic is likely to persist for as long as the capex cycle continues.

What UK operators should watch next

Several developments in the coming quarters will shape the environment for UK firms buying AI and cloud services.

First, pricing signals. As new data centre capacity comes online through 2026 and into 2027, watch for changes to reserved-instance pricing, spot-compute rates, and committed-use discount structures. A meaningful drop in unit costs would confirm that the supply expansion is translating into buyer-friendly economics.

Second, regulatory alignment. UK and EU data residency requirements continue to evolve. The hyperscalers' European expansion plans, funded by the capex wave, will determine whether UK firms can run AI workloads domestically or must accept cross-border data flows. Board members should track announcements on UK-based data centre builds from all four providers.

Third, model portability. The degree to which open-weight models, such as Meta's Llama family, gain traction in enterprise settings will influence how much bargaining power buyers retain. If proprietary models dominate, lock-in risks increase. If open alternatives prove competitive, procurement teams gain negotiating leverage.

Finally, the durability of cloud growth rates. AWS at 28 per cent, Azure at 40 per cent, and Google Cloud at 63 per cent are exceptional figures for businesses of this scale. Any deceleration would signal that the market is absorbing capacity more slowly than the builders anticipated, potentially accelerating the shift towards buyer-friendly terms.

The $650bn capex wave is, above all, a bet on sustained enterprise demand for AI compute. UK operators sit on the demand side of that equation. Understanding which providers have the clearest path to returns, and which are building on less certain foundations, is now a core part of vendor strategy.