Citi said consensus estimates continue to undershoot the revenue and profit potential associated with AI-driven demand, pointing to a number of data points within the hyperscaler ecosystem that support a stronger outlook.
The bank highlighted that hyperscaler cloud revenue growth accelerated to 48% in the second quarter from 39% in the first quarter. At the same time, the collective backlog acceleration increased to 151% versus 143% in the prior quarter, a measure Citi interprets as consistent with ongoing acceleration in enterprise adoption of cloud and AI infrastructure.
On returns, Citi noted that Cash Returns on Cash Invested over the last 12 months remained relatively steady at 28%, compared with 29% in the first quarter. Across the hyperscalers, Cash Returns on Cash Invested was 28% in the second quarter, effectively flat with the first quarter despite an inventory build at Alphabet tied to TPU sales.
According to Citi, returns are being supported by higher compute prices driven by strong AI demand. That dynamic, the bank said, is contributing to accelerating revenue and backlog growth while also supporting margin expansion. Hyperscaler EBITDA margins held at 43% in the second quarter, unchanged from the first quarter.
Quantifying the disconnect with market expectations, Citi estimates that consensus is underestimating revenue growth by 500 basis points for 2027 and by 900 basis points for 2028, after accounting for the acceleration in capital expenditures beginning in 2024 and continuing thereafter.
Capital spending expectations for the sector have moved higher following results. Citi said consensus raised combined 2026 capital expenditure forecasts by 0.8% after earnings, taking the aggregate 2026 capex forecast to $730 billion, which the bank noted represents a 77% increase versus 2025.
Individual company guidance changes were highlighted as evidence of the broader spending shift. Meta revised its guidance range to $130 billion to $145 billion from a prior outlook of $125 billion to $135 billion. Google increased its range, with the midpoint rising by about 7% to a band of $190 billion to $205 billion, a move Google attributed to "an acceleration in delivery of capacity to meet growing demand." Amazon’s outlook was lifted by $20 billion to roughly $220 billion, a revision the company tied to higher memory costs.
Citi also pointed to cloud service providers’ revenue and backlog trajectories as supportive of stronger AI-related enterprise spending. Google Cloud Platform in particular led with an almost 20-point quarter-over-quarter growth acceleration, and the combined backlog for Google, Microsoft and Amazon expanded to $1.69 trillion in the second quarter from $1.45 trillion in the first quarter.
What this means
- Data points across revenue growth, backlog expansion and sustained cash returns suggest the hyperscaler ecosystem is scaling faster than consensus expects.
- Higher compute pricing and increased capacity delivery are contributing to revenue acceleration and preserved margins at scale.
- Marked increases in projected capital expenditures reflect continued infrastructure intensity and commitment to AI-related capacity build-outs.