AI inference, productivity tools offset Gen-AI margin headwinds

This article was written by Bloomberg Intelligence Senior Industry Analyst Mandeep Singh. It appeared first on the Bloomberg Terminal.

The shift to inferencing from training bodes well for greater revenue from generative AI, which will be key to companies’ operating margin in the next 12 months. Alphabet generates billions of dollars in sales from customer gen-AI workloads running on GoogleCloud. Gen-AI Overviews monetization has also offset any decline in query volume. Smaller peers like Snap and Pinterest could see gross-margin headwinds from AI compute costs amid pressured ad pricing. Uber, Airbnb, DoorDash and other online marketplaces might seek to offset margin pressure with a ramp-up in ads from promoted listings in their mobile feeds.

Inferencing mix in hyperscale capex

Hyperscale companies have seen significant positive revisions for capital spending in 2025 as they boost AI-related outlays amid growing demand for AI inferencing in all kinds of application workloads. Alphabet and Meta are developing their in-house large language models (LLMs) to support a variety of AI services across their products. Chatbot and copilot monetization has improved with subscription offerings, and as they evolve with more capabilities, could emerge as contributors to revenue.

Hyperscale capex in 2025 is set to rise 45% this year, tapering from 73% a year ago.

Smaller peers’ margins hurt by model costs

Digital-ad peers have seen negative revisions to Ebitda-margin estimates year to date, driven by higher compute costs emerging from additional AI features. The effect could be more pronounced for smaller internet companies vs. larger ones such as Google and Meta, which have trained their own LLMs for use across their families of apps. These AI capabilities may help drive top-line growth and engagement, while productivity tools might aid in improving leverage on fixed costs.

So far, Reddit has been an outlier in the digital-ad peer group with its gross-margin expansion.The company may see its high-margin data-licensing revenue grow beyond Google, with other large foundational model providers including OpenAI and Anthropic, likely to pay for its content.

Marketplace consolidation offsets margin pressure

Ride-share and delivery companies have seen positive revisions to Ebitda margin aided by scale and better operating efficiency. Lower incentives for supply and customer acquisition have also helped margin expectations for larger online marketplaces, including Uber, Airbnb and DoorDash, along with the ramp-up of advertising and subscription offerings. Still, traffic-acquisition costs could increase amid higher compute expenses and the potential disruption from the rollout of robotaxis and autonomous vehicles from Waymo and Tesla, potentially pressuring ride-sharing take rates. Recent deals including DoorDash-Deliveroo and Prosus-Just Eat Takeaway suggest more consolidation.