Doug Anmuth

Dara acknowledged intense Brazil food-delivery competition raising 2-wheeler supply costs, but said Uber holds share and the low-margin business limits the P&L impact.

Dara reaffirmed Waymo as an important partner in Austin and Atlanta but stressed Uber won't depend on one partner given a growing multi-partner AV ecosystem.

Balaji detailed insurance renewal improvements, risk offloading to third parties, and resulting price reductions driving trip elasticity. L.A. trip growth trends significantly better than California and rest of country.

Dara cited Reserve's success as proof of on-demand to planned transition, noted 700,000 hotels via Expedia partnership with 10% Uber credits for Uber One members.

Dara described NVIDIA partnership to build a real-world data collection factory targeting over 3 million hours of rideshare-specific data, combined with advanced simulation to democratize AV development.

Dara outlined requirements: partner deals, vehicle capital, depot acquisition, charging infrastructure, government relations for regulatory clearance, and safety case development.

Dara described Nvidia's Hyperion as reference architecture for L4, Stellantis deal for 5,000 initial vehicles, with fleet ownership expected to be financialized over time.

Dara outlined cross-platform programs including top tabs, personalized upsells, and membership, noting cross-platform consumers spend 3x more with 35% better retention.

Dara described both in-session conversion benefits and delayed frequency effects from lower prices, with July trends showing acceleration and profit per ride still up year-over-year.

Prashanth explained the $20B is additive to ~$3B remaining, representing 12% of market cap. Share count already reduced 1% in Q2 with multi-year active execution planned.

Dara highlighted very high AV utilization, strong opt-in rates, healthy ratings, and plans to increase vehicle count in Austin with Atlanta expansion coming.

Dara noted favorable short-term elasticity from passing along lower insurance costs, with some mix shift toward international trips and travel sector affecting overall price.

Prashanth detailed three insurance improvement drivers: tech-driven driver safety dashboards, stable carrier pricing with captive insurance leverage, and state-level regulatory reform.

Dara noted AV deployments are still very small scale but the Uber network drives significantly higher utilization, with customers showing higher opt-in rates on repeat rides.

Li gave a detailed multi-part roadmap covering personalization, data infrastructure, sequence length, and LLM content understanding.

Zuckerberg framed it as a monetize-now versus build-future-assets trade-off rather than answering the training-versus-inference split.

Li declined to comment on the Manus acquisition status, offering no details on strategic rationale or timeline.

Li detailed how the adaptive ranking model uses 1 trillion parameter complexity with sub-second speed, routing to compute-intensive inference when conversion probability is higher.

Susan Li affirmed cash generation can support 2026 infra capex but declined to commit on FCF or new JVs.

Zuckerberg declined to commit to specific model or timing, framing the first releases as trajectory indicators rather than a single moment.

Susan Li declined to provide earnings/FCF or cash targets, emphasizing compute-capacity flexibility for 2026-2028 as the overriding priority.

Zuckerberg signaled a meaningful shift, Meta may not open source its largest frontier models for competitive and safety reasons.

Susan Li acknowledged exploring external financing partnerships for data center co-development, a significant shift vs prior quarter's stance.

Susan Li unambiguously stated Meta is funding Llama training infra itself, partnerships limited to hosting via AWS/Azure.

Susan Li acknowledged both drivers but did not break out the split, citing trade-related uncertainty.

Susan Li confirmed all three CapEx components grow in 2025, with servers still the biggest driver and core AI/non-AI dominant.

Zuckerberg acknowledged DeepSeek advances are being studied but defended heavy CapEx as a sustained strategic advantage.

Sundar acknowledged agentic coding as an area to improve but expressed confidence in staying at the frontier, citing the Gemini 4 pretraining run.

Anat: fund first from operating cash flow, then debt (expanded from ~$16B to ~$100B), with the equity raise aimed at a resilient balance sheet.

Ashkenazi confirmed 2027 CapEx will increase to meet robust demand but deferred specific figures to future earnings calls.

Schindler confirmed AI overviews and AI Mode are driving query growth and signaled upside in the 20% ad coverage rate due to AI-powered intent understanding.

Pichai said improving models across pre-training post-training test time compute with multimodal and agentic capabilities. Coding showing lot of progress. Integrating all together for products and Cloud APIs with lot of headroom ahead. Trajectory over past two years shows relentless innovation cadence confident about maintaining momentum. On TPUs said wide choice of accelerators makes Google Cloud attractive meeting customers where their needs are with end to end data center efficiencies driving overall momentum.

Schindler said Q3 Search growth across all major verticals led by retail and financial services. AI Overviews and AI Mode driving overall and commercial query growth. AI Overviews monetizing at approximately same rate providing baseline for richer placements over time. Paid clicks up 7% and CPCs up 7% year over year. Ashkenazi said productivity is ongoing not one-time including moderating headcount growth optimizing real estate leveraging AI internally with nearly half of all code generated by AI. Shorts lower revenue share helping gross margins.

Pichai said strong momentum especially in cloud with tight supply environment. Additional investment will play out in future years so both tight supply and increased investment are true at same time. Comprehensiveness of AI portfolio on GPUs and TPUs driving demand. Schindler said they manage business to drive great outcomes for users and attractive ROI for advertisers not to paid click and CPC targets. Paid clicks up 4% year over year in 10-Q but many factors affect metrics quarter to quarter.

Schindler confirmed AI overviews monetizing at approximately same rate with ads launched within overviews on mobile in US. Declined to detail click-through rates and conversion. Ashkenazi said efficiency approach unchanged focused on CapEx efficiency operating expense moderation real estate optimization and AI-first internal processes. $75 billion CapEx subject to rigorous demand-based allocation process.

Schindler confirmed AI overviews driving higher satisfaction and search usage with ads launched within overviews on mobile in US. Monetization at approximately same rate providing strong base for further innovation. Ashkenazi confirmed cloud capacity constrained in Q4 with demand exceeding available capacity. Lapping strong Q4 2023 deployment contributed to deceleration. Working to bring more capacity online through increased CapEx.

Alex declined ARPU guidance and launch timing but explained add-ons (e.g., Audiobooks+ doubling) structurally lift ARPU beyond price increases.

Christian clarified the EUR 200M full-year opex step-up is marketing and AI (not structural/headcount), with growth ex-currency/social-charges consistent Q2 to Q3 and moderating in Q4.

Gustav said early indications show a structural ARPU increase but cautioned it is too early to share specifics.

Gustav framed the opportunity clearly: solving copyright/attribution so existing creators can participate in AI via derivatives of existing IP.

Christian named specific drivers, pricing outpacing net content cost, improving ads, marketplace gains, and vertical/geographic expansion.

Daniel said the Apple case lifted U.S. conversions and enables direct communication plus a third a-la-carte monetization driver alongside subscriptions and ads.

Alex tied MAU to engagement/product improvement, attributed Q1 softness to the Q4 Wrapped pull-forward, and reaffirmed 2025 net adds in the four-year range weighted to H2.

Daniel framed 2025 as accelerated execution, doubling down on music with faster product velocity and discipline.

Olsavsky quantified the 650bps expansion and the 520bps ex-derivative figure, named the drivers (efficiency, capacity optimization, fixed cost control), and cautioned margins will fluctuate.

Jassy answered both sides: no frontier model is required for success, but Amazon is pursuing one for cost control, prioritization and speed.

Jassy said Project Rainier has gone very well and Anthropic is quite pleased. Trainium has 30% to 40% better price performance than comparable GPUs. Trainium is multibillion dollar ARR business and fully subscribed. Trainium3 just started shipping with 40% more price performance and expects nearly all supply committed by mid-2026. Already building Trainium4 for 2027 and having conversations about Trainium5. Combined Graviton and Trainium well over $10 billion ARR. Said this is an extraordinarily unusual opportunity to forever change the size of AWS and Amazon and will invest aggressively to be leaders.

Jassy said building a cluster with 500,000 plus chips going to 1 million is an infrastructure feat that is hard to do at scale. Differentiator is combination of AWS infrastructure capabilities built over long period plus Trainium chip performance and price performance. Many other customers interested in large Trainium clusters which will be accommodated with Trainium3.

Jassy said uncertainty remains on where tariffs settle especially China and who absorbs costs when prebuys deplete. Through first half no diminished demand or broad ASP increases seen. On AWS competitive growth gap said year over year percentages are function of base size with AWS meaningfully larger. Cited security differentiation and recent incidents at competitors. Said AWS has more demand than capacity and could be doing more revenue.

Olsavsky cited software optimization of server capacity, efficient custom networking gear, maximizing power usage in existing data centers, and Graviton custom silicon. Warned margins impacted by investment levels, competitive pricing, and GenAI service mix. Jassy described Alexa Plus with 100,000+ users showing very positive response. Can take complex actions like reservations, home automation.

Olsavsky said AI comes with lower initial margins and heavy investment but over long term margins should be comparable to non-AI business. On UPS, Jassy said UPS decided serving Amazon is lower margin for them and walked away from some volume. Amazon able to handle it with own logistics network.

Chesky said the strategy is emerging and consumer-first (Apple analogy), ruled out a white-label business, and framed enterprise as important but downstream of consumer demand.

Brian said RNPL is still early innings with desktop and merchandising expansion ahead. A full payments and pricing roadmap could deliver hundreds of millions in revenue each year.

Ellie cited three drivers for the raise: stronger nights forecast, durable higher ADRs, and monetization uplift from fee structure and insurance lifting take rate.

Brian said integration wasn't ready, needed account integration, messaging, personalization. Four principles: custom integrations, not appearing as commodity, not being a data layer, showing uniqueness of Airbnb. Declined Google Hotel Finder for same reason. Open to future integrations when well-developed.

Brian described services as 10 different businesses, testing attach rate in LA. 10% of service bookings from locals. Events build supply (Paris 50%+ home increase), policymaker relationships, brand trust.

Brian detailed co-host quality metrics and expansion to 15K co-hosts. Next: Japan and Korea. Ellie pointed to variable cost efficiencies in payments and customer service.