Brian Nowak

Zuckerberg walked through the portfolio of opportunities but declined to rank which scales first, deferring specifics to a later date.

Li explicitly declined to quantify 2027 CapEx but explained the planning framework of maximizing 2026-2027 capacity.

Zuckerberg gave a high-level framework (build leading models → scale products → monetize) but acknowledged no precise monthly plan for how each product scales.

Zuckerberg flagged subscriptions, advertising, shopping and the Manus acquisition as monetization levers but said detail will follow as models roll out.

Susan Li credited ad-ranking, ad-load redistribution, Advantage+ rollouts, and a 4pt FX tailwind as the Q1 2026 drivers.

Susan Li pointed to value-weighted conversion growth outpacing impressions and new model architectures enabling more data/compute for ads performance.

Susan Li declined to quantify but cited lapping Quest 3S and Q3 retail-partner procurement pull-in as the drivers.

Zuckerberg framed self-improvement as a key research thrust but declined to detail specific gating factors.

Susan Li walked through near-term items (adaptive recs, creator break-out matching) and longer-term LLM-in-recs bets.

Susan Li gave rich color on top use cases (info gathering, social interactions, writing assist, visual) and entry-point mix dominated by WhatsApp then Facebook.

Zuckerberg gave substantive detail on 17B-per-expert architecture, voice latency optimization, long context windows, and Behemoth-based distillation.

Susan Li gave a concrete timeline: MTIA ramping in 2025 for inference, expanding to training in 2026.

Zuckerberg gave broad vision on Meta AI + Llama 4 but deferred specifics; Susan Li provided clearer detail on MTIA custom silicon roadmap.

Dara named concrete milestones: live in 7 cities, on track for 15 by year-end, specific partner launches, plus service quality and per-vehicle utilization. Direct.

Dara attributed growth to structurally better membership benefits than competitors, global benefit expansion, no-fee grocery above $60, member days events, and 10% hotel credits.

Dara described a global playbook of expanding selection and reliability in sparse markets. Australia sparse markets 2x larger than average, with lots of growth remaining.

Balaji affirmed continued aggressive buybacks at a steady cadence, noting the company can simultaneously invest in growth and return capital given ~$10B free cash flow generation.

Dara argued hardware is fundamentally different from software platforms, with many OEMs and AV software players reaching the finish line. Uber is the winner-take-most 3P platform.

Dara emphasized organic category leadership in UK and France, gaining share in Spain and Germany. Highlighted partnerships with OpenTable, Instacart, and Ifood as competitive differentiators.

Prashanth said sparse markets growing 1.5x denser markets, roughly 20% penetrated. Wait-and-save product is an excellent match for suburban environments with lower car density.

Dara confirmed 99th-percentile Waymo utilization in both Austin and Atlanta, noted a positive halo effect on the overall system, but didn't update specific ride volume figures.

Dara attributed audience growth to product diversification, Moto ($1.5B, growing 40%), premium ($10B, growing 35%), reserve (growing 60%), plus teens and elder products expanding the addressable base.

Dara deflected on specific fleet size targets, focusing on quality and safety over incremental volume, noting the fleet will be over 100 vehicles soon.

Prashanth confirmed 19% year-over-year trip growth for three consecutive quarters, heavily led by audience growth, with sparser markets at 20% of trips and growing faster.

Dara provided a comprehensive overview of AV investment strategy across fleet depots, technical partnerships, and OEM supply, noting costs won't impact the three-year outlook.

Prashanth explained mobility profitability should be viewed at total company level, with continued investment in growth areas like membership and Uber Teens.

Jassy confirmed active conversations and said there is a real chance Amazon sells Trainium outside its cloud, without giving timing or return economics.

Jassy pointed to reserved 2027 and 2028 capacity and the barbell adoption curve, but did not commit to any deceleration in data center spend.

Jassy disclosed $364B backlog excluding the Anthropic deal worth over $100B, and confirmed reasonable breadth across multiple customers, not just one or two.

Jassy outlined vision for Rufus as best shopping assistant, noting 115% MAU growth and 400% engagement growth. Compared current agentic commerce to early search engine days.

Jassy said investment areas include selection expansion on both luxury brands and everyday essentials with one out of three units now everyday essentials. Speed of delivery improvements plus Quick Commerce with Amazon Now in India Mexico UAE showing triple shopping frequency. Perishables in thousands of cities with nine out of ten top items being perishables where available. On efficiencies cited regionalization extended from 8 to 10 regions improving inbound delivery and inventory placement. Making significant progress getting more units into each box. Robotics with over 1 million robots still a fraction of future potential.

Jassy said Amazon will always offer multiple chip options. Deep NVIDIA relationship will continue with significant purchases not constrained in buying. Annapurna chip team built Graviton with 40% better CPU price performance and doing same with Trainium on AI side. Expects Trainium3 to be about 40% better than Trainium2. Must deliver chip in volumes quickly and continue improving software ecosystem. Project Rainier proof point with Anthropic builds increasing credibility.

Jassy said AI market very top heavy with small number of large frontier models and large-scale applications like chatbots and coding agents like Cursor Vercel and Lovable running on AWS. Very large number of GenAI applications still in pilot or development. Cited Trainium2 with 30% to 40% better price performance as backbone for Anthropic Claude models and Bedrock. Highlighted SageMaker AI Bedrock Strands and AgentCore as differentiated middle layer. Said 85% to 90% of IT spend still on premises and GenAI inference will be another building block running close to existing data and infrastructure on AWS. Optimistic about combination of cloud migration AI acceleration and more capacity coming online.

Jassy focused on helping sellers through uncertainty, maintaining seller diversity and low prices, and managing inventory inflow to avoid fulfillment network congestion. Olsavsky said Q2 tariff costs not large given Q1 pre-buying but wider guidance range reflects general uncertainty. Advertising strong and some strength in April.

Jassy said Shreveport next-gen robotics showing speed, productivity, and cost to serve improvements. Plans to expand to new and retrofitted facilities. About a thousand generative AI applications built or in process across Amazon. Rufus shopping assistant, AI chatbot with 500 basis points better satisfaction, and 10% better inventory forecasting cited as examples.

Anat: takes a multiyear view and will invest as long as returns are attractive; still supply constrained with demand outpacing investment.

Sundar framed GenAI returns as very early innings with extraordinary opportunity but offered no quantified ROIC figure or timing.

Pichai outlined agentic consumer workflows as the key opportunity for search compute investment, noting the company is in early innings.

Pichai framed TPU sales through a customer-service lens rather than disclosing specific pricing strategy, citing capital markets and AI labs as target buyers.

Pichai said 2025 was about laying foundation getting models robust for agentic use cases with coding showing most progress. Universal Commerce Protocol launched at NRF with founding partners for interoperability. Now integrating into Gemini and AI Mode. 2026 is the year consumers will use agentic commerce. On YouTube said Genie will have wide applicability with multimodality and real world representation. All innovation from Imagine Veo Liria Genie brought to products and cloud customers. YouTube will incorporate tools with creators at center of experience.

Schindler said agentic experiences are additive helping people get stuff done and helping businesses. Working on agentic experiences across travel commerce and shopping verticals with attention to partner ecosystems. Introduced agentic checkout in AIO and partnership with PayPal. On Waymo Pichai said exciting opportunity to integrate Gemini multimodal experience and YouTube for dramatically better in-car experience. Will see newer experiences in 2026.

Pichai said 2.5 series models investing most in agentic direction with exciting progress including unreleased models. Main challenges are chaining sequences reliably with compounding latency and cost. Making robust progress at frontier with 12-month efficiency improvements on models. Expects 2026 to be breakout year for agentic experiences. Internally rolling out agentic coding workflows for software engineers with increased adoption over last few weeks overcoming earlier friction points.

Schindler said Q1 saw broad-based strength led by finance insurance retail healthcare and travel. Too early to comment on Q2 being only a few weeks in. De minimis exemption change will cause slight headwind from APAC retailers. Lots of experience managing uncertain times. AI Overviews continuing to drive higher satisfaction and usage with Q1 being largest expansion to date.

Pichai said AI overviews playing out positively with great metrics. More powerful multimodal AI coming including Lens Circle to Search and Project Astra capabilities. Deep Research expanding types of use cases for Search. Ashkenazi outlined efficiency priorities including technical infrastructure optimization managing headcount growth real estate footprint optimization organization simplification and using AI tools internally for coding and key processes.

Tony cited qualitative wins in search conversion and support automation from the replatforming but offered no hard quantification despite the ask.

Tony highlighted AI agents automating merchant onboarding with perfect quality. Already seeing P&L benefits from AI tools across merchant and support.

Ravi confirmed redundant parallel tech stack costs through 2026, potentially bleeding into early 2027 before phasing out.

Tony challenged the premise, saying customer emails suggest they are falling short. Cited this morning's email about wrong parking lot delivery. Opportunities for improvement everywhere, not one area harder than another.

Tony highlighted Deliveroo's leading positions in profitable Western European markets. Exchange of product learnings with Wolt drove improvements. Comfortable to increase international investment after proving management bandwidth.

Ravi attributed CapEx increase to merchant tablet refresh and autonomous delivery investments. Expects similar levels for rest of year within G&A guidance.

Tony said priorities largely unchanged, continued investment in all five areas. Proud of 2024 results as reflection of 2022-2023 decisions.

Ravi confirmed opportunistic and conservative buyback approach, focused on driving returns over longer period. Generated over $2B in shareholder value.

Glenn described Genius-Connected Trip integration as a 'superpower' creating more loyalty and value, but declined to share specific details, saying they'd announce when ready to roll out.

Glenn clarified constraint is deliberate testing discipline, not compute or cost. Also noted regulatory complexity around AI rules in Europe plus privacy and payment compliance requirements.

Glenn described the goal as personalized trip planning using proprietary data, cited natural language search as already live, and emphasized the incremental nature of AI progress.

Ewout cited investments in product, supply, marketing, brand, and alternative accommodations; confirmed Q2 US growth slightly above Q1 and beating the broader market.

Ewout pointed to social media channel scaling and traditional channel experimentation, noting the ROI mix effect laps in Q2 and marketing leverage is expected to continue full-year.

Ewout explained high end is unchanged but low end expanded to reflect second-half macro uncertainty; if uncertainties don't play out, results would match Q4 guidance.

Glenn acknowledged the concern but offered qualitative reassurance about trusted brand and data advantages rather than specific risk mitigation strategies.