UBER (Trans): $10bn AV bet; buybacks yield to M&A for now

Dolphin Research's Trans of UBER FY26 Q2 earnings call.

I. Key takeaways

1. Shareholder returns: buybacks paused for M&A, to resume within months

a. Policy: the company has reiterated using approx. 50% of FCF for buybacks.

b. YTD: repurchased about $3.5bn so far this year. In Q2, capital allocation shifted tactically to M&A, deploying roughly $4.0bn to buy Delivery Hero SE shares in the secondary market to move fast on the opportunity.

c. Resumption cadence: with the M&A now announced, buybacks are being rebuilt steadily. Management stressed this is measured in months, not a year.

2. Outlook and trends (qualitative)

a. US Mobility: maintains the view of sequential acceleration through the year. Q2 saw acceleration in trips and GBV, and the team keeps its H2 view unchanged.

b. Delivery: reported growth in Q3 will be dragged by M&A base effects (see item 5), but organic growth is accelerating. Overall Q3 trends look relatively healthy.

c. P&L impact from AV investments: as deployment and scaling approach, P&L effects will emerge. The company will quantify them for investors as it has historically.

3. Key metrics this quarter

a. Scale: GBV rose 22% YoY to over $58bn, above the top end of guidance, marking a fourth straight quarter of 20%+ growth. Non-GAAP EPS grew 35% YoY.

b. Cash flow: TTM FCF topped $10bn for the first time in history.

c. Segment profitability: Mobility OPM remained strong at 7.6%. Delivery revenue margin was broadly stable.

d. Volume: the platform handled roughly 300mn trips per week. AV contributed several hundred thousand per week, less than 0.5% of total.

4. Take rate optics: declines largely accounting-driven

a. Mobility revenue margin fell nearly 500bps YoY, with ~400bps entirely due to a UK business model change. This is optical, reflecting a reclassification out of cost of revenue.

b. The remainder reflects deliberate investment, mainly into lower-price products such as Moto in Brazil.

c. Management urges focus on the net take rate disclosed in the 10-Q rather than the reported revenue take rate. On a net basis, take rate is broadly stable.

5. AV funding plan and M&A mechanics

a. The $10bn AV commitment spans multiple years, with most already disclosed via partnerships announced over the past year.

b. Structure has two buckets: equity investments in AV software partners with milestones, securing roadmap visibility and commercialization priority. And selective balance sheet use for ground AV infrastructure, including fleet ops, real estate, and OEM take-or-pay commitments, such as the multi-year 120k vehicle procurement.

c. Leverage and externalization: each $1 invested by the company has catalyzed $2.5 from other investors for partners. The team is advancing 'financialization' of the ecosystem with third-party capital vehicles to avoid overloading UBER's own balance sheet.

d. Delivery M&A base effects: Trendyol Go (consolidated in the second half of Jun 2025) will fully annualize from Q3. New contributions include Getir (announced this month) and the recombination with Careem, but Trendyol Go is much larger, creating a net drag on reported Delivery GBV and order growth.

II. Detail from the earnings call

> Note: management's full prepared remarks and supplements were posted on the IR site pre-call. The call featured only brief CEO opening comments, so section 2.1 is concise.

2.1 Management highlights

1. Operations

a. Q2 reached record users and engagement, supporting durable growth, margin expansion, and cash generation.

b. Growth translated into operating leverage: non-GAAP EPS +35% YoY and TTM FCF above $10bn position the company to invest from a position of strength across priorities.

2. Delivery Hero transaction

a. The deal will extend coverage to nearly 100 markets and roughly double the number of markets where the platform can offer Mobility and Delivery together.

b. Strategy is to replicate proven growth playbooks across more markets and bring Uber's tech and Uber One to more users.

3. Autonomous vehicles (AV)

a. The debate has shifted from whether tech can deliver competitive service to scale, reliability, and unit economics.

b. Industry structure view: AV is 'physical AI' and will mirror LLMs — multiple frontier models rather than one base model, layered with open-source. Recent signs include NVDA releasing Alpamayo with open weights, Wayve licensing in the UK, and Zoox approval to expand Robotaxi operations.

c. Key difference vs. base models: AV is a regulated physical system requiring market-by-market deployment and must address the temporal-spatial variability of rideshare demand. The commercialization layer is therefore critical.

d. Company role: aim to be the leading commercialization platform for AV globally, operating with discipline now while building capabilities that define the next decade of growth.

2.2 Q&A

Q: Over the next 12 months, what Mobility AV milestones matter most to validate the path?

A: Live deployments and the number of operating cities, split into with-safety-driver and fully driverless, are key. We are live in seven cities and could reach 15 by year-end.

New launches this year include Nuro-Lucid in the Bay Area, Zoox in Las Vegas, Wayve in London and Tokyo, BIDU in London, and Pony.ai and Verne in Zagreb, with potential additions. Looking to 2027, we plan to open more cities with existing partners and bring on new ones. Rivian should enter with a tightly integrated full-stack and competitive BOM costs, targeting 2028 for San Francisco and Miami. The NVIDIA path targets Los Angeles and San Francisco in 2027, expanding to 28 cities globally by 2028.

We track three things: city launches, accelerated data collection for next-gen end-to-end models, and the start of true commercialization. Volumes are still small, so we prioritize service quality, then vehicle utilization.

With Uber demand, partners can drive high per-vehicle utilization, often in the low-to-30s trips per car per day. That supports attractive monetization for these vehicles.

Q: Update on AV and automation for Delivery?

A: We have multiple partners on Delivery. On sidewalk robots, we work with Serve and Cartken, but I'm increasingly optimistic on drones, such as with Flytrex, and more drone partnerships are coming.

Drone unit economics need time to become sustainable, but the value is speed. AV couriers provide a safe, solid experience yet can be slower than humans, while drones can slash delivery times. A 30-minute dinner drop is already good, but 10–15 minutes is magical. So today we operate on sidewalks, and next we expect to take to the air with partners.

Q: What is driving US Mobility acceleration, and how much was one-off from the World Cup vs. sustainable?

A: The World Cup helped, broadly in line with expectations. The momentum we see in the US reflects themes we have discussed since the start of the year and is much broader than one-offs.

We expected sequential acceleration through the year, and Q2 delivered faster trips and GBV growth. We maintain that view for the rest of the year. Drivers are threefold. First, insurance turned from a headwind to a tailwind, and we recycled savings back into the market, notably California. LA and SF are inflecting, with trip growth well above the national average.

Second, product velocity, with traction in premium (Reserve, U4B, Black) and value (Wait & Save). U4B grew 40% YoY, and Uber Health grew even faster.

Third, sparse markets. This long-term opportunity remains a strong growth engine: in sparse US markets, less than 10% of eligible consumers used Uber in the past 12 months vs. over 50% in dense markets. We are improving supply reliability and investing in marketing to drive awareness and trials. Overall, the acceleration is broad-based and sustainable.

Q: Why is new user growth so strong, and is it sustainable?

A: New users are at least at multi-year highs, driven by similar factors. First, low-price offerings keep expanding: globally with two- and three-wheelers, and in the US with Wait & Save, where users trade time for price.

These products bring new cohorts onto the platform, after which we can graduate them to core offerings. Second, sparse markets are growing much faster than dense markets for both Mobility and Delivery. We are building supply in these markets with healthy margins, combining faster growth and strong profitability, with ample headroom before high penetration.

Third, cross-platform usage. Historically it was Mobility to Eats, but only about 20% of users use both, and this cohort grows 1.5x faster than single-product users. As Delivery scales, the flow is now bidirectional globally and in the US. Fourth, new products: Women Preferred is adding many female drivers and riders, Uber Teens opens new segments, and high-margin U4B brings profit and enterprise customers.

Q: Competitive dynamics behind softer Brazil Mobility orders. Do other markets need heavier Moto and value investment?

A: Brazil is competitive. Mobility has long been intense, with competitors like DiDi, typical for LatAm. The real change is in food delivery: DiDi launched Didi Food and Meituan entered to challenge iFood, all competing for the same two-wheeler courier supply.

That supply shifts between food and rides, pushing up acquisition costs. We reallocated subsidies from the consumer side to the earner side.

The good news is overall share in Brazil held, but mix shifted from Mobility to Delivery due to heavier investment there. Two-wheel is a newer, inherently low-margin business, so P&L impact is limited, though it does weigh on orders. Importantly, this pressure is from outside our core battlefield.

Q: Media speculated about the Waymo relationship. Can you clarify, especially Austin and Atlanta?

A: Waymo is a key partner. We continue operating in Austin and Atlanta and expect to be there next year, with a strong product and tight operational collaboration.

At the same time, we must avoid dependence on a single partner. The AV field, like base models, is seeing many new entrants. We will keep delivering with Waymo while partnering broadly. We target 15 markets by year-end and well above that next year.

For scale: AV is in the hundreds of thousands of weekly trips vs. about 300mn weekly trips overall, under 0.5%. In AI search, estimates suggest 20% of searches have shifted to AI and 40% of users use AI in some way. Physical AV will penetrate slower and more cautiously, well below AI levels, with regulation a factor. We have time to build a competitive, attractive-commercial-terms partner ecosystem.

Q: How is AV regulation evolving, and how will policy shape adoption pace and geographic rollout?

A: We are highly regulated and regularly engage governors, mayors, and legislators who have genuine concerns about AI and AV on constituents: jobs, safety, congestion. Some concerns are real.

AVs perform well in our deployed markets, but issues have occurred, such as behavior around schools and school buses, interactions with emergency vehicles, and responses during power outages when lights fail. These need real discussion.

Washington, D.C. is debating such issues — for example, AVs cannot block roads during presidential motorcades. We need proper dialogue so new rules reflect all stakeholders. We've seen the risks of moving too fast: some AI data center efforts pushed ahead under NDAs and met public backlash. We support smart regulation and stakeholder engagement to enable sustainable innovation.

Sometimes you must slow down to build durable rules. We are very pro-AV but want regulation that balances interests and supports a lasting model.

Q: Why slow hiring below plan, and how do you balance reinvestment vs. dropping savings to the P&L?

A: We have a multi-year track record of disciplined headcount growth. This year we focus on two workstreams: converting AI spend into productivity gains to slow hiring, and organizational effectiveness.

On AI, it's early but we are getting low-cost developer productivity uplift. Near-100% of engineers use AI coding tools, and by our cautious metrics, per-capita code output has roughly doubled. We are measured in how we track this, but it already supports slower hiring. We also have targeted AI deployments in support and in marketing to improve CX and lower costs.

On org effectiveness, we periodically review structures. This quarter we executed precise reductions of 10%–20% in a few teams, yielding savings that are modest at the group level. We aim to stay disciplined and reinvest a portion of savings back into the market to create value for consumers and earners.

Q: How will the $10bn AV commitment flow through the P&L, and what is the margin impact?

A: Context first: the $10bn spans years, and most corresponds to partnerships announced over the past ~12 months. Two buckets: equity investments in AV software partners with milestones to secure roadmap visibility and priority in commercialization. Our lead role helps catalyze external capital, with each $1 from us driving $2.5 from others.

The second bucket is selective balance sheet use to seed ground AV infrastructure, such as fleet ops, real estate, and OEM support where take-or-pay is required. The multi-year 120k vehicle procurement largely sits here. The goal is to catalyze the ecosystem, while advancing financialization with third-party capital so it is not solely on UBER's balance sheet.

As for P&L and cash flow, we will provide more visibility as we approach deployment and scale. When P&L effects emerge, we will quantify them clearly for investors.

Q: Given negative headlines, how do you view Lucid's ability to meet vehicle commitments?

A: Lucid just reported and appointed Silvio as CEO, who is taking bold steps to reset cost structure and focus on quality. These moves can be tough but are necessary and positive in our view.

I've spoken with Silvio and our relationship with Lucid is very close. Their B2B build and our AV program are highly strategic, with a large, monetizable order at $70k–$80k per vehicle, which is encouraging. Nuro and Lucid are tightly integrated on AI and vehicle training, including weight tuning and API integration, and progress is solid.

Lucid is backed by PIF, also a major shareholder of UBER, with a strong board representative. PIF is a long-term fundamentals investor and continues to support Lucid. We believe UBER + Nuro + PIF support, combined with Silvio's actions, is the right setup for Lucid to deliver on commitments.

Q: GBV +22% at constant FX vs. revenue +19%. How to read the take rate?

A: Recall the UK business model change, mainly affecting Mobility. Delivery revenue margin is broadly stable, so focus on Mobility: revenue margin fell nearly 500bps YoY, with about 400bps from the UK model change, an optical effect from moving costs out of cost of revenue.

The remainder is deliberate investment into value products, notably Moto in Brazil. I would point investors to the net take rate in the 10-Q rather than reported revenue take rate. On that basis, take rate is broadly stable, and Mobility OPM remains strong at 7.6%.

Q: Can you quantify AI-driven gains in conversion and spend as the product becomes more intuitive?

A: AI does a few jobs. First, better, more effortless UX — for example, Cart Builder lets you snap a dish or jot a recipe and builds your cart via AI. Users love it, and these carts are often 2x the size of non-AI carts, lifting AOV and experience.

Second, not purely base-model work, but larger models can make smarter predictions with broader signals. They ingest historical, real-time, and cross-platform signals to surface more relevant content. That could be better-targeted ads for higher conversion, organic ranking to highlight local restaurants, discount ranking to save users money, or AI alerts for likely out-of-stock items with substitutions — all improving conversion or core UX.

We are extremely early. Rather than a single step-change, we operate quarter by quarter, year by year, optimizing systems. Expect AI to contribute to AOV, service quality and reliability, and surfacing the right product at the right time. For instance, three-quarters of trips on Uber start with personalized destination suggestions, no typing needed.

Q: 'Steadily rebuilding buybacks to normal levels' — what timeline: years, quarters, or months?

A: TTM FCF just exceeded $10bn, and our stated principle is to allocate roughly 50% of FCF to buybacks. We have repurchased about $3.5bn year-to-date.

In Q2 we tactically pivoted to M&A, primarily purchasing Delivery Hero SE shares with around $4.0bn to act quickly on the opportunity. With the announcement behind us, we are steadily rebuilding buybacks. This is measured in months, not quarters or a year.

Q: What integration lessons apply to deals like Delivery Hero — what works and what pitfalls to avoid?

A: We have done smaller analogous deals before, and the lesson is to have a clear integration thesis before signing. For Delivery Hero, we have high confidence in the plan: migrate to our global platform, follow a proven playbook, run on a realistic timetable, and be conservative in communicated synergies.

We have diligenced the asset in detail. Our Delivery runs on a single platform, while Delivery Hero, excluding Baemin, runs on a different stack. There is significant opportunity to bring them onto a modern stack like Uber's. Assuming closing in H2 2027, we would spend 2028 planning and building, and execute major migrations in 2029, an appropriate timetable for a deal of this scale.

Synergies will come from areas with high certainty: single platform migration, overlap, infrastructure, duplicative roles and services, and shared services such as payments and cloud. We stand by the synergy targets communicated two weeks ago.

Q: In markets with AV products not yet on the Uber network, how are Mobility engagement and retention trending?

A: In more mature AV markets, business trends are strong, as seen in order growth from both existing users and new cohorts. SF, LA, and Phoenix all saw faster trip growth in Q2 vs. Q1, and our category positions are higher than a year ago.

Think about Google search: AI penetration in search is far higher than AV penetration in rides, for obvious reasons of physicality and regulation. Yet even with an estimated 20% shift to AI search, Google query volumes still rose YoY as AI expanded the market. Google entered later but won on distribution and brand, and we believe similar dynamics apply to us.

Also, our opportunity, especially in the US, extends well beyond current AV cities. Only about 30% of GBV and 25% of profits come from the top 20 cities, with the long tail of thousands of smaller cities and suburbs remaining the main growth and profit engine for years. We will keep investing in both human drivers and AV, and the US remains a strong market for both teams.

Q: What is AV Labs building, and how does it complement partners?

A: The industry is shifting from heuristic, if-then systems to end-to-end models, especially for physical AV. Players like Waymo and Nuro have worked on AV for years, but end-to-end models that learn from massive data and make human-like decisions are rapidly taking over.

With AV Labs, we want UBER to be an accelerator using real, rideshare-specific data to speed L4 model development. We are building hundreds of vehicles with Robotaxi-grade sensors operating in rideshare scenarios to collect a superset dataset and share it with all partners. Multi-partner ecosystems force each partner to gather unique, long-tail datasets. We can capture high-fidelity, rideshare-specific data once and bring scale benefits to AV development for everyone.

We see development pace accelerating and want to be part of that acceleration. AV Labs is designed as a strong partner and accelerator for the entire AV ecosystem.

Q: Delivery is cycling M&A bases. What was organic growth in Q2 and what does Q3 imply?

A: Recall Trendyol Go in Turkey closed in the second half of Jun 2025, so it will fully annualize from Q3. We also announced a second Turkey deal, Getir, which will add to growth from Q3 onward, and Careem recombination is a smaller deal.

Net-net, because Trendyol Go is much larger than the latter two, reported Delivery growth in GBV and orders will face a headwind. Underneath, organic Delivery is accelerating, with strong trends in the US and internationally and category share gains across major markets. So organically we are accelerating, while inorganics shift from one large to smaller deals. Overall, Delivery should present a relatively healthy Q3 trend.

Risk disclosure and statements:Dolphin Research disclaimer and general disclosures