The Disruption That Doesn't Arrive

Introduction — Why I bought Uber

Uber is my most recent addition to the portfolio and one that has followed a similar story to past holdings that have proven very successful. It is a story where risks are being overly priced into the stock — risks that will eventually fail to have any material impact, and which have the potential to become a tailwind to the business. That risk is autonomous vehicles, which I discuss in detail below.

This investment follows the same pattern as a number of my past and present holdings: one where disruption fails to materialise. The clearest example is Google, which I leaned into heavily during the period when ChatGPT was supposed to disrupt Search. As Google's recent appreciation has shown, Search was not disrupted — it was strengthened by AI, and growth has accelerated. I believe Uber will be no different.

The pattern I look for is specific, and worth naming because it is the whole basis of the position. A dominant business is confronted by a technology the market believes is existential. The narrative is simple, the threat is easy to describe in a headline, and the defence requires understanding the incumbent's actual moat — which is usually distribution, data, or network density rather than the technology itself. The market discounts the entire business on the narrative. Meanwhile the fundamentals keep compounding, and the multiple quietly detaches from the performance.

That detachment is exactly what has happened here. Over roughly fourteen months to August 2026, gross bookings grew from $162.6bn to $193.5bn, free cash flow rose from $6.9bn to $10.0bn, and GAAP operating margin expanded from 6% to 11%. The business kept compounding. The stock did not — shares sit around 23% below the highs near $101 touched last autumn. The disconnect between operating performance and price is the opportunity.

Q2 2026 · Reading the Numbers — Ignore the Revenue Line

Before anything else, a decoding note, because Uber's reported figures currently mislead on first reading. Gross bookings grew 24% while reported revenue grew only 12%. The entire gap sits in Mobility, where bookings grew 22% but revenue grew 1%, and take rate fell from 30.7% to 25.4%.

Uber attributes this to "business model changes," which reduced total revenue growth by eight percentage points. This is almost certainly a shift to agency or net-revenue accounting in certain markets — the UK VAT treatment being the archetype — under which Uber books only its commission rather than the gross fare. My inference, not a company disclosure.

The evidence that this is presentational rather than economic is decisive: Mobility segment operating income still grew 28% to $2.2bn, and margin on bookings actually rose from 7.3% to 7.6%. Profit is unaffected. Only the revenue line was restated by accounting geography.

The practical consequence is that any screen or model anchored on Uber's revenue growth produces a false read — and a good deal of the market is running exactly that screen. Gross bookings and segment operating income are the measures that matter.

Source: Uber Q2 2026 earnings release and supplemental data, quarter ended 30 June 2026. Take-rate and margin-on-bookings figures derived from disclosed data. GAAP net income includes a non-operating, non-recurring equity revaluation gain; underlying GAAP earnings are closer to $0.8bn.

Growth — 20%+ at Enormous Scale, and Where It Comes From

The most remarkable feature of this business is that trips and gross bookings are still compounding above 20% on a base of roughly 15 billion annual trips. Companies this large are not supposed to accelerate. Uber added more first-time users in the trailing twelve months than in any comparable period in five years.

Trip growth of 18% decomposes into 16% more users and 2% more frequency. So this is an acquisition story, not an engagement story — a point I return to in the risks. The question that matters is where 16% more users come from at this scale. Management identified the drivers on the call, and they are more durable than I expected.

Sparse markets — the structural runway

The single most important disclosure of the quarter, and the one I keep returning to: fewer than 10% of eligible consumers in sparse markets have used Uber in the past twelve months, versus over 50% in dense markets.

Sparse markets are low-density geographies — suburbs, exurbs, smaller towns — inside countries Uber already operates in. This is not new-country expansion. Rideshare is a liquidity business, and thin density historically broke the economics: fewer riders means fewer drivers, longer waits, worse service, fewer riders. Uber concentrated on cities because cities worked.

Dense urban cores are now roughly half saturated. The periphery is almost untouched, and it contains far more people. What changed is that lower-cost formats and better matching finally make thin liquidity viable. This is the most durable growth lever Uber has, it requires no acquisitions, and it is quantified.

New categories and formats

• Two- and three-wheelers. Motorbike rides (Uber Moto, Uber Boda) and auto-rickshaws (Uber Auto). These reach a customer for whom a car ride was never affordable, and they work precisely where cars struggle — congestion, narrow streets, short trips. Driver supply is far easier to build because a motorbike costs a fraction of a car.
• Wait&Save. Trades speed for price. Well matched to sparse markets because it reframes structurally longer wait times as a deliberate customer choice rather than a service failure.
• Trust and safety formats — Women Preferred, Uber Teen. Worth separating from the price plays. These unlock demographics that were never price-constrained; they were trust-constrained. A different kind of addressable-market expansion.
• Premium and business. Reserve, Black, and Uber for Business growing 40% YoY. Non-UberX trips grew roughly 110% YoY. The product ladder now spans a three-wheeler to a chauffeured airport reservation.
• Insurance savings recycled into price. US insurance costs turned from headwind to tailwind, and rather than banking it Uber is cutting fares — visibly in California, where LA and San Francisco trip growth now outpaces the rest of the country. A cost tailwind converted directly into demand.

Margins — Operating Leverage Is Doing What It Is Supposed To

Growth alone is not the case — plenty of companies grow bookings without ever converting it. What makes Uber a compounder rather than a story is that profit is growing considerably faster than volume, and the gap is widening.

Non-GAAP operating income grew 40% against 22% bookings growth. Margin on gross bookings expanded from 3.3% to 3.7%. Corporate overhead grew 18% against 24% bookings growth — the definition of operating leverage. GAAP operating margin has expanded from 6% to 11% over roughly fourteen months.

Both segments are contributing. Mobility segment operating income grew 28% with margin on bookings rising to 7.6%. Delivery — now nearly the same size as Mobility at $27.5bn of bookings and growing faster — grew segment operating income 38% with margin expanding to 3.8%. The old mental model of Uber as a rideshare company with a loss-making food business attached is several years out of date.

Three structural forces should continue this. Fixed costs spread across a larger base. Mix shift toward higher-margin revenue lines (advertising, subscriptions, merchant-funded offers). And the cross-sell flywheel lowering acquisition cost per incremental product — only 20% of consumers use both Rides and Eats today, and cross-platform users grow 1.5x faster than single-product users.

The counterweight, which I take seriously, is that management is deliberately choosing to reinvest rather than let margin flow through. The $10bn AV commitment, the Delivery Hero acquisition, and cutting fares with the insurance savings are the same decision repeated. That is why the Q3 EBITDA guide came in light. It is a choice about pace, not a failure of the model — but it does mean margin expansion arrives more slowly than the underlying leverage would allow.

Optionality — Three Engines the Market Is Not Paying For

Beyond the core, Uber has three businesses compounding underneath the headline numbers. Two are already material and disclosed only in commentary; the third is the one everyone mistakes for a threat.

Note that none of these is a reportable segment. Advertising revenue is booked inside Mobility and Delivery — sponsored listings in the rider app, promoted placement and sponsored search in Eats. There is no advertising line anywhere in the segment financials.

This has two consequences worth understanding. It inflates reported take rates, because ad revenue has no corresponding gross bookings. And it flatters segment margins — part of Delivery's margin expansion to 3.8% is high-margin advertising mixing in rather than better delivery unit economics. The two cannot be separated from public disclosure. If Uber ever breaks advertising out as a discrete line, that will itself be a signal that management wants it valued separately.

The Autonomous Vehicle Question — Why I Think the Disruption Fails to Land

The bear case is easy to state, which is part of why it is priced so heavily. Consumer switching costs between ride apps are near zero. If Waymo or Tesla offer a cheaper, better robotaxi through their own app, what keeps anyone on Uber?

It is a fair question and I do not dismiss it. But the argument rests on a premise I think is wrong: that the scarce asset in this industry is the ability to drive autonomously. I think the scarce asset is demand — and that autonomy is on a path to becoming a purchasable component.

Scale takes far longer than the narrative assumes

Building fleets is a manufacturing and capital problem, not a software problem. Uber runs roughly 300 million trips a week; AV trips are currently under 0.5% of that. Waymo — the clear leader — operates around 4,000 vehicles. Even an extremely successful rollout addresses low single-digit percentages of the market for years.

Until an operator has metro-wide density, its ETAs are long and its coverage is patchy. And consumers do not choose on brand or even principally on price — they choose the shortest wait. A robotaxi service can win the six blocks it serves and still lose the default position for the city.

The early evidence points the other way

This is the part of the thesis I weight most heavily, because it is empirical rather than theoretical. In Phoenix, Los Angeles and San Francisco — the markets where robotaxis are most widely available outside the Uber app — Uber's trip growth accelerated versus the prior quarter.

That is the first real-world test of the cannibalisation thesis, and so far it suggests robotaxi availability is expanding the category rather than displacing Uber. Autonomy lowers the cost and raises the appeal of not owning a car; the beneficiary of a larger rideshare market is the aggregator holding the demand. Small sample, early days — but it is evidence, and the bear case currently has none.

Commoditisation, and why Nvidia matters more than the headline

Nvidia is not building bespoke technology for an Uber fleet. It is doing something more consequential: turning autonomy into a reference platform anyone can buy. DRIVE AGX Hyperion 10 — two Thor SoCs on Blackwell architecture, 14 cameras, nine radars, LiDAR, twelve ultrasonic sensors, plus DRIVE AV software — is described as making any vehicle L4-ready. Stellantis, Lucid and Mercedes-Benz are building on it. Aurora, Volvo and Waabi extend it to freight. Uber and Nvidia are jointly building an AI data factory on Nvidia's Cosmos platform, targeting 100,000 vehicles from 2027.

This is Nvidia doing to autonomy what it did to AI compute: becoming the arms dealer that lets everyone compete. Waymo's advantage is vertical integration and a decade's head start. Nvidia's play makes that head start purchasable. When supply commoditises, the winner is the aggregator holding the demand, the proprietary data and the network effects.

Uber can afford to be behind

This is the asymmetry I think the market misses. An AV developer must build technology, raise capital, manufacture vehicles, and then acquire demand from zero. Uber already has the demand. The moment its fleets go live they monetise instantly at high utilisation, because the riders are already in the app.

Uber is therefore not in a race to be first. It is in a position to wait, let others fund the technology risk, and buy the winning components once they are commoditised. Its roughly 30 AV partnerships exist precisely to ensure no single supplier gains monopoly leverage over it.

The hybrid network solves a problem pure AV fleets cannot

Management raised a point I had underweighted: demand is enormously variable. Friday night, a rainstorm, a stadium emptying, New Year's Eve. A pure-AV operator must either build enough fleet for peak demand — leaving expensive capital idle most of the week — or fail its customers at exactly the moments they care most.

A hybrid network solves this elegantly. AVs handle the predictable base load at high utilisation; human drivers flex in to absorb peaks, and their capital cost sits on someone else's balance sheet. That is structurally cheaper than any pure-AV fleet can be, and it is not a transitional arrangement — it is a durable architecture.

There is also a smaller but persistent point: a portion of riders will not fully trust driverless technology, and will actively choose a human driver. That may be a shrinking cohort, but it is a real one, and it can only be served by a network that has both.

Even the bear case is survivable

Finally, the segment most exposed to AV disruption is Mobility — and Mobility is becoming a smaller share of the business. Delivery is nearly the same size, growing faster, with faster margin expansion, and is essentially untouched by robotaxis. Advertising and subscriptions are compounding on top.

So a scenario where AVs genuinely take material Mobility share is damaging but not catastrophic — particularly from a multiple that already discounts it. The market is applying an existential discount to a risk whose realistic downside is a slower-growing segment inside a diversifying business.

Capital Allocation — The $10bn Commitment and What It Does to Margins

Uber is not building vehicles. It sold its in-house self-driving unit to Aurora in 2020 and has no intention of returning to vehicle or software development. The commitment splits into roughly $2.5bn of equity stakes (Lucid, Rivian, Nuro, Wayve, WeRide, Waabi) and roughly $7.5bn of milestone-linked fleet purchases, targeting 120,000 vehicles. Concrete orders include at least 35,000 Lucid vehicles, up to 50,000 Rivian R2 SUVs, and at least 5,000 Stellantis L4 vehicles from 2028.

The obvious objection is margin stacking: if Uber buys fleets from manufacturers who need their own profit, how good can the economics be? Three factors offset it.

• Buyer leverage is enormous. Rivian expects 65–67k total deliveries in 2026. A 50,000-unit order is not a customer relationship — it is an existential volume commitment that fills the factory. That is the position from which near-cost pricing is extracted.
• The equity stakes recapture supplier margin. Uber owns over 11% of Lucid. If Lucid earns a margin selling Uber 35,000 vehicles, Uber recovers a share of it — plus upside if the order itself re-rates the supplier. Partial vertical integration without the capex or execution risk of manufacturing.
• Hardware costs fall from here. A standardised, mass-produced Hyperion-class compute and sensor set on a Blackwell cost curve is exactly the mechanism by which AV hardware gets cheap. Uber is buying at the most expensive point of that curve today.

Split of the $10bn between equity stakes and fleet purchases is an estimate assembled from company announcements and press reporting, not a single disclosed breakdown.

Framed simply: Uber is spending $10bn to convert itself from a pure marketplace into a partially asset-owning operator, betting the profit pool moves to the fleet layer. I think that is the right bet, but it is a genuine change in the business model, and it justifies some of the multiple compression the market has applied.

Delivery Hero — A Large, Expensive, Strategically Sound Acquisition

Announced 16 July 2026, after the quarter closed: €41.50 per share in cash, $14.8bn equity value ($13.7bn adjusted for prior stake purchases). Uber already held roughly 37% economic ownership; with Prosus committing its ~17% stake, Uber's total economic interest reaches approximately 53%. Delivery Hero is separately selling 14 overlapping markets to SSW Partners for around $1.6bn to manage antitrust exposure.

Combined pro-forma 2025 gross bookings of $236bn across 99 markets, with cross-platform markets rising from 34 to 58 and 50m+ eligible cross-platform users added. Uber guides to over $1.2bn of run-rate synergies and high-single-digit percentage non-GAAP EPS accretion by year three.

Two cautions. The premium is steep — roughly 127% over the unaffected three-month VWAP before 8 May, though only about 34% over the pre-announcement average since the stock had already run on speculation. And with completion expected in H2 2027, 2028 planning, and 2029 migrations, execution risk is live and long-dated.

The strategic logic is nonetheless sound. Food delivery is scale-dependent and consolidating, and doubling the cross-platform footprint feeds the Uber One flywheel that is demonstrably working.

Risks — What Could Make Me Wrong

The thesis is that a specific risk is overpriced. The discipline that makes that thesis credible rather than complacent is stating precisely what would falsify it.

Valuation — Deliberately Conservative, and Still a 20% CAGR

I have not tried to model the bull case. The point of the exercise is to establish what the return looks like if very little goes right — because if the conservative case clears my hurdle, the optionality is free.

On three-year projections of 12% revenue growth — roughly half the current rate — operating margins expanding to 18% from about 14% today, and a 25× multiple, the implied return is approximately a 20% CAGR from today's valuation. That excludes buybacks, where shares outstanding have been falling 2–3% per year and fell 3.7% in the latest quarter.

Why each assumption is conservative

• 12% revenue growth against 22% current bookings growth, five consecutive years above 20% constant-currency, and a sparse-market runway with sub-10% penetration. This assumes a sharp and permanent deceleration.
• 18% operating margin against demonstrated operating leverage — 40% operating income growth on 22% bookings growth, GAAP margin already up from 6% to 11% in fourteen months — plus mix shift into advertising, subscriptions, and eventually higher-margin AV trips. This assumes the leverage largely stops working.
• 25× exit multiple assumes essentially no re-rating. If the AV fear resolves and the market re-rates toward a compounder multiple, that is upside not counted here. Equally, if it de-rates toward a capital-heavy operator, this is where the case is most sensitive.
• Buybacks excluded. At 2–3% annually, share count reduction alone adds meaningfully to the per-share figure. Treat it as an uncounted margin of safety.

Monitoring — What I Am Watching, and What Would Change My Mind

A thesis that cannot be falsified is a belief, not an investment. These are the specific, checkable signals I will track each quarter.

Signals that would confirm the thesis

• AV city count reaching the stated target of 15 by year end, from seven today. Concrete and falsifiable.
• Non-Waymo AV ride quality in Austin and Atlanta replacements, Dallas, Las Vegas, Abu Dhabi. Proving suppliers really are interchangeable is what validates the entire aggregation model.
• Continued acceleration in markets with heavy robotaxi presence — the single most important empirical test of the cannibalisation question.
• Advertising run rate compounding above 40%, and any move to disclose it as a discrete line.
• Uber One membership and cross-platform share — the 20% dual-product figure rising would confirm the flywheel.

Signals that would break it

• Any other AV partner moving toward a direct consumer app. Waymo alone is an outlier; a second defection is a pattern, and the aggregation thesis fails.
• Q4 delivery growth failing to recover once Trendyol fully laps — that would mean the "organic acceleration" claim was spin.
• Frequency staying flat below 6.2 trips per user while user growth decelerates. That is the point at which growth quality has genuinely eroded.
• Margin on bookings ceasing to expand for two or more consecutive quarters — the leverage mechanism failing rather than being deliberately spent.
• US Mobility failing to accelerate through 2026, against management's explicit guidance. A checkable commitment.
• Adverse AV liability rulings materially raising insurance assumptions for owned fleets.

Conclusion — Position Summary

Uber is compounding gross bookings above 20% at a scale where that should not be possible, converting it into 40% operating income growth, generating over $10bn of trailing free cash flow, and shrinking its share count — while trading at a valuation that discounts an existential threat currently representing under 0.5% of its trips.

Three engines the market is not paying for — Uber One past 50 million members, advertising at a $2.5bn run rate growing 50%, and the AV option itself — sit on top of a sparse-market runway with sub-10% penetration and a Delivery business now nearly as large as Mobility and growing faster.

Against that, the business is genuinely becoming more capital-intensive, the strongest AV partner has chosen to compete rather than distribute, and my valuation depends more on the multiple holding than I would like. Those are real, and I have set out above exactly what would tell me I am wrong.

But the shape of this is one I recognise. A dominant business, a narrative disruption, a multiple that has detached from performance, and fundamentals that keep compounding while the market looks elsewhere. I bought Google into the same set-up and for the same reason. I expect this to resolve the same way — not because the risk is imaginary, but because it is overpriced.

Sources & Disclosure

Basis of preparation. Compiled from Uber's Q2 2026 earnings press release and supplemental data (quarter ended 30 June 2026), management commentary from the Q2 2026 earnings call, company announcements regarding Delivery Hero and the Nvidia partnership, and contemporaneous press reporting on the Waymo partnership. Figures cited are as reported by the company; segment margin and take-rate calculations are derived from disclosed figures. Where estimates or inferences appear — notably the interpretation of "business model changes" as agency accounting, and the split of the $10bn AV commitment — they are identified as such in the text. Advertising and Uber One figures are drawn from management commentary rather than segment disclosure, as neither is a reportable segment.

Disclaimer. This is a personal research note recording my own reasoning for my own capital allocation. It is not investment advice, a recommendation, or an offer to buy or sell any security. It has not been prepared in accordance with regulatory requirements designed to promote the independence of investment research. I hold a position in the security discussed and my views are therefore not disinterested. Forward-looking assessments are inherently uncertain and may prove incorrect. Anyone acting on this material does so entirely at their own risk and should consult a suitably qualified and regulated adviser.