Intellegix National — September 06, 2026

The Uber AV Thesis — What Wall Street Is Actually Betting On

Let's start with the BMO Capital call, because it's a useful entry point into the larger story. Brian Pitz at BMO reiterated an Outperform rating and a $119 price target on Uber this week, and the language in that note is worth parsing carefully. He's not just saying "ride-sharing is going to grow." The specific framing is that Uber's "expanding AV infrastructure and growing partner base should position it as the preferred mobility platform for autonomous vehicle manufacturers." That is a very particular thesis, and it's one that has profound implications for market structure.

The key word there is "preferred." What Pitz and other Wall Street analysts who share this optimism are essentially betting on is that Uber won't need to build its own autonomous vehicles. Instead, they'll be the distribution layer — the platform through which every AV manufacturer reaches riders. Think of it like how Apple's App Store doesn't write the apps but takes a cut of every transaction. Uber's play, if this thesis holds, is to become the App Store for autonomous mobility.

And the "three continents" framing is doing a lot of work in that headline. Uber has AV partnerships that are now active or in advanced development across North America, Europe, and parts of Asia. That geographic breadth is actually a moat — it's not just scale, it's regulatory diversity. A company that has learned to navigate the different licensing regimes in California, Germany, and Japan simultaneously has institutional knowledge that a pure-play AV startup operating only in Phoenix simply doesn't have.

Wall Street consensus on Uber is broadly bullish for exactly this reason, Marcus, but I want to introduce a note of skepticism here that I think is important. The "preferred platform" thesis assumes that AV manufacturers will want to route through Uber rather than build their own distribution. But we've already seen Tesla make clear that it wants to own the full stack — the car, the software, the rider relationship. Waymo has been experimenting with its own direct-to-consumer app in San Francisco and Phoenix. The assumption that AV makers will cede the customer relationship to Uber is not settled.

That's a fair tension to name. The counter-argument from Uber's side would be that the economics of building a consumer-facing mobility brand are brutal. You need density, you need trust, you need marketing infrastructure. Uber has spent 15 years and billions of dollars acquiring those things. A company like Renesas, which makes chips, is not going to want to also run a consumer app and manage surge pricing algorithms and handle customer complaints at 2 in the morning. So the question is really: which AV players will want the full vertical, and which will prefer to just put cars on an existing network?

And the answer probably isn't binary. Some — Tesla, maybe Waymo — go full vertical. Many others partner with Uber. And that's actually enough to make the BMO thesis work, because Uber doesn't need to be the exclusive platform. It just needs to be where the majority of rides happen. At 7,722 on S&P futures this morning, slightly softer than Friday's close, the broader market context suggests investors are in a mild risk-off mood, but Uber's AV narrative has been relatively insulated from those macro swings because the thesis is so long-dated.

Right — you don't buy Uber at $119 because you think Q3 earnings will beat by four cents. You buy it because you believe the 2035 mobility economy looks structurally different from today's, and Uber is positioned to extract rent from that transition. That's a different kind of investment decision, and it requires a different kind of conviction.

Which brings us directly to what Andrew Macdonald said at the end of August, because his remarks are essentially the internal narrative that Uber is now willing to make public — and that shift in communication strategy is itself worth noting.

---

The End of Car Ownership — Macdonald's Claim and What It Actually Requires

Uber's COO Andrew Macdonald made a statement that I've been thinking about all week. He predicted that within 15 to 20 years, most people will no longer need a driver's license, and he called personal car ownership "the most inefficient asset" a person owns, pointing to the fact that cars sit parked approximately 98 percent of their lives. Now, I want to take that seriously as a claim rather than dismiss it as corporate boosterism, because the underlying logic is genuinely interesting — and so are its failure modes.

The 98 percent figure is real and has been cited in transportation research going back at least a decade. The average American car is in motion for about 30 to 60 minutes a day and stationary the rest of the time. From a pure asset utilization standpoint, that is a terrible ratio. You wouldn't run a manufacturing line that way. You wouldn't staff a restaurant that way. The only reason we've accepted it is that the alternative — coordinating shared mobility at the moment you need it — has historically been too unreliable, too expensive, or too undignified to substitute.

And what Macdonald is arguing is that autonomous vehicles break all three of those barriers simultaneously. They're potentially cheaper per mile at scale because you've eliminated the driver cost, which is currently 60 to 70 percent of a ride-share fare. They're more reliable because they operate 24 hours a day without human fatigue. And the dignity question becomes less relevant when the vehicle arrives in four minutes, is clean, and offers a consistent experience.

But here's where I want to push back, Sarah, because the 15-to-20-year timeline for "most people" is doing an enormous amount of work. The United States has roughly 290 million registered vehicles and a physical geography that is deeply hostile to the autonomous vehicle thesis in its purest form. Suburban and rural America — which is where the majority of the country's population actually lives — does not have the density required to make shared autonomous mobility economically viable on the timeline Macdonald is describing. A farmer in western Kansas is not going to summon an Uber AV to get to the grain elevator at 5 a.m.

That's exactly right, and I'd add a regulatory dimension. The driver's license is not just a permission slip to drive. In the United States it functions as a de facto national ID for millions of people — you need it to board a domestic flight, to open a bank account in some states, to vote in certain jurisdictions. Saying "most people won't need a driver's license" is a statement about transportation infrastructure, but it's also a statement about identity infrastructure that Macdonald almost certainly wasn't intending to make. Those systems are not going to transform on the same timeline as a ride-share app.

Where Macdonald's prediction is most likely to prove accurate is in dense urban cores — New York, San Francisco, Chicago, Boston, London, Singapore, Tokyo. In those environments, the math already works or is close to working. The question is whether "most people" means "most people in dense urban environments" or "most people on Earth," and those are very different predictions. If it's the former, he might well be right on schedule. If it's the latter, 15 to 20 years is almost certainly not enough time.

What I find most interesting about Macdonald's remarks is the strategic intent behind making them publicly. Uber is essentially shaping the Overton window for what the transportation future is supposed to look like — and by positioning car ownership as an inefficiency rather than a freedom, they're trying to shift cultural norms, not just describe technical reality. That's a communications strategy as much as a technology forecast.

And it has regulatory implications too. If the public narrative is that autonomous vehicles are an inevitable efficiency improvement, regulators feel pressure to clear the path rather than obstruct. If the narrative is that autonomous vehicles are a corporate land-grab on a fundamental aspect of American life, you get a very different regulatory environment. Uber knows this, which is why their executives are out making these sweeping claims publicly. The 98 percent figure is real. The 15-to-20-year prediction is aspirational. The strategy of conflating them is intentional.

---

Renesas, Autoware, and the Open-Source Hardware Race

Here's a story that got almost no mainstream coverage but is genuinely important if you care about how the self-driving industry's technical architecture is going to be decided. Renesas — a Japanese semiconductor company that most people outside the chip industry haven't heard of but which supplies roughly 40 percent of the automotive MCU market globally — has joined the Autoware Foundation. And a demonstration of Autoware running on Renesas's R-Car X5H hardware is planned for Automotive World 2026, which runs September 9th through 11th at Makuhari Messe, just outside Tokyo.

Let me explain what Autoware is for listeners who aren't deep in the automotive software world, because this matters for context. Autoware is an open-source software stack for autonomous driving — think of it as Linux for self-driving cars. It was originally developed by Tier IV, a Japanese AV startup, and it has become one of the most widely adopted open-source AV software frameworks in the world, with deployments in robotaxis, delivery vehicles, and industrial transport across multiple countries.

The significance of Renesas joining the Autoware Foundation isn't just that a chip company is supporting an open-source project. It's that Renesas is committing to building hardware that is explicitly optimized for Autoware — the R-Car Gen 5 SoC, which stands for System-on-Chip, is being co-developed with Tier IV to run Autoware natively. That's a much deeper integration than just "we support this standard."

And the strategic logic here is interesting from multiple angles. For Renesas, the play is to become the preferred silicon for the open-source AV ecosystem, in the same way that Nvidia became the preferred silicon for machine learning research through CUDA and the academic open-source community. If Autoware runs best on R-Car hardware, every company building on Autoware defaults to Renesas chips. That's a distribution strategy disguised as a technical collaboration.

For the broader industry, the Renesas-Autoware partnership represents a genuine alternative to the proprietary closed-stack approach that companies like Tesla and some others have pursued. A world where the AV software layer is open-source and the hardware is commoditized is a world where a startup in Seoul or Nairobi can build a self-driving vehicle without needing to license technology from a handful of American or Chinese companies. That has real implications for global competition in this space.

The September 9th to 11th demonstration window is worth watching closely. Automotive World is a serious trade event, not a press conference — the audience is engineers and procurement executives from major automakers. If the Autoware on R-Car X5H demo performs well in front of that crowd, you're going to see purchase discussions that could shape the supply chain for vehicles being manufactured in 2029 and 2030. These procurement cycles are long, which is why the news is happening now even though the impact won't be visible to consumers for years.

There's also a geopolitical dimension here that I want to flag. Renesas is Japanese. Autoware was developed by a Japanese company. The demonstration is happening in Japan. As the U.S.-China semiconductor competition has intensified over the past several years, Japan has quietly positioned itself as a third pole in automotive chip technology — one that is trusted by both Western and Asian automakers in ways that purely American or purely Chinese suppliers currently aren't. This partnership is partly a technical story and partly a market-positioning story inside that geopolitical context.

Well said, Marcus. And it's worth noting that the open-source dimension of this story actually connects to the Uber narrative we were just discussing. If Uber's platform thesis depends on a broad ecosystem of AV manufacturers — rather than one or two dominant players — then the existence of open-source frameworks like Autoware actually helps Uber. More manufacturers using a common software base means more vehicles that could potentially integrate with Uber's network, rather than a fragmented landscape of proprietary systems that each require separate integration work.

---

Uber's AI Economics — The Tenfold Surge That Didn't Break the Bank

There's a data point inside Uber's recent AI infrastructure reporting that I think deserves its own conversation, because it has implications well beyond Uber. The company has seen AI usage surge nearly tenfold — and I want to be precise here, this is internal AI usage, meaning the models and inference infrastructure that power things like routing, pricing, fraud detection, and driver matching — while AI spending costs have stabilized. Usage up tenfold, costs flat. If that ratio holds broadly across the enterprise AI industry, it changes the investment calculus for AI infrastructure in a significant way.

Let's break down why that ratio is possible, because it's not magic. There are two things happening simultaneously. First, Uber and companies like them are increasingly adopting open-weight models rather than paying for proprietary model APIs at the same rate. Open-weight models — where the weights are publicly released and companies can run them on their own infrastructure — are dramatically cheaper to operate at scale than paying per-API-call to a model provider. You spend more upfront on hardware, but the marginal cost of each inference drops substantially.

Second, Uber has invested heavily in what they're calling "smarter routing infrastructure" — which in plain terms means they've gotten much better at deciding which queries actually need a large, expensive model and which can be handled by a smaller, cheaper one. Not every decision in a ride-share platform requires the same level of AI sophistication. Figuring out whether a driver is likely to cancel a trip in the next two minutes might need a fairly simple predictive model. Optimizing dynamic pricing across 40,000 drivers in a city during a stadium event might need something more powerful. Uber has apparently gotten much better at routing workloads to the right model tier.

The signal this sends to the broader market is that we may be moving out of the "spend freely on AI and figure out the economics later" phase that characterized 2023 through 2025, and into a phase where companies are genuinely disciplined about which AI investments generate measurable returns. That's not a bad thing for the industry — it's actually a sign of maturity. But it does have implications for the AI infrastructure companies — the hyperscalers, the GPU manufacturers, the API providers — who have built their business models around the assumption that enterprise AI spending would grow without constraint.

Investors in Nvidia, in the hyperscalers, and in companies like Anthropic and OpenAI should be watching the Uber AI cost curve very carefully, because Uber is not a small or unsophisticated customer. If one of the world's most data-intensive consumer technology companies is finding ways to do ten times the work for the same dollar spend, other large enterprises are going to demand the same efficiency from their AI vendors. The era of "just throw it at GPT-4" without thinking about cost-per-inference is ending.

There's a broader economic principle at work here too — it's essentially the same dynamic we've seen in every technology infrastructure wave. Early adopters overspend because they're exploring what's possible. As best practices emerge and tooling matures, the cost-per-unit-of-value drops dramatically. Cloud computing followed this path from 2008 to 2015. Mobile app development followed it from 2010 to 2018. Enterprise AI is following it now, and Uber's tenfold usage-to-cost ratio is one of the clearest data points we have that the maturation is happening faster than most predicted.

And this is where I want to add the Curriculum Corner context, Marcus, because Uber's AI efficiency story is interesting on its own, but the bigger question is what happens when Uber's AV and AI advantages become so significant that regulators start asking whether the company is using its platform position to foreclose competition. That's exactly the question antitrust law is designed to answer.

Walk us through it, Sarah.

So the starting point for antitrust in the United States is the Sherman Antitrust Act of 1890 — yes, 1890, which means the legal framework for managing the most powerful technology companies of the 21st century was written before the automobile was invented, which is itself a fascinating limitation. Section One of the Sherman Act prohibits contracts or conspiracies that unreasonably restrain trade. Section Two is the one that applies most directly to Uber's situation — it prohibits monopolization, meaning the willful acquisition or maintenance of monopoly power, not just having it.

Here's the key distinction that most people get wrong: having a large market share is not illegal. Dominating a market because you built a better product is not illegal. What Section Two prohibits is using monopoly power to harm competition through exclusionary conduct — things like predatory pricing designed to drive out competitors, or using control of one market to unfairly advantage yourself in another. The classic analogy is a railroad company that owns the only tracks in a region — that company can't charge competing freight companies rates that make it impossible for them to compete while giving its own freight division preferential pricing.

Applied to Uber: if Uber becomes the dominant platform for AV deployment globally, the question regulators will eventually ask is whether Uber is using that platform position to disadvantage AV manufacturers who try to run their own direct consumer apps, or to lock in terms that make it economically impossible for competitors to emerge. Market share alone — even 70 or 80 percent of urban autonomous rides — isn't sufficient for a Sherman Act Section Two violation. The conduct matters as much as the share. But the bigger Uber gets in AV, the more carefully antitrust lawyers at the DOJ and the European Commission are going to be watching those conduct questions.

That's a really useful frame, Sarah, and it connects to something I want to flag for listeners: the current DOJ antitrust division has shown it's willing to bring cases that would have been considered far-fetched five years ago. The Uber of 2026 is not the Uber of 2016. It's a much more politically visible target, and the moment it starts generating the kind of market concentration that the BMO thesis implies, expect regulatory scrutiny to accelerate.

---

The AV Supply Chain — What the Industry Architecture Means for Cities and Workers

Let's zoom out from the individual company stories and talk about what the convergence of all these threads actually implies for the physical and economic landscape of cities. Because if Uber's COO is right even directionally — if car ownership does decline significantly over the next two decades — the second-order effects are staggering in a way that I don't think has fully entered the public conversation.

Start with parking. American cities have an estimated 800 million parking spaces — roughly three spaces for every registered vehicle. Those spaces occupy an enormous percentage of urban land, often in the most valuable parts of cities. If car ownership declines by, say, 30 to 40 percent over 20 years in dense urban areas, what happens to parking garages? They're already being designed for conversion in some cities — there are mixed-use residential buildings going up in Chicago and Seattle right now that were parking structures five years ago. That trend accelerates dramatically in an AV-heavy future.

Real estate is just one dimension. The auto industry directly employs about 8 million people in the United States when you include dealers, mechanics, parts manufacturers, and assembly workers. A decline in personal vehicle ownership of the magnitude Macdonald is describing would put structural pressure on all of those categories. Not immediately, not catastrophically — but persistently, over the same 15-to-20-year window. The political economy of that transition is going to be enormously complicated, particularly in states like Michigan, Ohio, and Indiana where automotive employment is still a significant share of the industrial base.

The insurance industry is another sector that doesn't come up enough in these conversations. Personal auto insurance is a roughly $300 billion annual market in the United States. If liability in an AV world shifts from the individual owner to the vehicle manufacturer or the platform operator — which is the direction the legal consensus is moving — that market doesn't disappear, but it transforms completely. Personal auto policies become product liability coverage. That's a completely different business model for insurance companies.

And then there's the question of who captures the economic value of all those reclaimed hours. The average American spends about 273 hours a year driving. In an AV world, that becomes productive or leisure time. Some of that value will be captured by the AV platforms through in-vehicle advertising, entertainment subscriptions, or productivity services. Some of it will diffuse to workers who become more productive during their commutes. The distribution of that reclaimed time value is actually one of the more interesting economic questions of the next decade, and almost no one is seriously modeling it yet.

I want to add one more dimension, which is the global south angle. The AV revolution as currently envisioned by Uber, Waymo, and Renesas is being built for dense, well-mapped urban environments with reliable wireless connectivity and clear road markings. Lagos, Cairo, Mumbai, and Nairobi — cities that are growing faster than any city in the developed world — have urban environments that are dramatically harder for current AV systems to navigate. The risk is that the AV transition creates a two-tier global mobility system: wealthy, planned cities get autonomous fleets that reduce costs and improve safety, while the cities with the greatest transportation needs and the least infrastructure remain dependent on human drivers for much longer.

That's a point that doesn't get nearly enough attention in the American and European press coverage of AV technology. The companies making these investments are rationally focused on the markets where the technology works today and where the regulatory environment is navigable. But the humanitarian case for autonomous mobility — reduced traffic fatalities, lower transportation costs for low-income workers — is strongest in places where the technology is furthest from deployment. That gap between where AV tech is needed most and where it's being built first is a real tension in this story.

---

Market Context — Reading the S&P Futures Signal on a Sunday

Let's take a few minutes to situate all of this in the broader market context, because the financial backdrop matters for how these technology narratives play out. S&P futures are sitting at 7,722 this Sunday morning, down about 32.75 points or 0.42 percent from Friday's close of 7,754.75. That's a modest softening, not a dramatic selloff, but it's worth understanding what's driving the mild risk-off tone heading into next week.

The broader narrative through most of August was that the market had digested the Fed's posture on rates fairly well — inflation data has been trending in the right direction, and the labor market, while softer than 2024's peak, hasn't cracked. But September historically tends to be the weakest month of the year for equities, and traders are positioning accordingly. A half-percent dip in S&P futures on a Sunday morning is not alarming; it's seasonal texture.

For the AV and tech sector specifically, the macro environment creates an interesting dynamic. These are long-duration investments — the cash flows that justify Uber's $119 price target from BMO, or that justify Renesas's investment in R-Car Gen 5 silicon, are largely expected to materialize in the 2030s. Long-duration assets are disproportionately sensitive to interest rate expectations. If rates stay higher for longer than the market currently prices, the discounted value of those future cash flows comes down, and AV-narrative stocks face more headwinds than their near-term earnings would suggest.

On the other hand, the Uber AI cost efficiency story we discussed earlier — the tenfold usage surge with stable spending — is exactly the kind of evidence that supports near-term earnings credibility. If Uber can demonstrate that it's managing its cost structure intelligently even as it invests in AV infrastructure, it becomes easier for analysts to maintain bullish targets even in a rate environment that's less than ideal. The operational discipline narrative and the long-term AV narrative are actually reinforcing each other for Uber specifically right now.

For retail investors trying to make sense of all this: the Uber-AV story is a genuine long-term thesis that deserves serious attention, but the 15-to-20-year timeline for meaningful car ownership displacement means that near-term stock performance will be driven more by quarterly execution — driver supply, rider demand, cost management — than by progress on AV deployment. The two time horizons exist simultaneously, and confusing them leads to bad investment decisions in both directions.

The Renesas piece is harder for most retail investors to access directly since it trades primarily on the Tokyo Stock Exchange, but it's worth watching as an indicator of how the automotive chip supply chain is positioning. Japanese automakers — Toyota, Honda, Nissan — are some of Renesas's largest customers, and if those OEMs start moving toward Autoware-based AV systems built on R-Car silicon, that's a revenue mix shift that would show up in Renesas's margins over a 3-to-5-year window.

---

IDF Strikes in Southern Lebanon — Regional Escalation and the Broader Middle East Picture

We've been focused on the AV and technology story all morning, but we need to acknowledge the breaking news coming out of the Middle East this Sunday. The IDF began a wave of strikes in southern Lebanon on Sunday morning, according to Israeli and Lebanese media. This follows a drone attack attributed to Hezbollah. Lebanese media are reporting that one of the buildings struck served as a hospital, with at least two people reportedly killed. This is an active and developing story, and we're working with early-stage reporting, so I want to be careful about overstatement here.

The context for this exchange matters. The relationship between Israel and Hezbollah has been in an uneasy phase since the conclusion of the last major conflict cycle, with periodic skirmishes across the Blue Line — the UN-demarcated boundary between Israel and Lebanon — continuing on a low-level basis. Each escalation carries the risk of triggering a response cycle that exceeds either side's intent to control. A Hezbollah drone attack followed by IDF airstrikes in southern Lebanon is not, unfortunately, an unusual sequence, but the involvement of what Lebanese media are describing as a hospital raises the diplomatic stakes considerably.

The international law dimension here is one that Sarah knows well. Strikes on medical facilities are among the most heavily scrutinized actions under the laws of armed conflict, even when military justifications are offered. Israel has in the past argued that Hezbollah uses civilian infrastructure — including medical facilities — to store weapons or shield military operations, which under the laws of war would affect the protected status of those structures. But the burden of proof for that justification is high, and the reputational and diplomatic costs of a hospital strike are substantial regardless of the legal argument.

The United States, the European Union, and the United Nations will all be watching the casualty figures and the IDF's stated justification closely. The current U.S. posture toward Israel has been one of continued security support while expressing private concern about civilian casualties — a balance that becomes harder to maintain as incidents like this accumulate. What happens in the next 24 to 48 hours — whether this remains a single exchange or triggers a broader escalation — will determine whether this becomes a multi-day international story or remains a localized incident.

For listeners trying to understand the broader regional architecture: Hezbollah's military capacity has been a contested variable since the 2024 conflict. Some assessments suggest the organization has substantially rebuilt its drone and rocket capabilities; others suggest significant degradation. Sunday's drone attack, if confirmed, would suggest their long-range strike capacity is more intact than some intelligence assessments had indicated. That's a significant data point for Israeli defense planners, and it likely informs the scale and targeting of the IDF response we're seeing this morning.

We'll continue tracking this story as reporting develops. It's a reminder that even in a week dominated by technology and business narratives, the physical security landscape in the Middle East remains the most immediate daily driver of geopolitical risk for the global economy — given the region's centrality to energy markets and shipping routes.

---

The Convergence Question — And What If We're Wrong About AV Timelines?

I want to spend this segment on the broader convergence question — what happens when all four of these threads we've been discussing today actually come together. Uber's platform strategy, Renesas's open-source hardware push, the AI cost efficiency curve, and the COO's 20-year prediction. If you believe all of them simultaneously, you get a picture of a transportation system that looks radically different from today's. But I think it's worth being precise about which of these claims we're most confident in — and then genuinely interrogating the weakest links.

Let me rank them by confidence level as I see it. The AI cost efficiency story — tenfold usage growth with stable spending — is the most near-term and verifiable claim. Uber reported it, we can track it quarter to quarter, and it's consistent with trends we're seeing across enterprise AI adoption more broadly. High confidence that this is real and directionally sustainable.

I'd agree. The Renesas-Autoware hardware integration story is also relatively high confidence, in the sense that the partnership exists, the demonstration is happening next week, and the technical merits of co-designed hardware and software are well established. The question isn't whether the collaboration is real — it is — but whether R-Car Gen 5 will actually become the dominant AV silicon in the way Renesas is positioning for. That's a competitive outcome, not a technical fact.

The Uber platform thesis — that they become the preferred distribution layer for AV manufacturers — I'd call medium confidence. The logic is sound, the precedents from other platform markets are supportive, but there are real competitors pursuing full vertical integration, and the regulatory risk we discussed is genuinely uncertain. A single major antitrust action could restructure this market in ways that the BMO price target doesn't reflect.

And then there's the 20-year car ownership prediction from Macdonald. This is the one I want to spend some time really interrogating, because it's also the claim that's doing the most work in the industry's public narrative right now. Everyone seems to be agreeing with the direction if not the timeline, and I think that consensus deserves pushback.

So let's do the "What If We're Wrong?" exercise. The confident prediction on the table is this: autonomous vehicles will make personal car ownership economically irrational for a majority of Americans within 15 to 20 years, driving a fundamental restructuring of transportation, insurance, real estate, and manufacturing. What's the strongest case that this is wrong?

The strongest counterargument is probably cultural and geographic rather than technical. Car ownership in America is not primarily an economic optimization problem. It is a cultural artifact, a freedom technology, and in large portions of the country a practical necessity that no urban-optimized AV fleet can substitute for. The Ford F-150 has been the best-selling vehicle in America for 44 consecutive years. The people buying F-150s are not doing so because they've done a rigorous cost-per-mile analysis and concluded it beats Uber. They're doing it because they haul things, because they live on land where a ride-share wait time would be 45 minutes, because the truck is part of their identity.

That's the cultural vector. The economic vector counterargument is that the total cost of ownership of a personal vehicle — even sitting parked 98 percent of the time — may actually be competitive with AV mobility for suburban and exurban families that make multiple daily trips in multiple directions simultaneously. The Macdonald pitch works perfectly for a single urban professional who commutes in one direction once a day. It works much less well for a family of four with two parents running kids to three different activities, grocery shopping, and visiting aging parents in a suburb with no shared mobility density.

The third failure mode for this prediction is regulatory capture in reverse — not the companies capturing regulators, but the existing automotive and fuel ecosystem using political power to slow AV deployment. Internal combustion vehicle manufacturing supports millions of jobs in politically decisive states. Auto dealer associations have already successfully lobbied against direct-sale models at the state level in multiple states. Organized labor in the trucking and taxi industries has real political leverage. The technological capability to displace personal car ownership may arrive well before the political permission to do so at scale.

So what's the signal to watch for? If we're wrong about the 15-to-20-year timeline, the early indicators will appear in specific data: personal vehicle registration numbers in dense urban markets — New York, San Francisco, Boston — should show meaningful year-over-year decline by 2029 or 2030 if the AV substitution is happening at the pace implied. If those numbers are flat or still growing in 2030, it's strong evidence that the transition is slower than Macdonald's framing suggests.

Additionally, watch the used car market. AV adoption at scale would create a sustained inventory surge in used vehicles as urban owners exit the market. If used car prices remain strong into 2030 and beyond, that's a market signal that the substitution is not materializing at the predicted pace. Conversely, a sharp and sustained decline in used vehicle prices in cities where AV coverage is most dense would be early evidence that the thesis is tracking.

The honest answer is that Macdonald is probably right about the direction and probably optimistic about the pace, and both things can be true simultaneously. A 30-to-35-year timeline for the same transformation would be more consistent with how infrastructure-dependent technological transitions have historically unfolded. That's still a profound change — it's just not a change that today's 40-year-old driver needs to plan for in the next budget cycle.

---