The AI Investment Reckoning: Uber’s President Says the Bill Has Come Due

On a Tuesday morning that should have been just another earnings cycle footnote, Uber’s president dropped a truth bomb that sent shockwaves through Silicon Valley’s C-suites. The cost of artificial intelligence, once treated as an unlimited expense account for the world’s largest tech companies, is becoming “harder to justify” [1]. This isn’t a startup founder lamenting burn rates or a venture capitalist grumbling about inflated multiples. This is the president of Uber Technologies, Inc.—a company operating in approximately 70 countries and 15,000 cities worldwide, with over 202 million monthly active users and 10 million active drivers and couriers—publicly questioning the ROI on the most hyped technology of the decade [1].

The timing is exquisite. Just days earlier, Pope Leo XIV released his first encyclical, Magnifica Humanitas, from Rome, with the co-founder of Anthropic at his side, calling for AI to be “disarmed” in service of the common good [3]. The Pope chose the language of “disarmament” deliberately, he admitted, “because this moment needs words capable of attracting attention, awakening” [3]. Meanwhile, nuclear startup Deep Fission is attempting to go public again, seeking an IPO that could raise $157 million—a bet that the insatiable energy demands of AI infrastructure will eventually need a new power source [4]. These three narratives, seemingly disconnected, are actually threads of the same story: the AI industry is entering its most uncomfortable phase yet, where promises of infinite returns collide with the hard mathematics of balance sheets, moral philosophy, and physics.

The Math That Keeps CFOs Awake at Night

Let’s start with the numbers that matter. Uber’s president didn’t just make a vague statement about belt-tightening. The phrasing—“harder to justify”—is a carefully calibrated piece of executive communication [1]. In corporate parlance, this signals that the internal models used to calculate return on invested capital for AI initiatives no longer produce the optimistic projections that justified blank-check spending in 2023 and 2024.

Consider what Uber actually does with AI. The company’s core business—ride-hailing, food delivery via Uber Eats, courier services, and freight transport—runs on a massive optimization engine that matches supply with demand in real-time [1]. This is not speculative AI research; it is applied machine learning that directly impacts whether a driver shows up in six minutes or sixteen. Uber has pioneered dynamic pricing algorithms, route optimization, and demand forecasting. If its president says even these proven, revenue-generating applications of AI are becoming harder to justify, the implication for companies with less mature AI deployments is dire.

The problem is structural. The cost of training and deploying large models has not followed the exponential decline curve that many investors assumed. While inference costs have dropped for some architectures, the frontier models—the ones that actually deliver step-change improvements in capability—require exponentially more compute, data, and energy. The open-source LLMs ecosystem has democratized access to smaller models, but the race for state-of-the-art performance remains a game only the wealthiest can play. Uber, despite its massive scale and 202 million monthly active users, now signals that it has hit the point of diminishing returns [1].

This is not a confession of failure. It is an admission of maturity. The low-hanging fruit has been picked. The next tranche of AI improvements requires investments in infrastructure—custom silicon, data center capacity, energy procurement—that do not map neatly onto quarterly earnings targets. When a company like Uber, which processes millions of transactions daily and has a direct financial incentive to squeeze every ounce of efficiency from its algorithms, says the math isn’t working, the entire industry should take notice.

The Vatican Weighs In: A Moral Framework for a Reckoning

It would be easy to dismiss the Pope’s encyclical as a religious document with limited relevance to the boardroom. That would be a mistake. Magnifica Humanitas arrives at a moment when the tech industry desperately searches for a narrative that justifies its continued spending [2]. The Pope’s critique is not Luddite rejectionism; it is a sophisticated argument about power concentration.

The encyclical decries “the concentration of technological power in a few global players” [2]. This is not abstract theology. It directly comments on the market structure of the AI industry, where a handful of companies—Google, Microsoft, OpenAI, Anthropic, Meta—control the foundational models, the compute infrastructure, and the data pipelines. When Uber’s president says AI spending is getting harder to justify, he implicitly acknowledges that the benefits accrue disproportionately to the platform providers, not the application-layer companies like Uber.

The Pope’s choice to invoke Gandalf and the language of “disarmament” is particularly striking [3]. He admits “the word is strong,” but argues that “this moment needs words capable of attracting attention, awakening” [3]. This calls for a fundamental reorientation of how we think about AI—not as a tool for competitive advantage, but as a technology that must be “disarmed” to serve the common good [3]. For a company like Uber, which has spent years fighting regulatory battles over labor classification, safety standards, and market access, the Pope’s framework offers an alternative vision: what if the goal of AI investment was not maximum efficiency and profit extraction, but equitable distribution of benefits?

The presence of Anthropic’s co-founder at the encyclical’s release in Rome is not incidental [3]. Anthropic has positioned itself as the “safety-first” AI company, building models with constitutional AI principles baked in. The company’s alignment with the Vatican’s message suggests that even within the industry, there is recognition that the current trajectory is unsustainable—morally, politically, and perhaps economically. If the Pope and the AI safety movement converge on the same critique, the window for unfettered AI spending may close faster than most executives realize.

The Energy Elephant in the Server Room

Meanwhile, a nuclear startup called Deep Fission is trying to go public, again, seeking $157 million [4]. The company’s story is a microcosm of the AI industry’s most uncomfortable dependency: the physical infrastructure required to power all this intelligence.

The numbers are staggering. Training a single large language model can consume as much electricity as a small town uses in a year. Inference—the act of actually using the model to generate responses—is even more energy-intensive at scale. Every ChatGPT query, every Uber route optimization, every AI-generated image requires electricity, cooling, and networking hardware. The AI industry has run on a tacit assumption that energy costs will continue to decline, or that renewable energy will scale fast enough to meet demand. Deep Fission’s IPO attempt suggests that the market is beginning to price in the possibility that neither assumption holds [4].

The company is seeking $157 million, which in the context of AI infrastructure spending is pocket change [4]. Google and Microsoft are spending billions on data centers. The fact that a nuclear startup struggles to raise a relatively modest amount tells you something about investor sentiment. The TechCrunch report notes that “investors may have trouble buying the nuclear startup’s story” [4]. If the market is skeptical about nuclear power for AI, it means the industry’s energy problem does not have an obvious near-term solution.

This creates a vicious cycle. AI spending is getting harder to justify because returns are diminishing and costs are rising. One major cost driver is energy. If energy costs continue to rise because AI demand outstrips supply, the ROI calculations get even worse. Uber’s president effectively says that the company has reached the point where the marginal benefit of another AI investment is no longer clearly positive [1]. The energy costs are a significant part of that equation, even if they are not explicitly mentioned in the public statement.

The Application Layer Squeeze

To understand why Uber’s statement is so significant, you have to understand the company’s position in the AI value chain. Uber is not an AI company in the sense that OpenAI or Anthropic is an AI company. Uber is an application-layer company that uses AI as a tool to optimize its core business [1]. This distinction is critical.

The application layer is where the economic pain is most acute. The foundation model companies—OpenAI, Google, Meta—can justify massive spending because they build the platforms that everyone else will use. They sell picks and shovels during a gold rush. But the application-layer companies, the ones actually using AI to run their businesses, must show concrete ROI. When Uber’s president says AI spending is getting harder to justify, he speaks for every company that has invested in AI without seeing a proportional increase in revenue or profit [1].

This is the classic innovator’s dilemma playing out in real-time. The companies that build the infrastructure capture most of the value. The companies that use the infrastructure to improve their existing businesses capture less value, because the infrastructure costs eat into their margins. Uber, with its 202 million monthly active users and 10 million drivers, has the scale to negotiate favorable terms with AI providers [1]. But even that scale is not enough to make the math work indefinitely.

The implication for smaller companies is brutal. If Uber struggles to justify AI spending, what hope does a mid-sized logistics company or a regional delivery service have? The AI industry has sold a vision of democratized intelligence, where every business can benefit from advanced models. The reality, as Uber’s statement makes clear, is that the economics of AI favor the platform providers, not the platform users. The vector databases and retrieval-augmented generation pipelines that promised to make AI accessible to everyone are still useful, but they cannot overcome the fundamental cost structure of the technology.

The Hidden Risk: What the Mainstream Media Is Missing

The mainstream coverage of Uber’s statement will likely focus on cost-cutting and the end of the AI hype cycle. That interpretation is too narrow. The real story is about the structural fragility of the AI industry’s business model.

The AI industry has run on a combination of venture capital subsidies, low interest rates, and the promise of future monopoly rents. The foundation model companies have spent aggressively, not because they are profitable, but because they race to capture market share and establish network effects. The assumption has been that once the market consolidates around a few winners, they can raise prices and extract monopoly profits. Uber’s statement suggests that this assumption may be flawed [1].

If the application-layer companies—the actual customers for AI services—reach the limit of what they can justify spending, then the foundation model companies have a revenue problem. They cannot raise prices indefinitely if their customers already question the value proposition. This creates downward pressure on AI pricing that the industry has not fully accounted for.

The Pope’s encyclical adds a regulatory dimension to this economic pressure [2]. If governments begin to take the “disarmament” framework seriously, the AI industry could face new constraints on how it develops and deploys technology. The concentration of power that the Pope decries is not just a moral issue; it is an antitrust issue [2]. Regulators in the US, EU, and elsewhere already scrutinize AI market concentration. The Pope’s intervention provides moral and intellectual ammunition for those who want to break up or regulate the AI giants.

Deep Fission’s struggles to go public add an energy dimension [4]. If the AI industry cannot secure reliable, affordable energy, its growth will be physically constrained. The $157 million that Deep Fission seeks is a tiny fraction of what the AI industry spends on compute [4]. But the fact that investors are skeptical of the nuclear story suggests that the market does not believe the energy problem has a near-term solution.

The Reckoning Has Begun

Uber’s president did not say that AI is overhyped or that the technology doesn’t work. He said that the spending is getting harder to justify [1]. That is a much more nuanced and dangerous statement for the industry. It means the technology works, but the economics don’t. It means the value is real, but the cost of capturing that value is too high. It means the industry has reached the point where the next dollar of AI investment has a lower expected return than the last dollar.

This is the moment that every technology cycle faces. The early adopters capture outsized returns because they apply the technology to the highest-value problems. As the technology matures, the remaining opportunities are smaller, harder to capture, and more expensive to pursue. The industry must transition from a growth-at-all-costs mindset to a discipline of capital efficiency. That transition is always painful.

The convergence of Uber’s economic realism, the Pope’s moral critique, and Deep Fission’s energy struggles suggests that the AI industry is entering a new phase. The era of blank-check spending is ending. The era of hard questions about ROI, power concentration, and physical constraints is beginning. The companies that survive this transition will be the ones that can answer those questions honestly, not the ones that continue to sell the dream of infinite intelligence at zero marginal cost.

For the rest of us—the users, the workers, the citizens—the question is whether the industry can navigate this transition without breaking the promises that made AI so compelling in the first place. The Pope wants to disarm AI in service of humanity [3]. Uber wants to justify its spending to shareholders [1]. Deep Fission wants to power the future with nuclear energy [4]. These are not contradictory goals, but they are in tension. Resolving that tension will define the next decade of technology, and the outcome is far from certain.

---

References

[1] Editorial_board — Original article — https://www.theverge.com/transportation/937116/uber-ai-investment-hard-to-justify

[2] Wired — What Pope Leo XIV’s First Encyclical Says About the Power of AI — https://www.wired.com/story/what-pope-leo-xivs-first-encyclical-says-about-the-power-of-ai/

[3] Ars Technica — Citing Gandalf, Pope Leo says we must "disarm" AI — https://arstechnica.com/tech-policy/2026/05/citing-gandalf-pope-leo-says-we-must-disarm-ai/

[4] TechCrunch — Nuclear startup Deep Fission says it’s going public, again, and I have questions — https://techcrunch.com/2026/05/23/nuclear-startup-deep-fission-says-its-going-public-again-and-i-have-questions/