Back to blog

Daily Intelligence

AI Is Generating Revenue. Now Comes the Bill for Energy, Capacity and Trust

October 22, 2026 · 8 min read

AI Is Generating Revenue. Now Comes the Bill for Energy, Capacity and Trust

The most revealing story this Thursday does not pit artificial intelligence against the real economy: it shows they are already the same conversation. SpaceX earns more than half its revenue from its AI division, while oil once again reminds us that digital infrastructure depends on an unpredictable physical world. Meanwhile, a multibillion-dollar judgment against Meta brings another raw material back into focus: trust. Today's argument is straightforward: growth is not enough; what matters is what remains after paying for capacity, energy and accountability. This October 22 edition uses information available by 7 a.m. Madrid time and does not present developments that have yet to happen as today's completed market results.

Macro/Energy: risk does not vanish because oil is still flowing

Bloomberg's overnight coverage, visible through its indexed excerpt, describes oil advancing after deadly attacks in Saudi Arabia raised concerns about supply. That excerpt identifies the concern; it does not provide enough detail to reconstruct every event or offer a live quotation. For the economy, the important distinction is between barrels missing today and the risk of barrels missing tomorrow. Both can raise contract, shipping and insurance costs, but they need not last equally long or require the same response. In an interview published this morning by CNBC-TV18, Julius Baer's Mark Matthews argues that geopolitical price spikes tend to moderate and that the AI investment cycle retains fundamental support. That is a market opinion, not a guarantee. If relief comes through normalizing supply, the effect is constructive; if it comes because expensive energy destroys demand, falling crude prices can coexist with a weaker economy. For a business, this distinction matters more than correctly predicting the next movement in a barrel's price.

Geopolitics: what a promise of stability is worth

Reuters reported yesterday that most Gulf stock markets retreated as diplomacy between Iran and the United States stalled. That signal and Bloomberg's overnight headline do not establish that the entire region is moving in a single direction; they do explain why markets continue paying for protection. Announced talks are not equivalent to safe shipping routes, and an open route does not automatically remove additional insurance and transport costs. The practical implication is to separate three levels: what governments say, what happens to physical flows and what companies ultimately pay. Confusing them encourages premature celebration of an easing in tensions or indefinite extrapolation of the worst case. For Europe, dependent on imported energy, the useful question is not simply whether oil is available, but how much delivery costs and how reliably it can be contracted. Technology is not insulated either: building computing capacity requires equipment, logistics and electricity on timelines much less flexible than political rhetoric.

AI/Tech: SpaceX begins to look like a cloud provider

Xataka's feed opens this morning with a figure that changes how a company should be understood: SpaceX's AI division accounts for approximately 52% of quarterly revenue. Its report, drawing on corporate results and Reuters, puts total third-quarter revenue at about $14.6 billion and AI revenue at about $7.6 billion. Reuters' indexed reporting also confirms the strong rise in revenue and investment. These results were announced on Tuesday; they are not an accounting surprise first disclosed today. The decisive distinction is what the company sells: according to Xataka's coverage citing Reuters, much of the business comes from renting data center capacity to other developers. This is not evidence that a single model has conquered every application. It is evidence of infrastructure demand. For enterprise customers, it could broaden supply alternatives; for investors, it calls for examining contracts, utilization and customer concentration, not merely a chatbot's demonstrations. Those are different questions from who currently leads a benchmark.

The other half of that story is the bill. Xataka reports quarterly capital expenditure of approximately $19.6 billion, with about $16.4 billion directed toward AI. Revenue and investment are not interchangeable accounting measures: capex is not deducted in full from the same quarter's profit, but it does require financing and consumes resources before earning a return. Nor does contracted revenue ensure that every machine will remain profitable for years. Lower computing rental prices, faster hardware obsolescence or delayed electricity connections would change the calculation. The lesson is not that growth is imaginary, but that it deserves better questions. How much cash does each additional unit of capacity generate? How long do customer commitments last? Who bears energy costs? The machinery business can be excellent without resembling software sales, and valuing both with the same intuition can be expensive. The timing of cash receipts and obligations matters alongside the apparent size of the opportunity.

Meta offers the counterpoint: not every technology bill sits inside a data center. Reuters reports a $25 billion court order in New Mexico concerning misleading privacy statements, in a case arising from the Cambridge Analytica scandal. NBC's coverage and Xataka corroborate the story; Xataka also reports the company's intention to appeal. A judgment exists, but there is neither a definitive economic outcome nor a basis for assuming that a full payment has already happened today. That distinction matters when interpreting earnings, provisions and valuations. Beyond Meta, the case connects governance and strategy: public commitments about data can have financial consequences years later. For any business introducing assistants and automation, knowing what information is collected, who receives it and how this is explained to customers is not merely a legal appendix. It is part of the product and of the trust that allows a company to keep selling it. An impressive technical capability does not make those obligations disappear.

The most immediately useful technology signal comes from security. Xataka describes a tool published by OpenAI for checking PHP versions in response to critical vulnerabilities. Its value lies partly in its limits: an external query can identify information exposed by a server, but may return UNKNOWN when information is insufficient. That means neither secure nor vulnerable. Even an apparent version requires comparison with the installed package and vendor advisories, because fixes may have been backported. The business lesson is not to install every new tool in a hurry, but to demand evidence appropriate to each decision. Inventorying services, confirming which installation actually handles requests and reviewing updates may deliver more immediate value than another automation announcement. This is not an audit of any specific website: it is an example of useful AI distinguishing between something actually checked and information that simply could not be found. A reassuring label is less valuable than understanding what a check can and cannot establish.

Markets: profit and cash tell different stories

SpaceX's results, Meta's litigation and energy risks pose the same question from three angles: what is growth worth after adjusting for the cost of sustaining it? This is analysis, not a description of today's European stock market opening. Matthews' optimistic interpretation provides a necessary counterweight: AI investment can translate into real productivity and earnings. But a valuable technology does not make every price paid for it attractive, or ensure equal rewards for every supplier. Four elements should be connected when reading accounts: revenue, operating margin, investment and cash flow. Rising profit can coexist with substantial financing needs; a legal obligation can be material without immediately becoming a cash outflow equal to its headline amount. The objective is not to distrust everything, but to prevent spectacular numbers from replacing analysis. Comparing equivalent periods also matters, as does distinguishing management forecasts from results already delivered. Strong demand today is useful evidence, but it does not settle the economics of capacity that will only become available later.

The next 24–72 hours: confirmations that could change the story

Through the weekend, the first watchpoint is energy: official communications and evidence concerning facilities, transport and supply continuity. A diplomatic headline deserves greater weight when it changes those conditions, not merely the tone of a press conference. The second is corporate: clarifications about contracted capacity, electricity costs and financing for AI expansion, alongside any verifiable procedural development in Meta's case. The third lies inside organizations: whether reviewing exposed services and vulnerable packages has become a task with an owner and deadline rather than an unexecuted message. These are monitoring priorities, not scheduled events whose publication is assumed. Before reacting to economic data or company results in the coming sessions, verify the official calendar and release time. This edition assigns neither numbers nor surprises to future releases and does not extrapolate Asian trading into US closing prices that remain unknown. The point of a short radar is to identify what evidence would change the argument, not to collect every possible headline.

Conclusion: three scenarios and one controllable decision

In the base case, energy tensions persist without a sustained deterioration in supply and computing demand keeps growing, but costs face closer scrutiny. The implication is to keep plans flexible, review contracts and require measurable returns from AI projects. In the bullish case, verifiable de-escalation reduces logistical risk and companies demonstrate that they can turn capacity into cash: accelerating deployments would make more sense where they already solve problems at known costs. In the bearish case, fresh disruption raises energy and financing costs while investment grows faster than revenue; preserving room to maneuver and staggering difficult-to-reverse commitments would take priority. I assign no numerical probabilities: these are frameworks for testing developments, not forecasts disguised as precision. Today's application is more modest and useful: choose one critical dependency—energy, a supplier, data or security—and check who controls it and what alternative exists if it fails. We cannot dictate the international environment, but we can reduce decisions built on assumptions nobody has reviewed. That is a more durable advantage than reacting fastest to every notification.

Sources and scope

AI Is Generating Revenue. Now Comes the Bill for Energy, Capacity and Trust | Adrian GC | Adrian GC