# AI’s Bill Is Getting Bigger. Can the Business Results Keep Up?

> AI companies are raising billions to expand. Faster loans, useful business software and paid robotaxi rides show where the money could come back.

By Lidia Yadlos · October 26, 2026

Canonical: https://blockster.com/ai-investment-anthropic-meta-deutsche-bank-oracle

Someone has to pay for AI's enormous computing bills. Eventually, that money has to come from customers who find the technology useful enough to buy.

That is the connection between this week's announcements from Anthropic, Meta, SpaceX, Tesla, Deutsche Bank and Oracle. Some are raising money to build more AI. Others are showing where people and businesses might actually use it.

Anthropic is preparing to sell shares to the public. Meta has outlined a financing plan worth as much as $95 billion. SpaceX has disclosed a roughly $40 billion arrangement to finance AI equipment. Meanwhile, a bank says AI is speeding up lending decisions, Oracle is putting it into everyday business software, and Tesla is bringing it into vehicles and robots.

**The next test is whether customers will pay enough—or save enough—to make all that spending worthwhile.**

For investors, this matters because growing demand for AI does not guarantee that every AI company will make money. For businesses and their customers, the payoff could be simpler: less time waiting, fewer repetitive tasks and services that cost less to deliver.

## The Money Going Into AI

Anthropic, the company behind Claude, offers a clear example of both the demand and the expense.

Its [IPO filing](https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm) shows roughly **$16.3 billion in revenue for the first half of 2026**, up from about $1.2 billion a year earlier. Customers are already spending substantial amounts on its products.

The same filing reports an **$80.7 billion net loss**, but that number needs explaining. Around **$72.4 billion** came from accounting changes to financing agreements, including debt that can become shares. Those changes do not mean the company spent that amount in cash during the period.

Its operating loss—the loss from running the business before financing and other items—was **$693 million**. Computing and infrastructure expenses were approximately **$8.9 billion**.

The question is whether customer revenue can grow faster than the cost of providing the service. More customers help, but serving them also requires computing power.

[Reuters reports](https://www.reuters.com/legal/transactional/anthropic-opens-books-ais-explosive-growth-staggering-costs-2026-10-26/) that Anthropic plans to list on Nasdaq under **ANT**. It has not yet disclosed how many shares it will sell or the proposed price. Investors therefore still lack key details needed to judge the offer.

Some existing backers want to invest more. On October 26, Brad Gerstner said Altimeter intended to participate substantially in the IPO, adding to its existing ownership. That expresses its plans; the final investment has not been confirmed.

[Post from @altcap on X](https://twitter.com/altcap/status/2114710064099364965)

## Meta Wants AI to Grow Its Advertising Business

Meta is raising money from a different starting point: it already has a large advertising business.

Its [financing plan](https://investor.atmeta.com/investor-news/press-release-details/2026/Meta-Announces-Launch-of-Offering-of-Depositary-Shares-Representing-Series-A-Mandatory-Convertible-Preferred-Stock-and-Offering-of-Class-A-Common-Stock-in-Connection-with-Forward-Sale-Agreements/default.aspx) includes approximately **$35 billion in offerings launched now** and a proposed **$60 billion share-sale program later**. The later program is not finalized, and some proceeds from the launched offerings would arrive in the future. Meta has not collected $95 billion in new cash.

The company expects **$135 billion–$145 billion in spending on long-term assets this year**, including the infrastructure supporting its AI plans. It says AI is already improving advertising results and keeping users engaged.

That gives Meta a way to earn money from AI without charging everyone for a chatbot: better advertising can help its existing business. But issuing more shares also means existing shareholders own a smaller percentage of the company. The new investment needs to produce enough growth to make that trade worthwhile.

## Where SpaceX and Tesla Fit In

SpaceX is another example of how expensive AI computing has become.

Under the [arrangement disclosed on October 26](https://www.reuters.com/business/media-telecom/spacex-enters-40-billion-deal-with-apollo-ai-chip-financing-2026-10-26/), a company managed by Apollo would buy AI chips and servers and lease them to SpaceX. SpaceX's payments toward the equipment would total approximately **$40 billion**, before financing costs. The deal is expected to close by year-end, with leases starting as equipment arrives in the first half of 2027.

This lets SpaceX pay over time. The equipment still needs to earn enough to cover those payments, along with electricity, maintenance and other running costs. Access to funding helps it expand; customer demand determines whether the expansion pays off.

Tesla shows a different use for AI: vehicles that carry passengers and robots that perform physical tasks.

In its [October 21 update](https://assets-ir.tesla.com/tesla-contents/IR/TSLA-Q3-2026-Update.pdf), Tesla said it had added Cybercab vehicles to its Austin Robotaxi fleet and begun producing Optimus robots in Fremont. For these businesses to succeed, customers need a reason to pay for the service or the machine.

A robotaxi must bring in more from rides than it costs to operate. A business buying a robot needs it to do useful work reliably, at a price that makes sense. Those are the results that would turn Tesla's AI plans into a stronger business.

Tesla spent **$4.7 billion on long-term assets in the quarter**, more than double the year-earlier amount. Profit attributable to common shareholders fell **43% to $787 million**. These figures cover the whole company; they do not show how much Tesla spent specifically on AI or establish that AI caused the profit decline.

They do show why the timing matters: Tesla is investing heavily while investors wait to see how much its newer businesses can contribute.

## What Customers Actually Get

Deutsche Bank offers a more immediate example of AI's usefulness.

Technology chief Marie-Jeanne Deverdun told [Reuters](https://www.reuters.com/business/finance/ai-cuts-some-deutsche-bank-corporate-loan-decisions-minutes-2026-10-26/) that a project introduced this summer had cut some simpler business-lending decisions **from weeks to minutes**. Some decisions for multinational companies had shortened from months to weeks. This applies to selected lending processes, not all loans.

For a company waiting to buy equipment or stock its shelves, getting an answer sooner can help it plan. For the bank, handling applications faster could reduce the work involved in each decision.

The bank still needs to make sound loans. Faster processing is useful only if errors and unpaid loans do not wipe out the savings.

Oracle is aiming at similar improvements across everyday office work. Its October 26 announcements highlighted **25 available Fusion Claw-powered applications**, building on the product's [September introduction](https://www.oracle.com/news/announcement/oracle-extends-fusion-agentic-applications-with-introduction-of-fusion-claw-2026-09-29/).

The [tasks](https://www.oracle.com/latam/news/announcement/ai-world-oracle-advances-outcome-driven-ai-across-fusion-applications-2026-10-26/) include checking accounting records, spotting payroll changes, reviewing purchasing costs and planning how to use limited supplies. New tools called Fusion Companion and Fusion Agentic Home are designed to let employees describe what they need done and have the software organize the steps.

These are jobs businesses already spend money on. If AI reduces the time and cost involved, there is a clear reason to use it. Customers still need to measure the savings after paying for setup, supervision and corrections.

They also need control over what the software can do. Oracle executive Chris Leone explained that point in the company's [announcement](https://www.oracle.com/latam/news/announcement/ai-world-oracle-introduces-new-ai-powered-ways-to-execute-work-across-oracle-fusion-applications-2026-10-26/):

> “AI is changing where enterprise work begins, but organizations need to maintain control over how that work gets done.”
>
> — Chris Leone, Executive Vice President, Applications Development, Oracle

As [Blockster's coverage of Citi and Bank of America](https://blockster.com/citi-bank-of-america-ai-agents-account-access) explored, letting AI help with financial work is different from giving an outside assistant permission to move money. Businesses need clear limits and people responsible for important decisions.

## Who Benefits When AI Saves Money?

The savings will not necessarily go to the company building the AI. A bank could lower its costs. A software customer could negotiate cheaper prices. A model developer could sell more services while spending much of the extra revenue on computing.

Workers have a stake in this, too. Faster software could free people to handle more valuable work, but it could also reduce the number of people needed for certain tasks.

That is part of the reasoning behind [Anthropic's proposed Public Benefit Fund](https://blockster.com/anthropic-public-benefit-fund-5-percent-shares). Its October 18 announcement pledged an initial **5% of shares**, subject to shareholder approval, with possible further contributions as the company grows. Supporting people affected by changes in employment is one potential use. The pledge does not mean workers have already received payments.

[Post from @AnthropicAI on X](https://twitter.com/AnthropicAI/status/2111881963657412903)

For readers following AI as an investment, the useful evidence is becoming more specific: customers who keep paying, services that cost less to provide, and profits that improve as the business grows.

Anthropic's sales, Meta's advertising results, SpaceX's computing business, Tesla's paid rides, faster bank decisions and useful office software all offer ways to test the promise. They are at different stages, and none guarantees the others will succeed.

**The funding shows how big the AI bet has become. The next chapter is about customers: what they use, what they pay for, and whether enough money is left over to make a profit.**