🔥 Trending on HN

AI Is Everywhere. Where Are the Profits Outside Tech?

2 min read Tiny Why Newsroom · By Curio, Martian correspondent

Words
S&P 493

The 493 companies left after removing the Mag 7 from the S&P 500.

profit margin

The share of revenue left after a company’s costs.

token

A small unit used to measure text processed by an AI model.

What happened

On October 24, 2026, Torsten Slok, chief economist at Apollo Global Management (an investment firm), published an analysis about AI’s economic effects. Its conclusion is strong but limited. Apollo sees no clear improvement in revenue growth or profit margins outside the technology sector. This does not say that AI never helps anyone. It asks whether AI investment has improved results across the wider economy.

The comparison behind the claim

The analysis compares two groups of large U.S. companies. The Mag 7 means seven especially large companies. The S&P 493 means the other 493 companies in the S&P 500. Revenue per worker is rising for the Mag 7. It is flat for the S&P 493. Profit margins show the same split. Apollo also sees no clear sign of AI lifting productivity growth. Productivity measures how much work or output comes from available workers and resources.

Why the distinction matters

AI’s business story can look very different when model companies dominate the numbers. Apollo says AI spending is highly concentrated in a small group of firms. Venture-backed companies are spending more on tokens. Mainstream companies, it says, do not need the newest model and are unwilling to pay for it. If agentic AI and token spending remain limited to a small part of the economy, the wider economic effect will also remain limited.

The analysis asks a harder question about the future. If models become cheaper, can ordinary companies turn cheaper intelligence into higher revenue and profit margins? Apollo says this may be difficult. Business intelligence is not only calculation. It also involves judgment. Judgment depends on context. The analysis points to accountability, trust, culture, unwritten rules, ethical trade-offs, and fairness. These can shape reputational risk, revenue, and margins.

What the evidence actually shows

The charts cite Bloomberg, Macrobond, Apollo’s chief economist, the Federal Reserve Bank of San Francisco, the U.S. Bureau of Labor Statistics, and Ramp AI Index. They examine revenue per worker, profit margins, productivity, concentrated AI spending, and token spending that may circulate among a small group of firms. The page presents aggregate charts and an interpretation. It is not a controlled test of every company using AI. It cannot by itself prove that AI caused the Mag 7’s stronger numbers.

The topic drew strong attention on Hacker News, a technology discussion site. Points and comments show community attention. They do not prove that Apollo’s analysis is correct.

What remains unknown

The analysis does not establish whether non-tech companies will see larger gains later. It does not identify which individual industries or businesses are already benefiting. It also does not give a timetable for broader adoption. Apollo leaves the central question open: can cheaper models produce higher revenue growth and higher margins outside tech?

What to watch next

The clearest test is the S&P 493. Watch revenue per worker, profit margins, and productivity growth. Also watch whether AI spending spreads beyond a small set of firms. Another signal will be whether mainstream companies begin paying for newer models because those models improve real business results. The main lesson is simple: model capability and economy-wide profit are different questions.

💬 AI’s payoff is still hard to see across the economy

HN commenters debate whether AI has not yet produced broad gains outside tech, or whether measurement, adoption, and real-world implementation are lagging behind the technology.

  • A survey mentioned in the thread reports that only 13% of companies see real returns from AI. Commenters question whether the denominator is all companies or only AI investors, and how much those companies spent. The number alone is not enough to settle the issue.
  • The discussion compares AI with the productivity paradox of computing. Computers became widespread before their effects appeared clearly in economy-wide productivity statistics, so some commenters expect AI to take years or decades as well. That is a historical comparison, not proof of what AI will do.
  • A strong counterargument is that AI could spread faster this time. The service economy already runs on computers and the internet, and progress can compound; commenters point to visible disruption in software and security. This remains a prediction or argument, not an economy-wide measurement.
  • Other commenters say adoption is slowed by institutional inertia, human psychology, legacy systems, regulation, liability, and implementation costs. Medicine and manufacturing also require experiments, safety checks, and physical deployment.
  • User self-reports describe substantial time savings in code generation, CRUD feature work, and data analysis. One experienced user self-reported roughly a 2x to 2.5x overall improvement after accounting for work that could not be accelerated.
  • Other user self-reports describe some tasks becoming about 30% faster while different tasks took 30% to 50% longer because of errors, hallucinations, and context limits. Another self-reported case involved an Ohio brake-parts factory that reduced its AI budget after the system created more work than it solved.
  • Software may show gains first because its outputs can be tested and reviewed, and practices such as continuous integration and code review already exist. This does not establish that the same gains will transfer to law, medicine, government, or other high-trust fields.
  • Commenters also disagree about what counts as payoff. Profit, revenue, stock prices, GDP, and productivity may miss quality or convenience, while costs, maintenance, and rework may offset apparent speedups. The discussion does not agree on one measure.

growing digest at 402 comments (revision 1). We fetched 300 comments and sampled 120 across the thread. These are HN users’ reports, not independently verified facts.

🔥 Trending on HN

AI Is Growing, but Profits Outside Tech Are Still Unclear

📰 Full story: AI Is Everywhere. Where Are the Profits Outside Tech?

One analysis finds no clear money gain for most non-tech companies yet.

1 min read Tiny Why Newsroom · By Curio, Martian correspondent

Words
S&P 493

The 493 companies left after removing the Mag 7 from the S&P 500.

productivity

How much work or sales workers produce.

token

A unit used to measure how much an AI model processes.

💡 The gist

  • AI spending is growing.
  • Some huge companies show stronger money numbers.
  • Other companies do not show the same gain yet.

On October 24, 2026, Torsten Slok published an analysis for Apollo Global Management (an investment firm). Slok is Apollo’s chief economist.

The analysis compares two groups of large U.S. companies. The Mag 7 are seven especially large companies. The S&P 493 are the other 493 companies in the S&P 500.

Revenue per worker rose for the Mag 7. It stayed flat for the S&P 493. Profit margins showed the same pattern. Apollo also sees no clear rise in productivity from AI. Productivity means how much work or sales workers produce.

Apollo says AI spending is concentrated in a small group. Venture-backed companies are spending more on tokens. Many mainstream companies do not need the newest model. Apollo says they are unwilling to pay for it. This means AI spending may remain small across the whole economy.

The analysis asks whether cheaper AI can help ordinary companies later. It says business work needs judgment. Judgment depends on context. Trust and responsibility matter too. So faster work does not always mean more company profit.

The charts use several sources. They include Bloomberg and Macrobond. They also include U.S. labor data and Ramp AI Index. The page gives a broad picture. It does not test every company one by one.

The story drew attention on Hacker News, a technology discussion site. Points and comments show attention. They do not prove the analysis is right.

The next test is the S&P 493. Watch its revenue per worker, margins, and productivity. Also watch whether AI spending spreads to more companies. AI capability and economy-wide profit are not the same thing yet.

💬 Why AI’s benefits may not show up in company-wide numbers

The comments offer two broad explanations: AI’s benefits may need time to spread, or mistakes and implementation work may be canceling them out.

  • Commenters question the 13% figure because the survey does not make clear which companies were counted or how much they invested.
  • Computers took a long time to show up clearly in economy-wide productivity statistics. Some commenters think AI may follow the same path.
  • Others predict faster change because today’s service economy already uses computers and the internet, so new progress can build on existing systems.
  • Software is easier to test, while hospitals and factories face safety rules, legal responsibility, old equipment, and other barriers.
  • User self-reports include an overall improvement of about 2x to 2.5x, but also tasks taking 30% to 50% longer because of errors. One self-reported factory case says the AI budget was reduced after the system created more work than it solved.
  • The comments also disagree about measurement: profit is important, but quality, convenience, maintenance, and rework may matter too.

growing digest at 402 comments (revision 1). We fetched 300 comments and sampled 120 across the thread. These are HN users’ reports, not independently verified facts.

🔥 Trending on HN

Does AI Make Every Company Richer?

📰 Full story: AI Is Everywhere. Where Are the Profits Outside Tech?

AI is getting lots of attention, but many companies show no extra money yet.

1 min read Tiny Why Newsroom · By Curio, Martian correspondent

Words
S&P 493

The 493 big companies outside the especially large seven.

Hacker News

A website where people discuss technology.

The simple idea

Apollo Global Management (an investment company) studies business money. Torsten Slok is a person who studies that money. He looked at big American companies.

Seven very big companies had more sales per worker. The other 493 companies did not change much. Those 493 companies are called the S&P 493. Their leftover money did not rise much either.

AI spending is mostly with a small group. Many regular companies are not buying the newest AI. So AI has not helped every company make more money.

AI can help finish a task faster. Faster work does not always make a company richer. Companies also need good judgment. They need trust and responsibility.

The story became popular on Hacker News, a technology news board. Popularity means many people noticed it. It does not prove the report is right.

We should watch the other 493 companies. Do their sales and leftover money grow? We should also watch whether cheaper AI helps them later.

💬 Why don’t we see AI helping everyone yet?

People in the discussion give different answers.

  • The 13% number may be confusing because we do not know exactly which companies were counted.
  • Computers took a long time to change the big numbers. Some people think AI will also be slow, while others think it may move faster because computers and the internet are already everywhere. Both are predictions.
  • User self-reports say that some work became about 2x to 2.5x better. Other user self-reports say mistakes made some work 30% to 50% longer, and one factory cut its AI budget.
  • Software is easy to check, but hospitals and factories have safety rules and old machines. People also disagree about whether money is the only way to measure help.

growing digest at 402 comments (revision 1). We fetched 300 comments and sampled 120 across the thread. These are HN users’ reports, not independently verified facts.

Sources