Innovation & Tech

AI Software Spending Could Hit $1.23 Trillion by 2035, Model Says

AI Insider's AI Market Model projects end-customer AI software spending at $629 billion to $2.19 trillion by 2035, with a $1.23 trillion base case hinging on three assumptions.

By Amara Osei

4 min read

Where AI Software Spending Could Go by 2035: Three Scenarios and the Assumptions Behind Them
Where AI Software Spending Could Go by 2035: Three Scenarios and the Assumptions Behind ThemTumisu / pixabay

What's News

  • Base case: AI software spending grows from ~$144bn in 2026 to ~$1.23tn by 2035
  • Scenario range by 2035: $629bn (lower) to $2.19tn (higher) — a 3.5x spread
  • Three assumptions separate the paths: AI ceiling share (25/40/55%) and software spend growth rates
  • Anthropic's S-1 (filed Nov. 2, 2026) reported Q3 2026 revenue of $16.96bn
  • FRI Wave 11: median superforecaster sees OpenAI plus Anthropic at ~$300bn run rate by 2030

End-customer AI software spending grows from about $144 billion annualized in 2026 to about $564 billion in 2030 and about $1.23 trillion in 2035 in the base case of AI Insider's AI Market Model, H1 2026 edition, updated Nov. 3, 2026 for Anthropic's S-1 filing. A lower case reaches about $629 billion by 2035; a higher case reaches about $2.19 trillion.

That 3.5x spread between the outer paths is the point. AI Insider presents them as scenarios built on stated assumptions, not forecasts. The figures are annualized run rates sized from company disclosures, filings and other evidence, with ranges — not audited company results.

The method mirrors government practice. The UK Government Office for Science's "AI Scenarios 2030," published June 15, 2026, groups five futures under three trajectories in which AI progress slows, continues or takes off, and stresses the real outcome will likely mix elements of several.

How Are the Scenarios Built?

The model treats AI software as a share of all software spending. Two things move. Global software spending — about $1.44 trillion in 2026, a figure drawn from Gartner's forecast — grows at an assumed rate. AI's share of that spending, about 10% today, rises along an S-shaped adoption curve toward an assumed ceiling.

Technologies typically spread this way: slowly, then faster, then slowly again near saturation. Anchoring the curve to today's share means each scenario needs only three inputs.

The assumptions split as follows:

  • Ceiling share of software spend AI could reach: 25% lower, 40% base, 55% higher
  • Software spend growth per year, 2026-2030: 10% lower, 12% base, 15% higher
  • Software spend growth per year, 2030-2035: 6% lower, 8% base, 10% higher

The ceiling matters most. At 25%, AI becomes a large category within software but most spending stays on conventional products. At 55%, most new software spending goes to AI products. "Nobody knows which is closer to the truth, which is why the model shows both alongside a middle case," the report says.

What Do the Three Paths Look Like?

All three start at $144 billion in 2026. By 2030 they reach $353 billion (lower), $564 billion (base) and $983 billion (higher). By 2035: $629 billion, $1,230 billion and $2,186 billion.

AI's share of software spending in 2030 lands at 17% in the lower case, 25% in the base and 39% in the higher. By 2035 those shares reach 22%, 37% and 54% respectively.

Growth slows in every case as AI's share nears its ceiling. In the base case, annual growth starts above 40% and falls to around 12% by 2035.

How Do Other Forecasters See It?

The Forecasting Research Institute's Longitudinal Expert AI Panel, in its Wave 11 survey of Aug. 27, 2026, found the median superforecaster expecting OpenAI and Anthropic to reach a combined run rate of about $300 billion by 2030 — with a 34% chance of exceeding $400 billion and an 18% chance of staying below $100 billion.

Company revenue isn't like-for-like with the model, which strips out sales between AI vendors. But the width of the distribution tells a similar story. The panel's reasoning also maps onto the ceiling assumption: forecasters with high estimates argued AI products will be priced against the cost of a worker rather than a software seat, while those with low estimates tied revenue to existing software budgets.

Why Show Three Paths Rather Than One?

The future depends on decisions not yet made: how many businesses move AI from pilots into everyday use, how quickly agents take on whole tasks, and how prices change as models become more efficient. Falling prices could expand use or hold spending down.

A single number would hide that uncertainty. A reader who expects AI to remain a feature inside existing software will lean toward the lower path; one who expects AI products to take most new software budgets will lean toward the higher. The full report places the three paths beside published bank and broker forecasts.

What Would Shift the Picture?

Three kinds of evidence would move the model between paths: how much business AI spending reaches everyday use at scale, price per task and whether lower prices bring more use, and the growth of software budgets themselves.

New disclosures feed in each quarter. Anthropic's S-1 filing — Form S-1/A, Amendment No. 1, filed Nov. 2, 2026 — reported third-quarter 2026 revenue of $16.96 billion, broadly in line with the run rate the model already held for the company.

The full H1 2026 report and the Excel dataset, including the scenario inputs, are available from AI Insider Intelligence. For corporates budgeting a decade out, the practical takeaway is that the AI line item's size in 2035 depends less on model breakthroughs than on whether AI is priced like labor or like software.

Source: theaiinsider.tech

Share this article:

More from Amara Osei

Amara Osei

Show full bio

Senior reporter covering companies, deals and strategy in entrepreneurship, startups and venture funding at Business Bearings.

2,503 articles

Related articles

« Previous article