SAN FRANCISCO -- November 13, 2026 -- Software companies selling AI as their primary product reported median ARR growth of 34% in 2025, 12 percentage points ahead of peers embedding AI in broader products, according to KeyBanc Capital Markets' 17th annual AI x Software Benchmark. The same core AI companies reported a median EBITDA margin of negative 12%, against negative 1% for the embedded group.
KeyBanc Capital Markets, the investment banking unit of KeyCorp, has renamed the survey, previously the SaaS Survey, and widened it to cover AI's effect on product strategy, monetization, unit economics and operating models. The bank did not disclose how many companies responded.
AI-native vendors grow faster but earn less per employee
Median 2025 ARR growth was 34% for core AI product companies, 22% for embedded AI companies and 9% for companies still developing customer-facing AI. The core AI group grew about 3.8 times as fast as the third group, by our calculation.
Embedded AI companies led on productivity. Their ARR per full-time employee was $201,000, against $117,000 at core AI companies. That is a gap of $84,000, or about 72% more per head. The margin gap between the two groups is 11 percentage points.
Charging separately for AI paid off in growth. Companies that do so reported 27% median ARR growth in 2025, versus 17% for those that do not, a 10-point difference. KeyBanc says the advantage has not yet shown up in net retention.
AI revenue is set to grow 128% in 2026 but reach only 12% of ARR
Respondents project AI-driven ARR growth of 128% in 2026, up from 74% in 2025. That would multiply AI-driven ARR by about 2.3 times in one year, yet AI would still account for only 12% of total ARR.
Overall median ARR growth was 24% in 2025 and is projected at 23% in 2026. Participants plan for a broad-based acceleration to 40% in 2027, a 17-point jump from the 2026 projection. Median EBITDA margin is expected to improve from negative 3% in 2025 to 1% in 2026, a 4-point gain.
Scott Peterson, Managing Director of Software Investment Banking at KeyBanc, said: "2026 is a proving year for AI-driven revenue."
Token costs push 44% of respondents over their AI budgets
Some 44% of respondents report AI spend over budget year to date, and 24% are over by more than 20%. By our calculation, that means roughly 55% of the over-budget group exceeds plan by more than 20%. Higher-than-expected token usage was the most cited cause.
Spending shows up in margins. Companies spending more than 5% of revenue on AI reported a median gross margin of 66% in 2025. Those spending less than 2% reported 78%, a 12-point difference.
Engineering absorbs AI first as headcount effects are expected to widen
Some 92% of respondents say AI is shortening product development cycles, and 91% deploy AI internally in engineering. Peterson says companies are reinvesting the productivity gains to ship more product faster rather than letting them drop to the bottom line.
In 2025, 65% of participants reported no AI-related headcount reduction or reallocation. Over the next 12 to 24 months, that share falls to 34%. R&D (45%) and customer service (37%) are the functions most likely to be affected.
Customers are also building. Some 57% of respondents say their customers are developing alternative products with AI coding tools. Only 20% offer fully autonomous agents to their own customers.