AI drives code to zero cost, but giants still dominate

The 2025 DORA State of AI-Assisted Software Development report shows AI has driven the marginal cost of writing code to near zero [DORA]. Distribution, trust, support and liability still represent roughly 80 % of a software business’s expenses, leaving the middle tier exposed.

AI acts as an amplifier: organizations with strong development practices reap productivity gains, while those with fragmented processes see little benefit [DORA]. In an Adidas pilot, loosely‑coupled teams achieved 20‑30 % higher output, whereas tightly‑coupled legacy ERP teams saw no measurable gain [DORA].

The data highlight three implications. First, a $0 marginal cost becomes a benchmark that competitors will chase. Second, 10‑person, VC‑backed startups that merely add features are squeezed from below by solo founders who can ship for free and from above by giants that bundle services. Third, discipline and systematization—not tool access—determine who converts AI into results.

Stack Overflow’s 2025 developer survey records a drop in AI favorability from 72 % to 60 % year‑over‑year, with 46 % of respondents expressing distrust in AI accuracy [Stack Overflow]. GitHub logged over one million Copilot‑generated pull requests between May and September 2025 [GitHub]. JetBrains’ usage metrics echo the same trend of widespread AI adoption but uneven impact.

Geographically, India, Brazil and Indonesia are emerging as major contributors to the developer ecosystem, reshaping the competitive landscape [GitHub]. Solo founders in these regions can leverage local compliance and lower hosting costs to challenge US‑based giants.