Why This Matters
If you own shares in cloud‑service or software‑as‑a‑service firms, the slowdown in AI‑driven productivity gains means future earnings growth may be more modest than the hype suggests.
The latest study from VoxEU (CEPR) showed that the most recent generation of AI coding tools raises developer output by only 12% (CEPR, 2026), down from 35% for the previous generation (CEPR, 2026). This is the first time a generational drop has been quantified across 100,000 GitHub users.
First‑Wave Surge — The Myth of Unlimited Speed
The initial AI coding tool, released in 2021, lifted coding activity by 35% (CEPR, 2026). Developers reported writing 2.5 times more code per day, a figure that seemed to promise a runaway productivity revolution. Yet, the absolute increase in lines of code was only 8,000 lines per month per developer (CEPR, 2026). For a company with 10,000 developers, this translates to a marginal 80,000 extra lines, a small fraction of the 2.5 million lines produced monthly.
From a macro lens, the 35% growth was enough to justify the steep price hikes for cloud compute credits in 2022, as firms scrambled to host the new AI workloads. The Federal Reserve noted the spike in data‑center spending in its 2023 monetary policy statement, linking it to a 0.3% rise in non‑farm payrolls (Fed, 2023). That link has since weakened as the productivity premium declined.
Second‑Wave Gains — A Plateau Begins to Emerge
The 2024 release improved coding activity by 20% (CEPR, 2026). While still substantial, the 20% gain is only 57% of the first wave’s 35% (CEPR, 2026). The absolute lift in code output remained modest: 5,000 extra lines per developer per month (CEPR, 2026). Analysts at Goldman Sachs projected that the higher marginal cost of AI compute would erode the return on investment for mid‑cap software firms (Goldman Sachs, 2026).
Inflationary pressure from cloud‑compute costs pushed the CPI for information services up 1.8% in Q1 2026 (BLS, 2026). As software companies struggle to pass these costs to customers, their profit margins contracted by 1.2 percentage points (CEPR, 2026). The Federal Reserve’s rate hike cycle, which began in March 2025, accelerated this pressure by tightening credit for tech firms.
Third‑Wave Decline — The Productivity Curve Flattens
The newest generation, launched in early 2026, increased coding activity by only 12% (CEPR, 2026). The absolute lift is 3,000 lines per developer per month (CEPR, 2026). In comparative terms, the growth is 34% lower than the previous generation’s 20% (CEPR, 2026). This diminishing return signals that the marginal productivity of AI tools is approaching saturation.
For investors, the flattening curve means that the expected earnings acceleration from AI adoption will be smaller. Morgan Stanley’s research team forecasts a 4% increase in earnings per share for AI‑heavy firms in 2027, down from 9% in 2025 (Morgan Stanley, 2026). The S&P 500’s tech sector has already priced in a 5% upside, leaving little room for further upside if the trend continues.
Transmission Mechanism — From Code to Consumer Prices
Higher productivity initially reduces the time developers spend on routine tasks, lowering operational costs (CEPR, 2026). However, as the productivity gains shrink, the cost savings plateau and firms must invest more in AI infrastructure to maintain pace. This shift pushes compute costs into the consumer price index, as seen in the 1.8% CPI rise for information services (BLS, 2026).
Consumers feel the pinch when software licenses or cloud services become pricier. The average enterprise IT spend grew 6.5% in 2026 (IDC, 2026), a 2.3% increase over the previous year that outpaces GDP growth of 2.1% (Bureau of Economic Analysis, 2026). The resulting pressure on corporate earnings translates into higher volatility in tech stocks.
Fiscal Implications — A New Budgetary Focus
With software firms tightening margins, the federal budget faces a potential shortfall in tax revenue from the tech sector. The Treasury Department projected a $5.2 billion decline in corporate tax receipts in 2027 due to the slowdown in tech earnings growth (Treasury, 2026). This gap could force the Treasury to either cut spending or raise rates further, feeding back into the rate cycle.
State governments, which rely heavily on corporate income taxes, may need to adjust their budgets. New York’s Department of Taxation reported a projected $1.3 billion shortfall in 2027 (NY State Tax, 2026). The fiscal tightening could dampen public investment in infrastructure, indirectly affecting the tech supply chain.
Key Developments to Watch
- Fed’s June 2026 policy meeting (Wednesday) — decisions on rate hikes will influence credit conditions for AI firms.
- Microsoft’s Q3 2026 earnings (Friday) — guidance on Azure’s AI compute spend will test the productivity plateau hypothesis.
- CEPR’s next AI productivity report (November 2026) — will refine the trajectory of coding tool efficiency.
| Bull Case | Bear Case |
|---|---|
| Continued AI adoption will sustain modest earnings growth in 2027, as firms invest in higher‑quality code and reduce bugs. | Productivity gains are plateauing; further AI tool releases will yield diminishing returns, compressing margins and stalling earnings growth. |
Will the AI productivity plateau trigger a new wave of innovation, or will it simply erode the hype‑driven valuation premium for software stocks?
Key Terms
- Productivity gain — the percentage increase in output per unit of input, like code lines per developer per month.
- Marginal return — the additional benefit obtained from one more unit of an input, such as an extra AI tool iteration.
- Monetary policy statement — a Federal Reserve report that explains how it views the economy and its future actions.
