Quick Thoughts: At the knee of the curve of disruption

Franklin Templeton Institute’s Stephen Dover interviews Jonathan Curtis, Co-Chief Investment Officer of Franklin Equity, on artificial intelligence (AI) and its impact on the economy and markets.

1. Markets are pricing real AI disruption, not a bubble

Recent volatility across equity markets, particularly within software and digital knowledge work, signals a growing recognition that AI-driven disruption is no longer theoretical. Attention is shifting from whether AI will reshape economic activity to how quickly it will alter business models, earnings durability and capital allocation.

Curtis emphasized that he believes this market behavior is not driven by speculative excess. Rather, investors are attempting to price tangible disruption to software economics, adjacent forms of digital knowledge work and companies whose earnings assumptions rely on potentially fragile AI infrastructure spending plans.

At its core, this environment reflects accelerating awareness that rapid advances in AI capabilities are beginning to translate into tangible economic consequences.

2. AI capability gains are hitting economics faster than fundamentals show

The pace of improvement in AI models continues to exceed expectations. Scaling laws remain intact, token costs are declining rapidly, and performance continues to improve across a widening range of economically meaningful tasks.

On benchmarks such as GDPval¹ (which emphasize real digital work rather than abstract academic performance), results continue to move decisively higher. Across a growing set of digital job categories, AI-generated output is increasingly preferred to human output.

3. Software is ground zero for disruption

Software development represents the first major category of digital knowledge work where advances in AI are translating into tangible and dramatic economic consequences. Over the past six months, software stocks have faced sustained pressure, a trend that has intensified alongside rapid progress in agentic coding capabilities from leading model providers.

AI is materially increasing developer productivity, altering the supply dynamics of software itself. As application development costs fall, barriers to entry decline and enterprises gain greater ability to develop internal tools rather than rely solely on third-party vendors. Investors are increasingly focused on how these dynamics affect software business models and long term cash-flow durability.

4. Competition among AI model providers is reshaping infrastructure spend expectations

Competition among AI model builders is intensifying with direct implications for capital allocation, particularly among the most capital-intensive and loss-making providers. Investors are increasingly questioning whether the full scale of announced infrastructure deployment plans will ultimately be realized.

This skepticism does not reflect the weakening of AI demand. Instead, it reflects uncertainty around which platforms are likely to capture enterprise and consumer share and how infrastructure spending may be distributed across providers. That uncertainty has placed pressure on companies positioned as the most direct beneficiaries of individual model developers.

5. Capital rotation is rational

Viewed through this lens, recent market volatility appears rational. Capital is rotating toward areas that clearly benefit from AI adoption and segments where disruption risk is perceived to be structurally lower, particularly within the physical economy.

Investment implications

Curtis expects volatility to remain elevated as AI-driven disruption unfolds along powerful exponential curves. This volatility should be viewed as a feature of the transition rather than evidence that underlying trends are faltering.

One of Curtis’ convictions remains firm: Demand for AI compute is likely to remain exceptionally strong. This supports a durable investment case for leading semiconductor manufacturers, suppliers of semiconductor capital equipment and companies aligned with the broader AI infrastructure buildout, including utilities, industrials and materials.

Curtis also sees underappreciated productivity gains ahead for early AI adopters with control of large, compounding datasets and a willingness to rebuild workflows around agentic systems, particularly across health care, financial services, customer service, marketing and technology.

Real-world constraints matter. Curtis acknowledges power availability, engineering and construction capacity, and natural resource inputs represent both bottlenecks and sources of underappreciated pricing power, growth and profit margin expansion for well-positioned providers.