TradingKey – The rapid development of artificial intelligence (AI) technology has shifted the spotlight from proprietary models like ChatGPT to open-source alternatives like DeepSeek in just a few years. Capital markets are now pricing in a new AI cycle: the transition from the AI hardware era to the AI software era.
Whether it’s Goldman Sachs’ Four Stages of AI Trading, Bank of America’s Five Trends in AI Software, or Morgan Stanley’s alpha in AI’s rate of change, Wall Street broadly agrees that AI remains the dominant theme in technology. However, a growing consensus suggests that opportunities are shifting from hardware infrastructure to software applications.
In early 2025, Goldman Sachs noted that global capital allocation is gradually pivoting from "hard tech" to "soft tech," a trend amplified by the launch of China’s large-scale model, DeepSeek.
According to McKinsey, the generative AI value chain consists of six segments: computer hardware, cloud platforms, foundational models, model hubs and machine learning, applications, and services. Compared to traditional AI value chains, foundational models represent a new addition.
[Generative AI Value Chain, Source: McKinsey]
McKinsey predicts that market opportunities across these segments will expand gradually over the next three to five years. The AI applications market is expected to experience the fastest growth, presenting significant value-creation opportunities for both established tech giants and new entrants.
Simplified, the AI industry chain can be divided into:
[AI Value Chain, Source: TradingKey]
According to Lu Qi, founder of MiraclePlus (formerly Y Combinator China), the launch of OpenAI’s ChatGPT marked a "paradigm shift" in AI development—from information perception systems (acquiring data from environments) to model knowledge systems (expressing information through models).
ChatGPT ignited a race for large AI models, with NVIDIA emerging as the biggest winner by supplying critical chips for model training. Its market cap once topped the U.S. stock market and remains firmly in the top three.
According to Deloitte, the semiconductor industry experienced robust growth in 2024, with sales surpassing expectations, achieving double-digit growth (of approximately 19%) and reaching $627 billion. Deloitte projects continued expansion, forecasting a record high of $697 billion in 2025, with target valuations of $1 trillion by 2030 and $2 trillion by 2040.
However, AI hardware stocks like NVIDIA face challenges due to cyclicality, slowing growth, and policy risks:
[Semiconductor Cycle, Source: Deloitte]
As NVIDIA’s hardware boom cools, Wall Street is turning bullish on software in 2025. Wedbush notes, "The AI software era has arrived."
Bank of America identifies five trends driving software growth:
Goldman Sachs outlines four phases of AI trading:
By 2025, Goldman sees Stages 1–2 under pressure, while Stage 3 (24 software firms among its 40 picks, including Palantir and Cloudflare) offers growth.