The gold rush triggered by the launch of ChatGPT has sent investors on a frantic hunt for the next big winner in the artificial intelligence sector. Initially, the spotlight fell squarely on chipmakers and networking giants, but as the industry matures, fund managers are shifting their gaze toward the hidden plumbing of the digital age. According to experts at Fidelity, the focus is moving away from simple processing power and toward a more critical, enduring constraint: memory.
Zach Turner, manager of the Fidelity Capital Appreciation Fund, suggests that while many traders chase short term spikes, the real value lies in the picks and shovels of AI infrastructure. He notes that as AI models evolve from basic chatbots into agentic systems capable of complex reasoning, they require staggering amounts of data to be stored and accessed instantly. With demand for memory hardware reportedly doubling every year while supply grows at a fraction of that rate, memory has become one of the most significant bottlenecks in the entire buildout.
This evolving landscape requires a strategic approach to portfolio management. Rather than viewing AI stocks as a monolith, some Fidelity managers categorize opportunities into three distinct tiers based on how long their competitive edge will last. At the top are companies with durable advantages in semiconductor manufacturing and testing equipment—businesses that remain essential regardless of which specific AI application wins out in the end. These firms provide high quality cash flows because their services are nearly impossible to replace quickly.
Further down the spectrum are medium and short term plays involving more commoditized products. While these companies often show explosive earnings growth due to current shortages, analysts warn that these advantages can fade once new competitors enter the market or supply chains stabilize. For professional investors, success in AI depends less on predicting which app becomes popular and more on identifying which physical components will remain scarce as hyperscalers pour trillions into development over the coming years.