The artificial intelligence sector is undergoing a sharp sentiment reversal in 2026, as a wave of selling pressure sweeps through stocks that had previously posted substantial gains. Morningstar research indicates that the prevailing “anything-but-AI” market posture has driven valuations for several well-capitalized AI-related companies down to compelling levels. According to the firm’s latest analysis, Nvidia, Taiwan Semiconductor Manufacturing Co., Meta Platforms, Alibaba, and Arista Networks are each trading at a significant discount to their respective fair value estimates.
“Generative AI remains the largest theme within the sector,” said Dan Romanoff, senior equity research analyst at Morningstar. “Software companies are embedding next-generation AI capabilities into their offerings, cloud providers are rolling out new services and scaling infrastructure, and select semiconductor firms are seeing surging demand for AI and data center chips.”
Below is a summary of the fundamentals and valuation profiles for the five stocks.
Nvidia shares currently trade roughly 32% below Morningstar’s fair value estimate of $260 per share. The company’s wide economic moat rests on its leadership across GPU hardware, the CUDA software ecosystem, and networking solutions tailored for AI workloads. The report highlights that elevated switching costs tied to the CUDA platform present a formidable barrier for rival chip designers seeking to gain traction in AI training. While major cloud vendors may pursue in-house development or alternative supply arrangements over time, such efforts are expected to only marginally erode Nvidia’s dominant positioning. Meanwhile, the company’s data center networking segment continues to scale in response to growing demand for AI cluster interconnectivity.
As the world’s largest dedicated chip foundry, TSMC commanded approximately 70% market share in 2025. The stock currently trades at an estimated 20% discount to Morningstar’s fair value estimate of $428 per share. The firm’s analysis points to sustained, long-term tailwinds from AI, Internet of Things, and high-performance computing applications, all of which require increasingly energy-efficient semiconductors. TSMC’s technological edge in leading-edge process nodes enables the company to generate excess returns even as the broader semiconductor cycle fluctuates. The ongoing industry shift toward the fabless model continues to provide structural support for foundry growth.
Meta Platforms—parent of Facebook, Instagram, WhatsApp, and Messenger—maintains a global user base of nearly four billion monthly active accounts. Its shares are priced at a roughly 33% discount to Morningstar’s $850 fair value estimate. The company is leveraging AI to refine its ad-targeting algorithms, a move aimed at boosting advertiser return on ad spend and lifting average revenue per user. Separately, Meta’s proprietary Llama large language model has been integrated as a chat assistant across its app ecosystem. Although the near-term monetization pathway for this feature remains undefined, the deployment is expected to contribute to incremental gains in user engagement and time spent on the platforms.
Alibaba ranks as the world’s largest online and mobile commerce operator by gross merchandise volume. The stock trades approximately 53% below Morningstar’s fair value estimate of $258 per share. Despite competitive encroachment from domestic rivals such as PDD and Douyin, the Taobao and Tmall marketplaces continue to generate steady cash flow that underpins expansion at AliCloud and the company’s global initiatives. Alibaba is divesting non-core assets and has expanded its share repurchase authorization to $35.3 billion through March 2027. Management has articulated a target to lift return on invested capital from single-digit levels into double-digit territory over the next several years.
Arista Networks supplies Ethernet switches and software primarily to data center operators. The stock is trading at a discount of roughly 28% relative to Morningstar’s fair value estimate of $175 per share. The company’s Extensible Operating System (EOS) delivers a programmable, modular networking architecture that distinguishes it from traditional hardware-centric, lock-in models. As demand for high-speed interconnects among GPU clusters intensifies to support AI training and inference, Arista’s established position in high-performance switching positions it as a direct beneficiary of AI-related capital expenditures. Concurrently, the firm is expanding its footprint beyond hyperscale cloud accounts into the broader enterprise campus networking segment.
