The tech sector in 2026 is being defined by a stark divide: semiconductor stocks tied to AI data centers are soaring, while software and information services companies are facing unprecedented losses. The iShares Semiconductor ETF has climbed 19% this year, building on a 40% gain in 2025, while the iShares Expanded Tech-Software Sector ETF has dropped 22% so far in 2026 after an 8% fall in the fourth quarter of last year. Analysts say the rise of artificial intelligence and autonomous coding agents is at the heart of the shift.
Semiconductor companies linked to AI infrastructure are benefitting from massive capital expenditures by tech giants. Amazon projects $200 billion in 2026 investments, Alphabet’s Google plans up to $185 billion, Meta Platforms anticipates $135 billion, and combined spending from Microsoft, Oracle, and CoreWeave could bring total AI data center investment close to three-quarters of a trillion dollars. Since OpenAI’s ChatGPT debut in November 2022, the semiconductor ETF has surged roughly 190%, reflecting the bullish sentiment.
Conversely, software firms are facing existential pressure. Companies like Chegg have plummeted 98% since ChatGPT’s release, while Salesforce shares are down 29% this year, trading at record-low price-to-earnings ratios. Investors fear that large language models (LLMs) like ChatGPT, Anthropic’s Claude, and Alphabet’s Gemini can produce software at dramatically lower costs, potentially displacing traditional enterprise software entirely. Coding has emerged as the most successful commercial use of LLMs, amplified by tools such as Anthropic’s Claude Code and other autonomous agents capable of performing complex programming tasks rapidly.
The rise of desktop agents like OpenClaw and Claude Cowork is intensifying the disruption. OpenClaw, an open-source agent for Mac computers, allows technically skilled users to automate multistep tasks and manage websites, software, and communications using LLMs. While powerful, these agents present serious cybersecurity risks, prompting firms like CrowdStrike to develop tools to detect and block unauthorized agent activity. Security experts warn that making tools like OpenClaw completely safe is nearly impossible, and improper use can expose sensitive data.
The combination of AI-driven coding, automated knowledge work, and powerful desktop agents is redefining what software means in a world increasingly navigated through prompts rather than traditional interfaces. OpenClaw’s developer recently joined OpenAI, symbolizing the accelerating shift toward AI-centric productivity. For now, the market narrative favors semiconductors over software, but analysts suggest that the software sell-off may present future buying opportunities once the market digests the new AI-driven landscape.
