TSMC soars on AI chip demand; IBM stumbles as customers shift to AI

TSMC delivered the week's biggest AI winner story. Taiwan Semiconductor Manufacturing Co. announced blowout earnings on Thursday, with second-quarter revenue jumping 36% year over year and net income rocketing 77.4% higher. The company also said it would increase spending to expand its US footprint. TSMC produces semiconductors for the world's most influential tech companies, including Apple, AMD, Nvidia, and Qualcomm. Annual revenue has surged from $75.99 billion（約12兆円） in 2022 to $122.56 billion（約20兆円） in 2025, a direct payoff from the AI boom. According to CFO Wendell Huang, TSMC's Q2 results were supported by strong demand, which he expects to continue into the current quarter as tech companies seek as many AI chips as they can get.

On the loser side, IBM preannounced its Q2 earnings on Tuesday with a stark warning: the company's revenue and earnings per share fell well below analyst expectations. Analysts had expected IBM to report adjusted earnings per share of $3.02 on revenue of $17.86 billion（約2.9兆円）, but the company posted adjusted EPS of $2.93 and revenue of $17.2 billion（約2.8兆円）. The culprit: customers are shifting their spending from IBM's mainframe systems to AI servers and memory and storage chips, signaling a fundamental reallocation of enterprise capital.

China's Moonshot AI also made waves this week. On Friday, the company revealed that its new Kimi K3 AI model has frontier-level capabilities. While it falls short in overall performance compared to state-of-the-art models like Anthropic's Fable 5 and OpenAI's GPT-5.6 Sol, K3 surpasses them on certain benchmark tests. The announcement underscores how quickly China's AI companies are catching up with the most well-funded organizations in the US. Like China's DeepSeek, Moonshot AI focuses on releasing open-weight models—models that users can download for free and customize—whereas Anthropic, OpenAI, and Google focus on proprietary models that users must pay to access. China's AI companies also generally undercut their US counterparts on usage pricing, making them more appealing to businesses that want to adopt AI without paying for access to Anthropic's Claude, OpenAI's GPT, or Google's Gemini.

The AI boom is reshaping corporate spending and global competitive dynamics. TSMC's 36% revenue jump reflects sustained demand from the world's largest tech companies—Apple, AMD, Nvidia, and Qualcomm—all racing to secure AI chips. CFO Wendell Huang's confidence that strong demand will continue suggests the chip supply chain remains a critical bottleneck as enterprises build out AI infrastructure. This same shift is devastating for legacy players like IBM, whose mainframe business is being cannibalized by customer demand for AI servers and memory storage.

Meanwhile, China's Moonshot AI represents a structural challenge to US AI dominance. By releasing open-weight models like Kimi K3—which surpass certain benchmarks of OpenAI's GPT-5.6 Sol and Anthropic's Fable 5—and undercutting US competitors on pricing, Chinese firms are winning over cost-conscious businesses that cannot justify proprietary model subscriptions. This mirrors the open-source playbook that has worked across software; the economics may shift pricing power away from the US incumbents toward lower-cost Chinese alternatives.

No comments yet. Be the first to share your thoughts!