What is the biggest narrative shift for AI in 2026?

2026 may become the turning point when AI inference workloads surpass training. According to Trading Platform, Bank of America analyst Vivek Arya's team has released a new research report stating that, despite concerns over funding, valuation, and interest rate fluctuations, continued growth in AI capital expenditures is driving strong performance in the chip sector, with industry focus shifting from training to the inference segment, which can truly deliver investment returns. Bank of America notes that by 2030, AI capital expenditures are expected to reach $1.2 trillion, and inference could ultimately account for the majority share—up to 75%. Although the market has many concerns about AI financing, valuation, and interest rate volatility, chip stocks have performed robustly at the start of this year. The Philadelphia Semiconductor Index (SOX) has risen about 13% year-to-date, marking the second-best January performance in the past 20 years, far outperforming the S&P 500's mere 1% gain. Notably, this rally has been led not by computing giants Nvidia and Broadcom, but by memory chip makers, semiconductor equipment suppliers, and analog chip companies. The AI capital expenditure commitments of hyperscale cloud service providers have inspired market confidence. These companies have emphasized three points: AI investment is crucial for maintaining double-digit growth; if not for supply constraints, sales growth could have been even higher; and there is no evidence of a "bubble." Bank of America Securities predicts that computing stocks will regain momentum based on these positive signals. This narrative shift will create differentiated opportunities for various types of chip suppliers. From GPUs, CPUs to ASICs, a variety of silicon solutions will find their positioning in the inference market, while continuing to impact memory and semiconductor equipment suppliers. Turning Point for Inference Workloads Inference is the key segment for realizing returns on AI investment and requires a range of optimized chip solutions for cost and performance. According to Bank of America Securities analysis, while training remains important, 2026 may mark the year when inference becomes the larger workload, ultimately accounting for the majority—or 75%—of the expected $1.2 trillion AI capital expenditures in 2030. During this shift, Nvidia maintains its leading position with the broadest product pipeline, covering training (Blackwell, Vera Rubin), inference (Rubin CPX, Groq, Vera CPU), and has supply assurance advantages. Broadcom enjoys strong collaborations with Google and Anthropic and is finding new opportunities with OpenAI, Apple, and xAI. Bank of America Securities believes concerns over Google's in-house computing (COT) risk are exaggerated, leading to excessive declines in Broadcom's stock price. AMD, as the reliable second supplier of Nvidia's general-purpose chips, saw its stock fall last Friday due to market fears about TSMC's 2nm process, but Bank of America Securities believes the process is progressing well. Opportunities remain for memory and semiconductor equipment suppliers, although volatility may increase after recent surges. Optical Connectivity: Real Demand but Overblown Gains Optical transceiver and component suppliers are the strongest chip sector performers after memory chips. With the expansion of AI clusters and bandwidth demand, the necessity for increasing optical connectivity is indisputable. Nvidia's upcoming photon switch (with a webinar to be held February 3 during LITE's earnings release this week, COHR will announce results in the same period) is a potential catalyst, and Marvell's acquisition of Celestial AI has further sparked interest in new optical architectures. However, it should be noted that, except for Meta, evidence of hyperscale cloud service providers' interest in co-packaged optics (CPO) is limited, mainly because of higher operational complexity and more bill-of-materials control shifting to Nvidia and Broadcom. According to top industry analyst group Lightcounting, CPO sales are expected to account for only about 1% ($500 million) of the roughly $46 billion Ethernet transceiver market in 2026/27. Copper cables remain relevant, and it is recommended to buy leading active cable provider CRDO. ~~~~~~~~~~~~~~~~~~~~~~~~ The above featured content comes from Trading Platform. For more in-depth analysis, including real-time interpretation and frontline research, please join [Trading Platform • Annual Membership]. Risk Warning and Disclaimer The market involves risk and investments require caution. This article does not constitute personal investment advice and does not take into account the individual investor's specific investment objectives, financial situation, or needs. Users should consider whether any opinions, viewpoints, or conclusions in this article are appropriate for their particular circumstances. Investing based on this article is at your own risk.