Technology Sector
AI-driven profits are fueling massive shareholder returns and IPO preparations across the sector, with Samsung, Anthropic, and OpenAI all capitalizing on the memory chip and generative AI boom. Meanwhile, regulatory scrutiny is intensifying on medical AI devices and social media platforms, with Meta facing a major lawsuit over alleged child addiction and the FDA establishing new oversight frameworks.
Samsung Electronics is planning a shareholder return program exceeding 100 trillion won, or approximately $71.75 billion, to distribute record profits generated by the AI-driven memory chip supercycle. The initiative will allocate about 50 percent of the company's free cash flow toward special dividends and investor payouts, with a formal board vote scheduled for the end of August. This plan follows a similar announcement by domestic rival SK Hynix, which unveiled a 40 trillion won share buyback and cancellation program while committing to return over half of its free cash flow from 2025 through 2027. Both memory manufacturers have faced criticism from investors for holding large cash reserves during periods of high profitability, leading to concerns about capital efficiency. Meanwhile, the intense concentration of AI infrastructure spending on a few suppliers is driving up component prices, with TrendForce projecting third-quarter DRAM contract prices to climb 58% to 63% and NAND flash prices to rise 70% to 75%. Investors responded favorably to the payout reports, pushing Samsung's stock up by more than 9% on August 20.
The Food and Drug Administration opened a public comment period on a discussion paper outlining potential regulatory frameworks for generative artificial intelligence-enabled medical devices. Comments are due through October 19, 2026. Led by the Digital Health Center of Excellence within the Center for Devices and Radiological Health, the initiative introduces a two-axis risk assessment framework alongside a competency-based premarket evaluation model. The proposed evaluation structure combines non-clinical device benchmarking with clinical confirmation, drawing an analogy to how physicians are trained and credentialed. Unlike traditional software, generative AI systems can accept open-ended inputs, perform multiple tasks, and generate variable outputs, creating distinct challenges including real-world performance degradation and unpredicted behaviors. The discussion paper also addresses risk-proportionate postmarket monitoring and considerations for foundation models and agentic AI systems. Acting FDA Commissioner Kyle Diamantas noted that the effort aligns with administration priorities to accelerate AI-powered medical products to market while upholding safety standards.
Calendly is entering the AI meeting note-taker market with a native transcription tool and a planned assistant named Callie. The scheduling platform's new product joins calls to record audio and video, transcribe conversations, generate summaries, and draft follow-up emails. CEO Tope Awotona noted that the company is targeting outward-facing professionals like sales and marketing workers whose days are packed with meetings. Calendly is also testing a system-audio transcription feature similar to Granola's approach, alongside Callie, an assistant designed to pull context from past meetings and scheduling data. The company faces a crowded field of established note-takers including Fireflies, Read AI, Otter, and Fathom. To address privacy concerns surrounding meeting bots, Calendly plans to notify participants before calls begin using its scheduling integration and will allow users to request that the AI leave the conversation.