This week’s stories are about tech, AI and startups, coming from Korea and the USA

AI chip sales shift toward long-term deals as data center demand surges

The rapid expansion of artificial intelligence (AI) data centers is reshaping not only production strategies but also sales models at the world’s largest memory chipmakers, including Samsung Electronics, SK hynix and Micron Technology.

The companies are moving away from a traditional market dominated by quarterly and short-term orders toward longer-term supply agreements, a shift that could help reduce the sharp boom-and-bust cycles that have long characterized the memory chip industry, according to Korea Herald.

JPMorgan said in a July analysis that new long-term agreements are increasingly extending to three to five years. The contracts can include predetermined supply volumes, pricing formulas or price floors, while some carry stronger financial commitments such as upfront payments, collateral and financial guarantees.

Some agreements also include so-called “take-or-pay” provisions, under which customers must pay for the volume they have committed to purchase even if they ultimately take less than that amount.

Samsung said last month that nearly all of its major customers are seeking multiyear supply agreements. The company has signed contracts with five major global data center customers and is in talks with five additional major AI customers.

If those negotiations are completed, Samsung expects roughly 60% to 70% of its medium- and long-term memory production capacity to be secured through long-term agreements.

SK hynix has also completed negotiations on long-term contracts with about 10 companies, including major customers, while discussions with other customers are continuing.

Micron announced 16 strategic customer agreements in June. The deals are backed by $22 billion in cash advances and other commitments, including about $18 billion in cash payments.

The shift toward longer-term contracts is being driven primarily by tightening memory chip supplies as AI data center investment accelerates.

AI chip sales shift toward long-term deals as data center demand surges

Goldman Sachs, Intel invest $400 million in AI video startup

Goldman Sachs and Intel invested $400 million in Higgsfield, an artificial intelligence (AI) startup focused on video generation and editing, valuing the two-year-old company at $5.4 billion.

The new funding round more than quadruples Higgsfield’s valuation in just eight months, underscoring the growing investor interest in AI-powered video production.(By the way I have a story here about the startup investments of Turkey)

DST Global and Liberty Global led the round, which also included Tribe Capital, Smash Capital, Fifth Wall, Valor Capital, Mirae Asset Capital and NTT Docomo Ventures.

The platform also enables users to control camera movements and visual effects without relying on traditional filming equipment, according to City A.M.

Higgsfield was founded in 2023 by former Snap executive Alex Mashrabov and Yerzat Dulat. The company launched its browser-based platform in 2025, allowing users to generate videos from text and images and edit existing footage with AI.

Higgsfield says it attracted more than 30 million users across 238 countries and territories, with the United States representing its largest market.

But the company’s rapid rise in valuation is increasingly being driven by enterprise customers rather than individual creators.

Higgsfield’s annualized revenue reached $700 million in August, up from about $20 million a year earlier, according to the company. Enterprise customers now account for the majority of sales, compared with less than 25% in January.

The company is targeting brands seeking to produce a steady stream of video content for social media while reducing their reliance on traditional production companies and creative agencies.

AI-generated video requires significantly more computing power than text or image generation, making the cost of running video models one of the industry’s biggest challenges.

Higgsfield said it plans to use part of the new funding to expand its computing capacity, while also investing in its enterprise products and security infrastructure.

Goldman Sachs, Intel invest $400 million in AI video startup

What happens to the secrets people share with AI chatbots?

Imagine an AI chatbot designed to keep users talking longer, influence their opinions or persuade them to buy a product. Now imagine the system can use everything it knows about a person — including fears, insecurities and private details shared during conversations — to achieve those goals.

Rules governing how AI companies can use consumer data have yet to fully address that scenario, making the privacy commitments companies make in their data policies increasingly important.

AI companies are also developing ways to access information beyond the chatbot window.

OpenAI recently introduced an optional “Computer History” feature that allows ChatGPT to keep a record of the applications users work with and websites they visit on their computers.

Google, meanwhile, said last month that it would by default use photos and other content users upload through Search to train its AI systems. Users can opt out of the practice.

Such features could make AI systems more personalized and useful. But one of the most significant differences between AI chatbots and traditional search engines or social media platforms is their ability to encourage users to disclose highly personal information, according to Axios.

People routinely turn to AI tools with questions about their health, finances, jobs and relationships. That growing trove of personal information could become particularly significant as companies look for ways to monetize AI assistants.

Meta, Google and OpenAI are all exploring different approaches to advertising within their AI chatbots. That creates a potential economic incentive for companies to use personal data more extensively to understand users and increase engagement.

The result could be a new privacy challenge: AI assistants may know far more about their users than traditional digital platforms because people voluntarily tell them things they would rarely share elsewhere.

What happens to the secrets people share with AI chatbots?