AIdriven

Published on 07/07/2026 – 11:28 GMT+2•Updated
11:34

The South Korean technology giant Samsung said on Tuesday it expects operating profit of about 89.4 trillion won (€51bn) for the April-June quarter, roughly nineteen times the 4.7tr won (€2.7bn) it earned a year earlier and more than it made in the previous three years combined.

ADVERTISEMENT

ADVERTISEMENT

The extraordinary numbers reflect the same force reshaping the memory industry worldwide: the race to build AI data centres has pushed chip prices to record highs.

According to Citi Research, average selling prices for DRAM memory rose 44% quarter on quarter, and NAND flash 53%, as AI demand spilled beyond specialised high-bandwidth memory into the conventional chips that go into phones, servers and PCs, with customers now chasing longer-term supply contracts.

The estimate beat analyst forecasts, but far from celebrating, the market sold.

Samsung shares fell by over 10% before closing nearly 7% lower, dragging rival SK Hynix and the wider Kospi index down with them.

Samsung’s stock has more than doubled this year alone, so a historic quarter was already priced in, and leveraged local ETF products tracking the shares have made them prone to outsized moves.

There was also a blemish in the numbers as revenue of 171tr won (€97.6bn), though up 129% year on year, came in slightly below forecasts.

“We believe the slight revenue miss was largely driven by more moderate DRAM price hikes than expected, which likely spooked investors who are increasingly pricing in structural strength in memory prices,” said Jing Jie Yu, an analyst at Morningstar.

Hanging over everything is durability.

Investors are increasingly asking whether the technology giants bankrolling the AI build-out can sustain their spending without piling up debt against a payoff that remains unproven, the worry behind last week’s chip sell-off across Asia.

Samsung publishes its full results, with a breakdown by division, on 30 July, a report the market will scour for clues about whether the boom is structural or simply another memory cycle nearing its peak.

Additional sources • AP

Grid operators warn the US heatwave could send electricity demand near record levels before the Fourth of July holiday.

Power grid operators in the United States are warning that a dangerous heatwave could put more strain on an electric grid already under pressure from surging energy consumption.

A stretch of extreme heat is expected to intensify across much of the central and eastern parts of the country this week, peaking from Tuesday through Thursday.

That heatwave is likely to continue through one of the busiest travel weekends of the year, as millions of Americans prepare for Fourth of July celebrations on Saturday.

Temperatures this week are forecasted to climb above 38 degrees Celsius (100 degrees Fahrenheit) from Boston to Washington, DC, pushing up demand for air conditioning.

The heatwave coincides with two major events on the US calendar. Saturday’s holiday marks the 250th anniversary of the US’s independence, and millions are expected to gather for barbecues, parades and fireworks.

The extreme temperatures also come as the FIFA World Cup has reached the knockout stage, with many host cities, including New York, Boston, Philadelphia and Washington, expected to feel the heat.

Humidity could push the heat index as high as 46 degrees Celsius (114 Fahrenheit) in some places, while overnight temperatures will offer little respite.

The US’s largest regional grid operator, PMJ Interconnection, is forecasting record summer electrical demand of 166.3 gigawatts for Thursday evening, surpassing the previous summer peak set two decades ago, in 2006.

The New York Independent System Operator (NYISO), the state’s grid operator, is also expecting electricity demand to approach record highs, while the Midcontinent Independent System Operator (MISO), which covers 15 states in the Midwest and South, could also see its peak demand record challenged.

Authorities at MISO say they will rely on PMJ for support in covering consumer needs.

In a May report, PMJ’s executives warned of a “fundamental mismatch between how fast demand is growing and how quickly new supply can be built and connected to the grid”.

New power plants, they said, now take twice as long to build and cost twice as much as they did a decade ago.

Meanwhile, there has been increasing pressure on electrical grids from new technology like data centres and electric vehicles.

In May, PMJ said hyperscale data centres were “adding load at an unprecedented pace”.

Experts say the artificial intelligence (AI) boom is colliding with climate change, with tools like ChatGPT, Gemini and Claude being processed in vast, energy-hungry data centres.

The most energy-intensive are the hyperscale facilities that require between 100 and 300 megawatts of electricity, enough to power hundreds of thousands of homes.

Many of those are concentrated in northern Virginia, which sits within PJM’s service territory and is widely described as the world’s largest data centre hub.

Researchers have also identified what they call a “data heat island effect”, finding that land surface temperatures around AI data centres rise by an average of 2 degrees Celsius (3.6 degrees Fahrenheit), with some locations seeing increases of up to 9 degrees Celsius (16.2 degrees Fahrenheit).

The National Weather Service in the US warns that long periods of extreme heat create significant stress on the body.

It has urged people to limit outdoor activity, stay hydrated and keep close to air conditioning or cooling centres.

A 2024 report from the Journal of the American Medical Association (JAMA) found that 21,518 deaths in the United States from 1999 to 2023 were heat-related.

The highest number came in the final year of the report’s analysis, 2023. That year, 2,325 people died from causes attributed to high temperatures.

The presidential chief of staff for policy on Tuesday proposed introducing public dividends to distribute the “fruits” from an artificial intelligence (AI)-driven economic boom.

Kim Yong-beom made the suggestion in a Facebook post, as the benchmark Korea Composite Stock Price Index (KOSPI), the country’s main stock index, was heading toward the record-high 8,000-point mark, driven by gains in chipmakers, including Samsung Electronics Co. and SK hynix Inc.

The companies posted record-high profits in the first quarter, highlighting their leadership in the global chip market amid the AI boom.

“The fruits of the AI infrastructure era are not the results generated by certain companies alone … they were produced on a foundation that all the people have built together over half a century,” the presidential policy chief wrote.

He argued that deliberating on how to use the proceeds would “not be optional but necessary if (the companies’) strategic advantage in the distribution network for AI infrastructure creates a structural upcycle and that, in turn, leads to record-breaking tax revenues.”

“Part of these fruits should be structurally returned to the people,” he said.

Kim referred to cases of foreign countries “socially institutionalizing structural excess profits,” such as Norway’s oil-generated profits in the 1990s, and suggested “public dividends” as the name for the program should South Korea introduce such a system.

The policy chief also listed a fund for young entrepreneurs launching startups, a pension program for the elderly and a fund for retraining in the AI era as possible areas that could benefit from the initiative, while stressing the need for social consensus in making such a decision.

“There’s a possibility that South Korea could become the first country to return excess profits from the AI era into people’s lives,” he noted.

Cheong Wa Dae later clarified that Kim’s proposal has nothing to do with any internal discussion or review at the presidential office, describing it as a “personal opinion.”

Copyright (c) Yonhap News Agency prohibits its content from being redistributed or reprinted without consent, and forbids the content from being learned and used by artificial intelligence systems.