Samsung, SK hynix and Micron have reportedly committed all of their DRAM and high-bandwidth memory output for 2027, with buyers locked into multi-year agreements at prices above this year's levels. HBM consumes far more wafer area per bit than ordinary memory, so each AI order tightens supply for PCs, phones and cars. Contract prices for a 16-gigabit DDR5 chip rose from about $8.70 in October 2025 to $48 in September 2026. Some server memory trades 30 to 50 percent above contract on the spot market. Nvidia reports results on Tuesday, with memory supply rather than wafer starts capping how many accelerators can ship.
Anthropic released Claude Fable 5.2 and Claude Mythos 5.2, two names for the same underlying model separated by how much cyber and biology work they will do. Fable is open to all paid users at $8 per million input tokens, while Mythos goes only to vetted security and life-science organisations, and the launch lands weeks before a possible stock listing.
Nvidia's board said chief executive Jensen Huang resigned on November 12, and the chief financial officer is running the company while directors search for a permanent replacement. A day later, a Reuters review identified more than $3 billion in Chinese artificial intelligence computing projects built around restricted Nvidia processors, even though Nvidia says U.S. export controls shut it out of China's AI data-center market. Chinese labs have also been trying to rent Nvidia capacity at data centers outside China, and the Commerce Department is moving to close that route.
Microsoft's email and collaboration services were still recovering on Friday after an outage that began the previous morning and hit Outlook, Teams, Copilot and Exchange Online for users in dozens of countries. Downdetector logged about 140,000 reports from 74 countries within an hour. Microsoft said some users could not reach Exchange Online mailboxes through multiple connection methods and that part of its infrastructure was carrying elevated load, then began shifting traffic across its global network. Messaging and meetings largely kept working, while mail, calendar details and Copilot did not. Governments and airlines reported disruption at service desks. Microsoft shares fell as the incident stretched past 24 hours, and the company has not said what caused the load spike on its mail platform.
Nvidia's board announced on Thursday that Jensen Huang has resigned as president, CEO and director, effective immediately, ending a 33-year run at the company he co-founded. Chief financial officer Colette Kress becomes interim CEO while keeping her finance role, and former Synopsys chief Harvey C. Jones becomes the board's first-ever chair. The exit follows disclosures that Huang had two extramarital affairs, one with a woman US intelligence agencies suspect was a Chinese asset. Nvidia shares fell 7.5% in after-hours trading, and the company reports quarterly results on November 17.
Nvidia shares rose about 4.5% in premarket trading on Wednesday, recovering part of a two-day slide that took the stock down 4.3% on Monday and 5.5% on Tuesday. The company has hired the law firm Wachtell Lipton to investigate whether a personal relationship involving chief executive Jensen Huang compromised company secrets or national security. Huang denies wrongdoing and Beijing has rejected the spying claims. Lawmakers have asked him to brief Congress on 18 November, a day after Nvidia reports quarterly results. The stock is still valued near $5 trillion, and AI hardware names remain the market's largest single source of index risk.
Google announced two new systems in its eighth-generation Tensor Processing Unit lineup, the TPU 8tx and TPU 8i Max, both due to be available next year. The 8tx uses hardware-level confidential computing so customer data and model weights stay isolated while a model runs, and pairs TPU 8t silicon with AMD's fifth-generation EPYC processors. The 8i Max packs two TPU 8i chips into one package, doubling compute, on-chip SRAM and high-bandwidth memory throughput in the same rack footprint. Google says nearly 100 customers use its TPUs, and the launch widens its challenge to Nvidia in AI accelerators. Both systems are due on the market next year, and neither is an entirely new silicon design.
Nvidia's board has opened an independent investigation after chief executive Jensen Huang told staff he had a romantic relationship with a Chinese businesswoman whom US intelligence suspects of spying for Beijing. Huang apologised to his family in an email sent on Sunday and said he would cooperate fully. The woman, Alice Huang, is a former executive at Megaspeed International, a Singapore-based Nvidia customer suspected of diverting restricted AI chips to China. The board said it has no reason to believe company secrets or national security were compromised, and Huang remains chief executive. Nvidia shares fell sharply on Monday as investors weighed the export control implications of the case.
The US Commerce Department has approved dozens of Chinese companies to handle Nvidia's H200 chips, according to reporting by the New York Times, including some with ties to firms under national security sanctions. Several approved entities were listed at addresses that appeared to be empty offices. The approvals raise questions about how export licences are checked once they are granted.
South Korea plans to spend 21.3 trillion won, about $16 billion, on semiconductors, physical AI and AI data centers in 2027, nearly double this year's 10.8 trillion won allocation. President Lee Jae-myung presented the figure inside a record 820.9 trillion won budget, up 12.8%, that also sets aside 4.7 trillion won for a sovereign AI model and money to train 2,000 chip-design specialists. About 2 trillion won goes to expanding production capacity and 2.1 trillion to power, water and industrial-complex infrastructure. Defense spending rises 8.2% to 73.3 trillion won. Lawmakers must still approve the package.