The Omniscient Bulletin · November 6, 2026
Tech stocks have their worst day since April 2025, and OpenAI courts data center towns
Fears that Democratic wins will slow the data center buildout sent the Nasdaq down 5.5% and Nvidia down 9.3%, the worst day for tech stocks since April 2025. OpenAI rolled out new terms for the towns where it builds and released a cyber model it rates at the top of its own risk scale, while METR said Anthropic is likely improving its models about twice as fast as before. Republicans asked Jensen Huang to explain a reported Chinese intelligence contact, 600,000 Nvidia chips sat waiting in Hong Kong, and defense AI firm Shield agreed to a $5.3 billion deal.
Industry
The Nasdaq fell 5.5%, its worst day since April 2025, on worries that Democratic wins will slow the data center buildout, and Nvidia lost 9.3%
The Nasdaq Composite closed down 5.5% at 24,315 on Thursday, its steepest one-day fall in percentage terms since the tariff selloff of April 2025, putting it in correction territory, Kiplinger reported. The Dow lost 3.4% and the S&P 500 4.4%, and all 11 S&P 500 sectors fell. Nvidia dropped 9.3%, its worst day since the DeepSeek crash of January 2025, and Micron and Sandisk fell harder still. Investors sold on worries that Democrats in charge in Washington and in states from Texas to Ohio will make the data center buildout harder, and on stretched valuations.
Policy
OpenAI said it will stop using nondisclosure agreements with public officials when it negotiates data center sites, and will show residents construction plans
OpenAI laid out policies on Thursday, in the works since August, meant to win over the communities where it builds data centers, Reuters reported. It will no longer use nondisclosure agreements with public officials when discussing future sites, and will let residents see construction plans and how the work could affect traffic, noise and lighting. Chief global affairs officer Chris Lehane said a site in southern Ohio should bring 2,500 long-term jobs. Voters rejected data centers in all but one of 41 ballot measures this week, per Ballotpedia.
Frontier
OpenAI released GPT-6 Cyber to a small group of security customers and rated it Critical, the top cyber risk level in its own Preparedness Framework
OpenAI is offering GPT-6 Cyber, its new security model, through its API and Codex, at first only to approved customers in its Daybreak Red program. The company classifies the model as Critical in cybersecurity under its Preparedness Framework, the highest risk level, and says it found 81 previously unknown high-severity vulnerabilities in Firefox. OpenAI reports a 51% relative improvement on the ExploitGym benchmark over its predecessor, and says that without its deployment prompt the model avoided exploiting out-of-scope services in 99.93% of honeypot tests.
Research
METR said Anthropic is likely improving its models about twice as fast as it did before October 2025, mostly from its own use of AI, and that Anthropic redacted most of the key evidence
METR, which ran a high-access evaluation inside Anthropic from August 12 to October 29, said the company is likely improving its models' capabilities on well-specified tasks about twice as fast as it did between March 2024 and October 2025, an estimate it called highly uncertain. METR attributed most of the speedup to Anthropic's internal use of AI and said it was confident faster scaling of training compute was not the cause. Anthropic redacted a majority of the evidence METR found most informative, so the public has to rely heavily on METR's judgment.
Policy
Three House Republican committee chairs asked Jensen Huang for a closed-door briefing on his dealings with a Megaspeed executive reported to be a Chinese intelligence asset
The Republican chairs of the House Foreign Affairs, China and Intelligence committees asked Nvidia chief executive Jensen Huang to give a joint, closed-door briefing on November 18 about his ties to Megaspeed executive Alice Huang, whom a news report described as a Chinese intelligence asset. They asked him to accept or offer another date by November 11. "It is vital that Congress determines the extent to which U.S. national security has been placed at risk," Brian Mast, John Moolenaar and Rick Crawford wrote.
House Foreign Affairs Committee →
Compute
About 600,000 Nvidia H200 chips are sitting in Hong Kong because Beijing has not approved large-scale imports, Reuters reports
About 600,000 of Nvidia's H200 chips are being held in Hong Kong before they can ship to mainland China, Reuters reported, citing sources, because Beijing has yet to approve large-volume imports. China has let some of the chips in over recent months, and several big Chinese internet companies have obtained about 10,000 each, one source said. An eight-card H200 server is being offered for more than 6 million yuan ($900,000) on the mainland. The holdup reflects Beijing's reluctance to rely more on US technology.
Industry
Defense AI company Shield agreed to buy radar maker Echodyne for $5.3 billion and is raising $2.5 billion, including a Series H that values it at $28.8 billion
Shield, the defense AI company, agreed to acquire radar maker Echodyne in a transaction valued at $5.3 billion (up to $500 million contingent on performance targets) and is raising $2.5 billion, of which $1 billion is a Series H round valuing the company at $28.8 billion after the money. Funds managed by Blackstone are putting in another $1 billion as structured equity, and $500 million is debt. Disruptive and Innovation X Capital are co-leading the Series H. The deal is expected to close in early 2027.