The Omniscient Bulletin · November 5, 2026
Voters turn on data centers, and investors split the AI buildout's winners from its borrowers
The midterms delivered the clearest verdict yet on the AI buildout: voters rejected data centers almost everywhere they were on the ballot, and in premarket trading investors sold the companies borrowing to build them while buying the hyperscalers that already own capacity. Google agreed to stand behind Anthropic's rent at a Kentucky campus, OpenAI described an automated pipeline that wrote 948 Linux root exploits, and Greg Brockman said OpenAI could be ready to go public next year. A bank regulator gave its examiners a framework for supervising AI, and Google confirmed buyouts abroad to contain costs.
Policy
Voters sided against data centers in nearly all 41 local ballot measures, Ballotpedia and CNBC found, and 56% told a CNBC poll data centers were a crucial or major factor in their vote
Voters backed data center restrictions or rejected pro-development measures in nearly every one of 41 local ballot measures in 10 states with results available Wednesday, per Ballotpedia and CNBC. The exception was Harwood, North Dakota, where voters rejected added review of industrial projects. Ohio had 21 of the measures. In a CNBC Voter Poll, 56% called data centers a crucial or major factor in their vote. Rep. Greg Casar, who chairs the Congressional Progressive Caucus, said after Democrats were projected to win the House that "AI company CEOs should be called to testify under oath."
Compute
Data center and power stocks fell in premarket trading after Democrats won Congress, with CoreWeave down 9.5% and Vistra 15.4%, while Amazon and Microsoft each rose 6%
Data center stocks slid in premarket trading Wednesday after Democrats won Congress: Applied Digital fell 8.8%, CoreWeave 9.5% and Cipher Digital 21%. Power suppliers fell too: GE Vernova 11.5%, Constellation Energy 9.2% and Vistra 15.4%. Amazon and Microsoft jumped 6% each as investors bet slower construction would make existing data centers more valuable. Siebert Financial CIO Mark Malek said, "The most vulnerable AI companies are the borrowers, not necessarily the biggest spenders."
Compute
Google agreed to backstop Anthropic's rent on 401 megawatts at a TeraWulf data center campus in Kentucky, and can take over the leases if Anthropic defaults
TeraWulf disclosed in an 8-K on Wednesday that Google agreed to backstop certain of Anthropic's obligations under 20-year leases for about 401 MW at its Hawesville, Kentucky campus, leases worth about $19 billion in commitments. The backstop starts once the final data hall under each lease begins. After a default or insolvency, Google can pay the rent due and assume the lease or, in some cases, pay a termination fee for equity in TeraWulf's Raylan subsidiary. It ends once Anthropic or a guarantor affiliate holds certain investment-grade ratings for at least three months.
Research
OpenAI says a pipeline running an unreleased cyber model produced 948 working Linux root exploits in a single run costing $1.2 million
OpenAI on Wednesday detailed an agent pipeline that used an upcoming cyber-focused model to find, reproduce and patch Linux kernel vulnerabilities. In a mid-October run, it reproduced 1,724 vulnerabilities within 30 hours and produced 948 exploits that let an unprivileged user gain root access, for about $1.2 million in API credits, or $1,279 per exploit. OpenAI said that, to its knowledge, the agents used at least 10 novel exploitation techniques, and that researchers review each exploit and patch before reporting to maintainers.
Industry
OpenAI President Greg Brockman said most of its revenue now comes from businesses and that he thinks it will be ready to go public in 2027
OpenAI President Greg Brockman said on Wednesday that enterprise sales now bring in most of the company's revenue and that he believes it will be ready to go public in 2027, Reuters reported. Speaking at the Wall Street Journal's tech conference in Napa, California, he said OpenAI has been ramping up sales to businesses, including its Codex agent, in competition with Anthropic, which Reuters says is likely to go public before year end. Asked on stage whether OpenAI would be ready for an IPO next year, he said, "I think so."
Policy
The US bank regulator OCC gave its examiners a framework for supervising banks' use of AI, which it calls Super Intelligence
The Office of the Comptroller of the Currency issued a framework on Wednesday for its bank examiners to evaluate banks' use of what it calls Super Intelligence, including agentic systems. A footnote in the framework PDF defines the term as the technologies and systems previously covered by the term artificial intelligence, citing Executive Order 14434. The release says examiners should not treat the technology as a separate risk category, and banks should not have to define a separate risk appetite for it. The OCC said it will update the framework regularly.
Industry
Google confirmed it expanded voluntary buyouts to staff in Canada, France, Germany, Japan and Switzerland this year as it reins in costs amid AI spending, Reuters reports
Google confirmed to Reuters that it began offering voluntary exit packages in June to some staff in Canada, France, Germany, Japan and Switzerland, as it tries to contain costs amid heavy AI spending. Google Cloud and DeepMind staff were excluded from earlier offers, and DeepMind was left out in Switzerland, Reuters reports. Three people familiar told Reuters that Swiss staff who accepted got a month's salary per year of service, up to 12 months. Microsoft offered buyouts to about 60,000 employees on October 26.