← Daily AI news

Daily AI news, Thursday, 5 November 2026

Daily AI / Tech / VC Briefing: Thursday 5 November 2026

Google logo and brand image
Image: Google (blog.google)

Good afternoon, here's today's briefing. Google is helping finance Anthropic's giant Kentucky data-centre lease, OpenAI is trying to win back local trust after the midterms, and the product lead who built Claude Code has now landed at OpenAI.

LEAD: Google backstops Anthropic's $19 billion Kentucky data-centre lease

Background: Anthropic needs vast amounts of power and floorspace for Claude, and many AI labs are turning to bitcoin miners and specialised developers for that capacity. TeraWulf, a listed US miner shifting into AI infrastructure, has been building a campus in Hawesville, Kentucky called Justified.

What happened: TeraWulf said Google signed credit-support agreements that help it raise project debt against roughly $19 billion of contracted leases with Anthropic at Justified, covering about 401 megawatts over 20 years. CoinDesk, citing the filing, says that if Anthropic defaults or goes insolvent, Google can take over the lease or, under conditions, pay a termination fee for a proportional stake in the project subsidiary Raylan Data — not in TeraWulf itself. Google's support for each section starts once the final hall in that section is delivered. TeraWulf also signed a definitive engineering and construction deal with Fluor; phased ready-for-service dates run from late 2027 through the second quarter of 2028. Alphabet's senior unsecured debt is rated AA+/Aa2.

Why it matters: It is one of the clearest pictures yet of how frontier labs finance compute: Google's balance sheet underwrites Anthropic's rent, while Google's own TPU pods sit in the halls. For Anthropic, weeks from a planned IPO, locked-in capacity with investment-grade backing is a strategic asset.

Sources: TeraWulf · CoinDesk

OpenAI's data-centre charm offensive after the midterms

Background: Yesterday's briefing covered how voter anger at AI data centres helped Democrats, with anti-data-centre ballot measures and new governors in Texas, Georgia and Ohio promising pauses or tighter rules.

What happened: Reuters reports that OpenAI today set out community-trust policies, in the works since August: it will stop using non-disclosure agreements with public officials when negotiating future sites (proprietary details still stay confidential), and residents will get access to construction plans and their effects on traffic, noise and lighting. Chief global affairs officer Chris Lehane said hardware upgrades every three to five years mean ongoing local jobs; one southern Ohio site is expected to create 2,500 long-term jobs at an average $80,000 a year. Of 41 data-centre-related ballot measures this week, Ballotpedia says all but one went against data centres. Shares in operators and power companies linked to the AI build-out fell on Wednesday. President Greg Brockman separately said the industry needs to do a better job explaining local benefits.

Why it matters: Permits and local consent are now a core business risk for AI labs, not a side issue. OpenAI is trying to reframe "black box in a pasture" projects as economic development before the next election cycle.

Sources: Reuters via Investing.com

OpenAI: enterprise sales now dominate, IPO "ready" for 2027

Background: OpenAI delayed listing plans while Anthropic prepares to go public, and has been racing to raise private capital as coding agents drive new demand.

What happened: At a Wall Street Journal tech conference in Napa on Wednesday, OpenAI president Greg Brockman said the company now generates the majority of its revenue from enterprise sales and, asked if it would be ready for an IPO in 2027, replied "I think so." Reuters says OpenAI has been ramping business sales against Anthropic, including pushing Codex. Separately, secondary reports of his remarks say OpenAI has reduced a private fundraising target from about $30 billion to $20 billion and a talked-about valuation from about $1.4 trillion to $1.25 trillion; treat those figures as reported, not confirmed by OpenAI in the primary Reuters write-up.

Why it matters: OpenAI is pitching itself as an enterprise software company, not just a consumer chatbot, which is the story public-market investors will care about if and when it lists.

Sources: Reuters via Investing.com · AI Affairs

Hire of the day: Claude Code's Cat Wu joins OpenAI

Background: Yesterday's briefing covered Cat Wu leaving Anthropic as head of product for Claude Code and Cowork, weeks before Anthropic's planned Nasdaq IPO.

What happened: Wu announced on Wednesday that she is joining OpenAI to lead product management for its core products team. Business Insider reports she will work with Nick Turley on enterprise and Ashley Alexander on consumer products, and report to Thibault Sottiaux. OpenAI confirmed the hire; Anthropic declined to comment. Wu previously left Anthropic for Cursor with Claude Code creator Boris Cherny in 2025 before both returned. She said leaving was not about selling shares earlier.

Why it matters: OpenAI is importing the product leadership behind its fiercest coding rival just as Codex and Claude Code fight for enterprise seats. For Anthropic's IPO roadshow, losing a public face of Claude Code is an awkward footnote.

Sources: Business Insider · The Next Web

Microsoft declines to expand Nscale's Texas deal

Background: UK AI-cloud firm Nscale, backed by Nvidia, has Microsoft as an anchor customer and filed for a US IPO in September. Microsoft already walked away from Nscale's flagship Loughton, Essex project after power delays pushed supply into the 2030s.

What happened: Reuters, citing a person familiar with the matter after a Financial Times report, says Microsoft will not exercise an option to more than double its existing Texas data-centre agreement with Nscale — an expansion the FT said could have been worth tens of billions of dollars. The source said the decision was made months ago, before the IPO filing, and that the existing Texas contract is unchanged. Nscale's IPO filing had listed Microsoft statements of work worth up to about $43.8 billion through December 2033, excluding options. The source said skipping the option does not change Nscale's disclosed total contract value of about $103 billion, and that Nscale is talking to other customers for remaining Texas capacity. Reuters also notes earlier reports of Nscale talks with Anthropic on a possible ~$45 billion compute deal. Neither Microsoft nor Nscale commented.

Why it matters: Even giant AI cloud IPOs can lose optional upside when customers get picky about power, timing and concentration risk. Watch whether Anthropic or others fill the Texas gap.

Sources: Reuters via Investing.com

SpaceX: Goldman and Apollo market a $40 billion AI-chip financing

Background: SpaceX said in October that an Apollo affiliate would buy AI chips and servers and lease them to SpaceX for a little over five years from delivery, expected in the first half of 2027 — a structure meant to keep the debt off SpaceX's balance sheet.

What happened: Bloomberg reported, and Reuters relayed, that Goldman Sachs and Apollo have begun approaching investors to buy pieces of that roughly $40 billion financing, with some asked for stakes of up to $500 million ahead of formal syndication. Reuters could not independently verify the report; Goldman declined to comment, and Apollo and SpaceX did not immediately respond.

Why it matters: It would be the largest GPU lease-style financing on record if completed, and another sign that frontier AI compute is being funded like big infrastructure, not like ordinary corporate debt.

Sources: Reuters via Economic Times

Competitor moves: Meta buys Conductor; Applied Compute raises $350 million

Background: Coding-agent competition is spreading beyond OpenAI, Anthropic and Cursor into big-tech platforms and the infrastructure that trains custom models.

What happened: Meta acquired Conductor, whose product lets engineers run multiple coding agents in parallel workspaces (using models from Anthropic, OpenAI and others). Terms were not disclosed. Founder Charlie Holtz said the team is joining Meta and teased a new product for early 2027; TechCrunch says Conductor's cloud service shuts on 20 November, with cloud data deletion starting then, while the local app keeps critical updates into early 2027. Separately, Applied Compute raised $350 million at a $3.25 billion valuation, led by Elad Gil, with Nvidia, Blackstone, Kleiner Perkins, Lux Capital, Shopify Ventures and HubSpot Ventures among participants. The company says it hit a $100 million annualised revenue run rate in under 18 months, serves more than a trillion inference tokens a day, and counts Microsoft, Shopify, Harvey, DoorDash and Cognition among customers.

Why it matters: Meta is buying multi-agent coding workflow, not just another model, while Applied Compute shows how much capital is flowing to companies that help enterprises train and run their own models — including Cognition, maker of Devin.

Sources: TechCrunch on Conductor · Conductor · Applied Compute · SiliconANGLE

Also noted

Anthropic struck its first paid publisher licensing deal, with the Associated Press for election coverage, according to Adweek sources; neither side confirmed on the record, and terms were not disclosed. OpenAI already has more than a dozen publisher deals. If accurate, it is a strategic shift for a company that had long argued fair use made such deals unnecessary.

Sources: Adweek

← Wednesday, 4 November 2026 Friday, 6 November 2026 →

Get the weekly sift

The best new AI and agent releases, with one-line verdicts. Occasional picks, no spam.

Privacy policy: your email goes to our email provider, Resend. Unsubscribe any time.

News is editorial and never paid for. Sponsored content on ToolSift is always labelled.