Econify.AI Newsletter, November 13, 2026

Nvidia's CEO resigns, and the financing behind AI capacity is now a question

The New York Times reported that a classified memo identified Alice Huang, an executive at the Singapore cloud provider Megaspeed, as a suspected Chinese intelligence agent. Jensen Huang admitted an affair on November 8 and resigned on November 12. CFO Colette Kress is interim CEO, and Harvey C. Jones is the board's first chair. Nvidia says its roadmap and customer commitments continue uninterrupted.

The business questions sit underneath. The Information reported that at least $7.5 billion of Nvidia chips may have reached Chinese companies through Megaspeed, and that Commerce is drafting rules to close the remote-access loophole. Commentators also pointed out that Nvidia has backstopped financing for smaller cloud providers and AI labs, and new leadership may pull that back.

Econify's take: Under Jensen Huang, Nvidia was one of the strongest open source AI advocates. We're watching whether his replacement is as supportive of open source or whether AI ends up under the control of large corporations, frontier models, and government.

Claude adds spreadsheets, whiteboards and shared dashboards, and one switch flips November 19

Anthropic launched Claude Whiteboards and Claude Sheets in beta and made Claude Dashboards generally available as Claude Data, all on paid plans. Whiteboards is a shared board where your team brainstorms with Claude. Sheets builds and edits spreadsheets with you, though Anthropic says its functionality is simple for now and it exports to Excel. Claude Data connects to a warehouse or CRM and lets your data team certify the queries everyone should use, so people stop building dashboards with conflicting numbers.

On Enterprise plans, Whiteboards and Sheets are off by default. Claude Data turns on by default November 19, 2026. Microsoft's stock reportedly fell on the news.

Econify's take: Entering office productivity could signal that Anthropic wants revenue beyond selling tokens. Another interesting question is whether this is also a signal that frontier models may be plateauing.

Anthropic's new top model targets big projects, and Anthropic still points daily work to Opus 5.5

Anthropic released Claude Fable 5.2 to all users and Mythos 5.2 to vetted cybersecurity and life-science organizations. Pricing is $8 per million input tokens and $40 per million output tokens (tokens are the chunks of text AI is billed on), with a faster mode at double that. Anthropic reported that 1,000 agents formalized the proof of Fermat's Last Theorem in 30 hours. It also recommends Opus 5.5 as the daily driver.

Reviews split. Every's review said Fable 5.2 can take a whole product from start to finish. Others found it expensive and no better than Opus 5.5 for routine work. One independent analysis put its cost at about $27 per task at maximum effort and about $9 at high effort.

Econify's take: We'd keep Sonnet 5.5 as the default and test Fable on one bounded, long-running project, such as a legacy system migration, with a named human reviewer. It's likely to be expensive, and few of our clients have work that justifies that price.

Google's Gemini 4 Carbon nearly matches the top model on a coding test at a fraction of the price

Google released Gemini 4 Carbon after holding it for the government's voluntary safety testing. On a bug-hunt benchmark from tester Pavel Hurin, which hides 105 known bugs in real software, Carbon found 57 and Fable 5.2 found 58. Carbon's run cost about $31, against nearly $140 for the Fable run. One render test produced similar results to Fable 5.2 at about a third of the cost.

Google DeepMind also said it used Carbon to formally verify security properties of the Signal messaging software in about a month, work it said could otherwise take teams decades of combined effort.

Econify's take: If your team already uses Google's tools, run Carbon against your current tool on one real task.

Other news

Anthropic's IPO roadshow is under way, with a proposed range of $70 to $78 per share and an implied valuation of $1.25 trillion to $1.4 trillion, which underwriters were reportedly considering raising to about $1.5 trillion at the top end, and trading is reportedly set to start November 19. AI stocks sold off sharply after Democratic midterm wins, and Sanders called for a national data center moratorium. The same week, OpenAI closed $11 billion at a $1.25 trillion valuation against a target of up to $30 billion, Microsoft began winding down an Oracle compute rental, and Google took over Crusoe's troubled Wyoming data center with $17 billion planned. Menlo Ventures estimates enterprise AI spending reached about $177 billion, while Apple cut its engineering teams' daily AI cap from $300 to $200. METR found Anthropic's model improvement pace roughly doubled, mostly through AI-assisted research. Eighth Rank researchers mapped a personal knowledge graph that macOS builds from your contacts and activity, and an Anthropic engineer urged reviewers to form a hypothesis before prompting agents.