All articles
Weekly Review

Weekly Tech Review: November 2–8, 2026 — Nvidia Opens an Investigation Into Its Own CEO, and the AI Money Gets Harder to Raise

PromptyBox Team· November 10, 2026
Nvidia CEO investigation and AI funding slowdown, featuring Jensen Huang, falling investment, and data centre restrictions against a dark cinematic background.

Two things happened this week that look unrelated and are not.

Nvidia's board opened an independent investigation into its own chief executive. And the two best-funded companies in artificial intelligence both raised money below what they were asking for.

For about two years the AI trade has run on the assumption that capital is effectively unlimited and the people running these companies are beyond question. This was the week both halves of that got tested.

Here is what happened, with the confirmed items separated from the reported ones.

Nvidia's board is investigating Jensen Huang

On 8 November, Nvidia's board opened an independent investigation into chief executive Jensen Huang. This is confirmed by the company's own statement, not inferred from reporting.

The sequence, as far as it is established:

  • Huang disclosed an affair with a former executive at another company.
  • A US policy memo had linked that person to China's Ministry of State Security.
  • The New York Times and Financial Times reported that US intelligence agencies assess she has been passing information to Beijing.
  • A committee of independent directors, with outside counsel, is running the review.
  • On 10 November those directors retained Joon H. Kim of Wachtell Lipton, a former acting US Attorney for the Southern District of New York, to conduct it.

The board states it has no reason to believe national security or company secrets were compromised, while saying the inquiry will examine all available information. It also expressed confidence in Huang and his leadership team. In a memo to staff, Huang said he had fallen short, supports the investigation and will cooperate fully.

Nvidia closed down 4.27% on Monday, erasing close to $200 billion in market value.

Some restraint is warranted here. No findings have been released. An allegation about a third party is not a finding about Huang, and the board's own position is that nothing is known to have been compromised. We are not naming the other person: she is not a public figure, and her identity adds nothing to the part of this that actually matters.

What does matter is the governance question. One individual is unusually central to Nvidia — to its strategy, its China relationships and its standing with regulators. A $200 billion single-day move on a governance question, before any finding exists, is the market pricing that concentration rather than pricing the allegation.

The money got harder

Three separate events, same direction.

OpenAI raised $11 billion at a $1.25 trillion valuation — below the $1.5 trillion it had sought. It has already filed confidentially to go public as soon as next year.

Anthropic began its IPO roadshow, at a reported $70–78 per share. Valuations in circulation run from $1.1 to $1.4 trillion, but Reuters reported some investors weighing a lower figure after seeing its latest financial disclosures.

Meta went to the bond market for €53 billion to fund its buildout.

Two of the most sought-after companies in the industry priced below ambition within a fortnight of each other. That is not a collapse — these are still among the largest valuations ever attached to private companies. But for a year every round cleared its target, and this week two did not.

Elsewhere in the capital story: Marvell agreed to buy Lumentum for $115 billion, and investors sold shares in both companies — a reasonably clear verdict on the price. Google committed $15 billion to AI. And Anthropic bought $3.4 billion of gas generators to power its data centres, which tells you where the actual constraint now sits.

Voters started saying no to data centres

Less covered and arguably more consequential: in the midterms, voters approved restrictions on AI data centres in dozens of communities.

Compute has become a local ballot issue. Every projection of AI capacity assumes the land, power and permits arrive on schedule — and permits are granted by exactly the local bodies that just received a mandate to say no. This is a slower constraint than funding, and a harder one to engineer around.

AI found another thing humans missed

On 6 November, Google DeepMind said AI agents it guided produced two cryptographic advances against legacy 1024-bit RSA, in a matter of weeks.

If that sounds familiar, it should. Our last review covered PHP2Shell, where a decade-old remote code execution flaw in PHP was found by two firms independently pointing language models at source code — and attackers were exploiting it four days before the patch.

Same pattern, higher stakes. 1024-bit RSA is already deprecated, so this is not an emergency. But the capability on display is: machine-assisted analysis finding structural weaknesses in systems that were considered settled, on a timescale defenders are not organised for.

Also this week

Anthropic released Fable 5.2 and Mythos 5.2, and shortly afterwards warned that model capability is advancing faster than safety measures can keep pace with — an unusual thing to publish during your own IPO roadshow. It also brought Claude into Google Meet to handle tasks during live calls.

Tesla plans to end Cybertruck production around mid-2027, freeing capacity for the Cybercab. Canva is restructuring more than a fifth of its workforce.

And a security note worth acting on: a researcher demonstrated that a plainly-worded question could pull account details out of Google's connected apps, and Google is now advising users not to connect apps holding confidential information. That is prompt injection arriving in consumer products — if you have connected anything sensitive to an AI assistant, this is the week to review it.

What the week actually said

The AI story has been told as a straight line: more capital, more compute, more capability. This week introduced friction at three separate points — governance, funding and permitting — and none of them is a technical problem that a better model solves.

The companies are not in trouble. But the assumption that money and permission arrive automatically took its first real knock, and that assumption was doing a lot of work in everyone's projections.


Sources: Nvidia Investor Relations, CNBC, Bloomberg, TechCrunch, NBC News, The Daily Upside, Crypto Briefing.

Using AI rather than reading about it? Our prompt library is free and needs no account.

Enjoyed this article?

Get trending prompts and new articles like this one, delivered weekly. Free, no spam, unsubscribe anytime.

Weekly Digest

Trending prompts & new articles, weekly. No spam.

Explore More on the Blog