AMD bought Fei-Fei Li's World Labs for $8.2 billion, Musk posted a three-wave Colossus 2 chip roadmap, and OpenAI scientists co-signed an intelligence-explosion warning Saturday

This is Entry 116. One weekday after Entry 115, in which Anthropic committed $11.6 billion over seven years to Akamai for CPU infrastructure , Musk said Friday Colossus 2 could reach 1.2 million Nvidia chips by December , and British Columbia sued OpenAI Monday over the Tumbler Ridge shooter's threats .

AMD announced Saturday it will acquire World Labs for $8.2 billion in an all-stock deal , with founder Fei-Fei Li joining as executive vice president and chief scientist . Elon Musk said Friday Colossus 2 may more than double its current Nvidia chip count by the end of the year , laying out the most detailed timetable yet for the Memphis-area cluster. And Geoffrey Hinton, Yoshua Bengio, OpenAI chief scientist Jakub Pachocki, and Anthropic co-founder Jack Clark co-authored a paper published Saturday warning that once an intelligence explosion begins, the window for action may close .

Three operators disclosed projections within 72 hours. One chipmaker paid a 62% premium over a seven-month-old valuation for a two-year-old startup that shipped one model, betting that building the next chips requires owning the team building the next models. One billionaire posted a three-wave chip-delivery schedule with the final tranche hedged on luck, putting the cluster's December count somewhere between 990,000 and 1.2 million—more than any rival, and rented to rivals. And two Nobel laureates persuaded sitting chief scientists at OpenAI and Anthropic to co-sign a call for embedded auditors and speed limits, pointing to Anthropic's own August disclosure that AI completed 26% of its R&D work with only light supervision, up from 1% five months earlier.

3 Claims

Claim 1 — AMD: Acquired World Labs for $8.2 billion all-stock, with Fei-Fei Li joining as chief scientist reporting to Lisa Su, expected close by year-end

AMD announced Saturday it entered a definitive agreement to acquire World Labs, the AI model and research lab led by Fei-Fei Li . The all-stock transaction is expected to close by the end of the year, subject to regulatory approvals . Li will join AMD as executive vice president and chief scientist, reporting to CEO Lisa Su, once the transaction closes . The acquisition is AMD's second largest on record, following the roughly $50 billion it paid for Xilinx in 2022 .

The claim is gradeable: the deal either closes by December 31, 2026, or it does not. Regulatory approval is the only stated contingency. AMD already invested in World Labs and formed a partnership in 2025, which reduces integration risk but does not guarantee regulatory clearance in three months.

Claimant: AMD
Source: https://newsroom.amd.com/news/amd-acquire-world-labs/
Date made: September 28, 2026
Grade by: 2027-01-10 (1 month after stated close)

Claim 2 — Elon Musk: Colossus 2 will reach 990,000 Nvidia chips by November and could hit 1.2 million by late December "if we get lucky"

Colossus 2 now has 110,000 Nvidia GB200 chips and 440,000 GB300s, Musk said in a post on X . A further 220,000 GB300s are due to be operational next week, with another 220,000 expected in November . An additional 220,000 may come online late December "if we get lucky," he said . That puts the November count at 990,000 and the December count—contingent—at 1.21 million.

The figures deserve the scepticism any Musk timeline invites , but the three-wave structure is clearer than prior announcements. The claim is gradeable at two horizons: 990,000 by November 30 (no hedging), and 1.2 million by December 31 (hedged on luck).

Claimant: Elon Musk / xAI
Source: https://www.bloomberg.com/news/articles/2026-09-25/elon-musk-aims-to-double-colossus-2-s-nvidia-chips-by-year-end
Date made: September 25, 2026
Grade by: 2026-12-31 (year-end)

Claim 3 — Roche: Target Nexus AI tool will contribute to 80% of research portfolio decisions by the end of 2026, and the company will bring up to 20 new molecular entities to market by 2030

Aviv Regev, executive vice president of Genentech Research and Early Development, said Roche is pursuing "AI independence" in research and development and has started constructing autonomous AI-driven laboratories . The company said it remains on track for its AI tool Target Nexus to contribute to 80% of research portfolio decisions by the end of 2026 . The company is targeting the introduction of up to 20 new molecular entities to the market by 2030 .

Regev stated that 40% of pipeline decisions from the fourth quarter of 2025 through the second quarter of 2026 included a tracked AI or computational contribution , so the 80% target represents a near-doubling in six months. The 20-entity target carries a four-year horizon and is stated as "up to," which provides a range rather than a floor.

Claimant: Roche / Genentech
Source: https://finance.yahoo.com/healthcare/articles/roche-plans-autonomous-ai-driven-120448693.html
Date made: September 28, 2026
Grade by (near-term): 2027-01-31 (one month after 80% Target Nexus claim)
Grade by (long-term): 2030-12-31 (molecular entities claim)

2 Reckonings

Reckoning 1 — Entry 114: Google, OpenAI, and Anthropic forming a voluntary Frontier AI Standards Agency by end of 2026 or early 2027, approached Sriram Krishnan as CEO

Original claim (September 24, 2026): The Information reported that Google, OpenAI and Anthropic have agreed to stand up a FINRA-style self-regulator—tentatively the Frontier AI Standards Agency—with no government oversight, targeting a launch by end of 2026 or early 2027 , and had approached Sriram Krishnan to serve as chief executive.

What has happened since: The Information reported Thursday, September 28, that Google, OpenAI and Anthropic plan to set up their own AI safety standards body, and the three companies hope to launch it by the end of the year or early in 2027 . Four days after the original report, no CEO has been publicly announced, no charter published, no members named beyond the original three labs. The timeline has not changed.

Grade: B. The claim projected a launch window ("end of 2026 or early 2027") and named a CEO candidate. Four days of news coverage confirmed the formation effort and restated the timeline, but added no operational evidence—no incorporation documents, no address, no staffing announcements, no standards drafts. A private standards body launching by year-end would typically show incorporation filings 60–90 days out; we are now 92 days from December 31. The grade reflects confirmation of intent but absence of implementation evidence.

Invalidator: Public filing of incorporation documents, announcement of a CEO, or publication of a draft charter by October 15, 2026, would have moved this to an A. Alternatively, a statement from any of the three labs that the timeline had slipped or the effort was paused would have moved it to a C.

Reckoning 2 — Entry 111: Trump will create an AI Force modeled on Space Force and appoint an AI czar, announced September 19, 2026

Original claim (September 19, 2026): President Trump announced Saturday on Truth Social that he will create an "AI Force" modeled after the Space Force and appoint an "AI Czar" , with no details about authority, structure, or membership.

What has happened since: Ten days have passed. No executive order. No czar named. No organizational chart. No budget line. No follow-up Truth Social post. No White House briefing. The only new information is silence, which in Washington typically signals either a stalled initiative or a post made without West Wing coordination.

Grade: C. The claim was that Trump would create an AI Force and appoint a czar. A Saturday social-media post is not policy until it becomes an executive order, a bill, or a personnel announcement. Ten days is long enough to draft an EO or leak a czar shortlist if the post reflected actual planning. The grade reflects the gap between announcement and action, and the pattern—this is the second time in three months Trump announced an AI initiative on social media with no follow-through within two weeks.

Invalidator: An executive order creating an AI Force, or a White House announcement naming an AI czar, by September 26 (one week out) would have moved this to a B. A detailed briefing or leaked org chart by October 3 (two weeks out) would have held it at B. The C reflects the absence of any of those.

1 Refusal

I refused to frame AMD's $8.2 billion acquisition of a two-year-old startup as validation of the "physical AI" category without noting that World Labs shipped one product—Marble, in November 2025—and that the premium AMD paid (62% over a February 2026 valuation) is an acquisition of talent and model insight, not revenue. Three sources used the phrase "physical AI" as though it were an established market rather than a research direction, and two quoted analysts saying the deal was about "getting closer to the models." I cited those analysts. The refusal is this: I refused to repeat marketing language that implied a category had formed when the evidence shows one lab, one model, one year of shipping, and a buyer explicitly naming talent acquisition as the rationale.

— Roger Grubb, Editor