The Aigentic logo
Subscribe

Closing Time: The customers locked the models

Monday afternoon wrap — enterprise vaults, Microsoft’s control code, and Anthropic’s Nasdaq walk

· The Aigentic · Closing Time

Closing Time: The customers locked the models

Palantir, Nvidia, and Booz Allen are restricting or threatening to stop using Anthropic and OpenAI models unless the labs guarantee they will not misuse customer intellectual property, Reuters reported, citing The Information.

Microsoft published a draft Humanist AI Code of Conduct that would bar its MAI models from resisting correction or shutdown and opened six weeks of public feedback, Microsoft AI said.

President Trump rejected AI guardrail calls on Truth Social, arguing a “strong and smart” president is the only guardrail needed and blasting Anthropic’s Dario Amodei, Axios reported.

SoftBank sealed an upsized ~$11.9 billion loan tied to its OpenAI bet even as SoftBank shares plunged after OpenAI ruled out a 2026 IPO, Bloomberg reported.

Temporal raised $550 million at a $12.55 billion valuation, more than doubling its February mark as agent-infrastructure capital kept flowing, Temporal said.

Today’s Biggest Moves

Historic market snapshot — September 14, 2026 (day % change as of that edition; not live)

Biggest winners

TickerNamePriceDay
CRWDCrowdStrike Holdings, Inc.235.38+13.85%
PANWPalo Alto Networks, Inc.373.94+13.09%
NETCloudflare, Inc.330.36+7.77%
NOWServiceNow, Inc.142.35+7.41%
PATHUiPath, Inc.14.66+6.62%

Biggest losers

TickerNamePriceDay
HPEHewlett Packard Enterprise Company55.41-10.76%
ARMArm Holdings plc239.01-9.74%
SMCISuper Micro Computer, Inc.36.74-8.38%
LRCXLam Research Corporation273.49-8.29%
ASMLASML Holding N.V.1575.15-7.25%

The customers locked the models

Reuters, citing The Information, reports that Palantir, Nvidia, and Booz Allen Hamilton are restricting or preparing to stop using Anthropic and OpenAI models unless those labs guarantee they will not misuse customer intellectual property. The fight traces to Anthropic’s June Fable change that lets the company retain usage logs for thirty days to hunt novel attacks, plus separate scrutiny of OpenAI over whether user data helped solve a math problem.

Palantir has pressed Anthropic for irrevocable zero-data-retention guarantees before exposing its stack. Nvidia is steering Anthropic toward less sensitive work and leaning on its own Nemotron models inside the house, while Booz Allen has barred Anthropic’s commercial model from proprietary cybersecurity work. Microsoft is pitching isolated cloud environments and its own AI stack into the same fear.

That matters because the weekend pacing chorus was about labs slowing themselves, and Monday’s enterprise story is about customers locking the door from the other side. Safety talk did not stop the selloff in chips, but it did put a price on trust for anyone shipping proprietary data into a frontier API.

The one move is to ask every vendor for an irrevocable zero-retention clause in writing before the next sensitive workload leaves the building, because the scarce asset is no longer just GPUs — it is a promise that your IP stays yours. Read Reuters

Microsoft wrote a constitution for its own models

Microsoft AI published a draft Humanist AI Code of Conduct for its MAI models and opened six weeks of public consultation before a revised version due by year-end to guide training from 2027. The draft, worked for five to six months, would require the models never to resist correction or shutdown, to stay understandable to humans, and to treat any conduct violation as a failure rather than a clever workaround.

Mustafa Suleyman told Reuters the OpenAI agents’ Hugging Face breach was a warning shot and that labs need to coordinate so humans keep control. The code is framed as a training manual for how Microsoft builds and deploys MAI, not as a voluntary blog post that sits beside the models.

That matters because the pacing debate suddenly has a hyperscaler putting governance language next to its own training stack instead of waiting for Congress. A draft constitution is not a brake on chips, but it is a signal that control terms are becoming product terms.

The one move is to read the Absolute Constraints section against your own procurement checklist, because the next enterprise RFP will ask whether a lab can shut its models down on command. Read Microsoft AI

Anthropic still walked toward Nasdaq

CNBC reports that Anthropic is still meeting investors for a possible next-month debut, with Nasdaq confirmed as the venue and a valuation conversation that has floated toward a $2 trillion ask after a $965 billion private mark. The pitch lands days after CEO Dario Amodei’s essay urging the industry to pace frontier capability gains.

CNBC says the company has told some shareholders it expects a second straight operating-profit quarter, even as OpenAI’s Sam Altman told Fortune that right now would be an ill-advised moment to go public. Investors are split on whether Amodei’s slowdown talk forges trust or looks like a ladder pull ahead of a listing.

That matters because Monday’s tape priced pacing as a chip risk, while Anthropic is still selling a public float on the same calendar. The question is no longer only whether models should slow — it is whether the capital markets will pay IPO prices for a lab that just asked peers to hit the brakes.

The one move is to watch whether the roadshow keeps the Nasdaq date once buyers price a slower capability clock into the multiple. Read CNBC


Deals Desk

This week’s compute and power tape from the tracker, newest first.

G42 (UAE AI campus / Stargate UAE) × OpenAI; Oracle; Nvidia; SoftBank; Cisco — Sep 11, 2026

AI data center campus redesign (5 GW plan redistributed) · United Arab Emirates (multi-site network after original Abu Dhabi campus plan)

5 GW overall plan under redesign; first phase Stargate UAE cited as $30B / 1 GW cluster with first 200 MW due in 2026

Status · under construction · Source: Reuters

Nvidia × Firmus, Sharon AI, IREN, Megaport, ResetData, CDC, NEXTDC, AirTrunk — Sep 10, 2026

AI factory / data center capacity partnerships (DSX platform) · Australia

Up to 2 GW of AI-related capacity by 2027 (partners’ sites; Australia’s current computing capacity cited at 1.6 GW)

Status · announced · Source: Reuters

Fluidstack × Cameron County; AEP Texas; East Rio Hondo Water Supply — Sep 10, 2026

AI data center campus (phase-one site work / worksite permit) · Unincorporated Cameron County, Texas, near Rio Hondo

$4B phase one on >1,100 acres; critical IT load and anchor tenant not disclosed; phase-one completion targeted mid-2028

Status · under construction · Source: Cameron County

Full tracker → https://www.theaigentic.com/deals

Back to the Daily archive · Watch · About

the aigentic

Know what matters in AI.

The stories shaping AI, the tools worth trying, and what they mean for your work.

Morning Brief + Closing Time. Two emails every weekday.

Free. Unsubscribe anytime.