Nscale raises $2.36 billion in pre-IPO notes led by Third Point. Nvidia (NVDA) commits $1 billion.
Nscale, which has filed to list on the NYSE, raised $2.36 billion in convertible loan notes led by Third Point, and Nvidia committed another $1 billion due in mid-November. The notes convert into shares automatically when the IPO is completed.
Google (GOOGL) plans to launch its first AI-chip satellite with Planet Labs next week
Google's first Project Suncatcher satellite, developed with Planet Labs and carrying Google TPUs, is scheduled to launch next week on a SpaceX rideshare. Separately, Google began testing a Gemini feature that calls businesses for Pixel 11 owners.
Quantum stocks compared: IonQ (IONQ) vs Rigetti (RGTI) vs D-Wave (QBTS)
IonQ, Rigetti, and D-Wave all sell quantum computers, but they use different technologies and run businesses of very different sizes. IonQ reported $80.1 million of second-quarter revenue, nearly 10 times Rigetti and D-Wave combined.
Three dot-com warning signs are back for AI. I'm watching for a fourth.
Chip and memory prices are climbing, bond buyers want wider spreads on AI debt, and the Fed is raising rates. Those are three warning signs from 2000. Here's the one I'm still waiting for.
IonQ (IONQ) jumps on a quantum error-correction claim and a first system for Nvidia's research center
IonQ shares jumped after the company said it built the first end-to-end real-time quantum error decoder on a single CPU, tested on simulated circuits of up to 408 logical qubits. IonQ also said its Superion 256 will be the first on-premises quantum computer at Nvidia's research center in 2027.
CoreWeave (CRWV) signs multi-year deal to power Harell Data's private biotech AI training platform
CoreWeave signed a multi-year deal to run AI training for Harell Data, which says models on its platform train on proprietary biotech datasets without the data leaving it. Data owners earn a revenue share on each training job. No contract value was disclosed.
Nvidia (NVDA) director Mark Stevens sold about $946 million of stock in three weeks
Nvidia director Mark Stevens sold another 1,366,000 shares, worth about $300 million, from a family trust on September 18. It is his third sale filing in three weeks, and none of them reports a Rule 10b5-1 trading plan.
Huawei unveils an 11-chip AI portfolio and says Ascend demand already tops supply (NVDA, AMD, INTC)
Huawei laid out an 11-chip AI infrastructure portfolio at Huawei Connect on September 17, pulled its next Ascend accelerator forward to early 2027, and said domestic demand is already outstripping supply. It is a direct challenge to Nvidia (NVDA), AMD and Intel (INTC).
CNBC calls Nvidia (NVDA) CEO Jensen Huang one of Trump's top AI allies. Huang is making the case against a slowdown.
CNBC describes Nvidia CEO Jensen Huang as one of Trump's most influential allies in the AI-safety debate. Huang is arguing against new rules and against halting chip sales to China, while Dario Amodei's camp wants the industry to slow down.
What is an AI kill switch? Shutting down modern AI means finding thousands of off switches, not one.
An AI kill switch is meant to shut down a dangerous AI model in an emergency. The catch is that modern AI runs across thousands of machines, so there is no single thing to turn off.
Nvidia-backed Nscale (NSCL) files for a NYSE IPO with $103.4 billion in signed contracts. Almost all of the compute behind them still has to be built and financed.
Nscale filed to list on the NYSE under ticker NSCL, reporting $103.4 billion in contract value as of Aug. 31. Anthropic and Microsoft are about 85% of it, but only about 25,000 of Nscale's 461,000 contracted GPUs were running.
California's Newsom signs an order to develop an AI "kill switch." Two years ago he vetoed a bill that would have required one.
Governor Gavin Newsom signed executive order N-9-26 directing California agencies to develop an AI "kill switch" and stronger safety rules for frontier models, with recommendations due in two months. Two years ago he vetoed a bill that would have required a shutdown capability.
Nebius (NBIS) is raising GPU rental prices up to 21% on October 1. CoreWeave (CRWV) told bond buyers it is charging more too.
Nebius will raise pay-as-you-go rates for renting selected Nvidia chips from October 1, its second increase in three months, with GPUs up 17% to 21% and memory up about 41%. CoreWeave told convertible-note buyers the same morning that it is signing capacity at higher prices too.
Why chip prices are rising in 2026. What Nvidia (NVDA), TSMC (TSM) and AMD have each actually said.
Nvidia is reported to have told customers to expect more than 15% on AI servers, TSMC to have raised advanced-node wafer prices 5% to 10%, Samsung foundry quotes up to 15%, and AMD about 10%. Four separate reports, and four very different levels of evidence.
Taiwan's chip industry has more than two openings per job seeker. The AI buildout needs people.
Taiwan's chip industry has more than two open jobs for every person looking for one. ASE needs at least 3,000 people this year, and one materials maker is calling old university classmates to staff a single new plant. Texas Instruments, GlobalFoundries, and ASE describe the same shortage.
Nvidia (NVDA) CEO Jensen Huang calls new antitrust laws for an AI slowdown "completely unnecessary"
Jensen Huang told Jim Cramer that new antitrust laws for an AI slowdown are "completely unnecessary." Altman, Musk, and Nadella endorsed pacing. Huang's disagreement is narrower: he says each company should decide for itself when a product is not ready.
Anthropic's CEO asked the industry to slow down. The company's reported $100 billion IPO is now aimed at mid-November
Dario Amodei published a call to slow the AI race on September 12. Within 48 hours Reuters reported Anthropic had picked Nasdaq, Nvidia was weighing a $10 billion anchor order, and the company told shareholders it would post a second straight adjusted operating profit. Sam Altman cited the same safety moment as his reason not to list.
The circular AI trade, mapped: Nvidia (NVDA) hardware is named in only one of Anthropic's four money loops
The circular trade everyone describes runs through Microsoft: up to $15 billion in, $30 billion of Azure out, Nvidia hardware named. Anthropic's three larger loops run on Trainium, TPUs and AMD, and only Amazon's discloses the money going back.
Odyssey's new AI drives cars, flies drones and plays Grand Theft Auto. Odyssey is optimizing the model for Amazon (AMZN) Trainium.
Odyssey published results for Odyssey-3, a shared world model it says can be adapted to robot arms, humanoids, cars, drones and video games with a small layer of task-specific control data. There is no stock to buy, but Amazon and AMD are on the cap table and Odyssey is optimizing the model for AWS Trainium.
TSMC (TSM) co-COO calls AI a "three-year-old superman." The company is cautious with what it lets AI see.
Y.J. Mii, co-chief operating officer of the world's largest chipmaker, told students in Taipei that AI is powerful but does not know right from wrong. He also said TSMC is more cautious with AI tools on sensitive R&D data, for fear of leaks.
SoftBank's SB Energy will sell $500 million of stock to Japanese investors before its US IPO
SB Energy will sell up to $500 million of new shares to Japanese investors as part of a US listing that may value it at $50 billion. Nvidia has committed $3 billion, OpenAI holds warrants worth $5.5 billion, and no data center capacity is operating yet.
Samsung is funding Euclyd, an Nvidia (NVDA) challenger that skips the memory Samsung sells
Samsung co-led a $230 million round for Euclyd, a two-year-old Dutch startup building an AI inference chip that replaces HBM with its own memory design. Former ASML chief Peter Wennink is chairman, and nothing ships until 2028.
Palantir (PLTR), Nvidia and Booz Allen are curbing Anthropic model use over data and IP fears
Three large companies have restricted Anthropic's models or told the company they will, over concerns a provider could retain customer data and IP, The Information reported. Anthropic announced a response two weeks earlier that is still rolling out.
Fujitsu (FJTSY) opens MONAKA server sales in November as global CPU rollout follows
Fujitsu opens sales of its MONAKA servers in Japan and Europe in November 2026, with shipments starting April 2027. The standalone processor is planned for cloud and data center operators in the quarter ending March 2027.