Mistral's $830m Debt Raise for Europe Data Centres

Mistral ("Frontier AI. In your hands"), a French AI startup led by ex-DeepMind Arthur Mensch, which has a "Donkey Kong room" in its Paris office, and not to be confused with Mistral Group (UK bus and coach provider), has raised money from a consortium of banks to run a new data center powered by Nvidia chips. A non-exhaustive list of banks involved include HSBC Holdings, Mitsubishi UFJ Financial Group, BNP Paribas ("The bank for a changing world") and Credit Agricole CIB.

Moore Threads IPOs in China

Shanghai-based Moore Threads is a Chinese competitor to Nvidia which has raised the equivalent of more than $1bn.  Investors rushed in betting Beijing's push to end foreign technology reliance is now fully on.

The company was founded five years ago in 2020 by former Nvidia executive Zhang Jianzhong. 

Moore designs GPUs which have proven very useful in AI training. While analysts say China made chips lag behind foreign counterparts - friendly policies and investor keenness for AI related businesses - creates a strong environment for continued innovation.

Money raised will go to fund new R&D.

Nvidia Makes Sales in South Korea

US chip maker Nvidia has struck a deal to supply its chips to the South Korean government, and local behemoths like Samsung, LG and Hyundai. Nvidia CEO Jensen Huang said this will allow Korea to produce "intelligence as a new export".

AI Acceleration Firm Groq Gets $750m New Funding

Groq is the creator of the Groq "LPU" (Language Processing Unit) which aims to accelerate AI inference by having a purpose built hardware accelerator that is optimised for running pre-trained models. 

Despite the hardware product, or rather in conjunction with it and enabled by it, Groq's approach is "software first" designed to give developers direct control and simplify hardware utilization (compare the Mojo approach). The hardware and software are tightly linked to enable higher performance. 

This tight coupling has an impact on the software stack - whereas Groq has a proprietary stack, Nvidia's GPUs are supported across open source products like PyTorch and TensorFlow.

You could say its sweet spot is low latency AI inference.

Groq's latest funding round of $750m was led by Disruptive. Other VCs participated as well. Its early funding rounds were from Social Capital (Series A and B) and TDK Ventures (Series B), the corporate venture capital arm of TDK Corporation.

Commonwealth Fusion Systems Raises $863m

Commonwealth Fusion Systems (aka CFS, founded in 2018) has raised money from investors including Nvidia's NVentures venture capital unit to complete a demo system to lead to construction of its first commercial plant (400MW) in Virginia (1MW is equivalent to 1 million Watts).

All added, Commonwealth has raised around $3bn, which is one-third of all capital raised in the fusion industry. Google has already agreed to buy half the power output from the project, estimated to complete in the early 2030s.

Perplexity Bids to Buy Google Chrome for $34.5bn

The bid comes amid speculation that Google will have to divest Chrome after a court found that Google had an illegal monopoly on online search.  The offer was sent to Google's parent Alphabet on 12 August 2025. This is a Jonah-Whale scenario as the bid is higher than Perplexity's own valuation of around $18bn and would be financed by outside investors. Perplexity said in the event of bid acceptance they would keep most of Chrome's existing team. Perplexity's backers include Jeff Bezos, Nvidia and SoftBank.

Maginative is an AI News Platform

Maginative is an AI news platform bringing the scoop on OpenAI, Google, Anthropic, Microsoft, Meta, NVIDIA and more.

Nvidia Said to Be Ready to Acquire ARM Holdings from Softbank for $40bn (25% over Original Purchase Price)

This would conclude a four-year holding period for Masayoshi Son, the deal hinging on the recent conclusion of a dispute between Arm and the head of its China joint venture, Allen Wu. Softbank payed $32bn for ARM back in 2016 (shortly after the UK voted to leave the EU).  To facilitate the deal, Softbank also reversed a move to strip out the IoT component of ARM. Nvidia recently overtook Intel to become the world's most valuable chipmaker.

Nvidia Q2 2018 Profits Up 40% But Competitors Crank up the Heat with Custom Silicon

Stellar Performance

Nvidia announced Q2 revenues of $3.12 billion a rise of 40% from 2017. Performance gains came from the gaming segment comprising graphics chips bundled with personal computers. Data center business grew 83% with demand from Google and Amazon for running computations in the cloud. Disappointing was the fall in demand for digital currency chips, leading to revenue decline of 54% from last year.

Competition from "Custom Silicon"

Competition may come from a number of areas. Google and Baidu are investing in custom silicon, with Google's hardware projects run by former Motorola president, Rick Osterloh. For machine learning specifically, Google has a Tensor Processing Unit (TPU) to run neural networks at scale. In the automotive client space, Nvidia's autonomous driving chipset, Xavier, is under threat as Tesla aims to adopt its own machine learning silicon.

Traditional Competitors

Advanced Micro Devices is planning to release graphics chips for machine learning based on 7nm technology next year. Intel is also wanting to up its game, hiring AMD chief architect, Raja Koduri, to head up its graphics card business set to deliver product in two years.  Nvidia will likely respond by upping the stakes in research and development. Expenses could run to $870m in Q3.

Keeping it Real

Nvidia's Jensen Huang presents at the SIGGRAPH Graphics Technology Conference (GTC). This builds real-time awareness of the company's latest product innovations and drives fan following.