Happy Wednesday!

📝 Focus

• Clouds over $Trillion Nvidia?

📊 In the Markets

• US Hawkish Pause ahead? No UK Pause. Stimulating China; Activating Toyota

• AMD's AI Push

📖 MoneyFitt Explains

🎓️  Absolutely Fabless

📝 Focus

Clouds over $Trillion Nvidia?

Chipmaker and AI winner Nvidia rose 3.9% and finally managed to stay above the $1 trillion market capitalisation (share price X number of shares) level into the close. Having briefly crossed the mark during the day on May 30th, it joins a very small band of US giants who've been valued above $1tn: Apple, Amazon, Microsoft, Alphabet, Facebook, now Meta, and Tesla. (The Nvidia price actually halved in 2022 on fears about slower spending on data centres, having previously rallied as a pandemic play. Locked down? Play computer games and punt crypto!)

.....▷ Nvidia is the first semiconductor🎓 company to hit a $1tn valuation as the prime play on Wall Street in the latest frenzy over artificial intelligence sparked, of course, by OpenAI's ChatGPT. In April, Nvidia debuted its powerful new H100 Tensor Core GPU, and last month it launched two new supercomputers, one using its H100 GPUs and the other using its own CPUs and no GPUs at all, showing growing dominance in the high-performance computing market.

.....▷ The company started making niche graphics processing chips mostly for computer gaming enthusiasts, but those GPUs turned out to also have characteristics used to power AI applications, including ChatGPT. The actual business models using AI are still being developed, but as of now, Nvidia's chips are the shovels the AI gold miners need while they figure it out. The big question is whether Nvidia's supply can keep up with demand. It's the only game in town!

.....▷ Or is it? There is competition brewing out there, not simply for the AI chips themselves (like AMD - see below), but with the whole idea of even having chips. The model of dedicated hardware with internal AI chips is challenged by cloud-based AI services where providers like Google, Microsoft (with AMD), and Amazon are designing their own chips while also buying loads of Nvidia's. Meta is as well, while its open-source AI language model LLaMA is also proving popular with developers. (Even Apple and Tesla are designing their own AI chips.) Dedicated AI hardware is designed specifically for neural network deep learning projects, with faster processing and lower latency (the time it takes for a computer to respond to a command), but it can be expensive and needs specialised expertise. Cloud-based AI services allow cost-effective and scalable access to AI capabilities without all that, though possibly with some performance and latency trade-off.

And as a pure design chipmaker, Nvidia is "fabless"🎓 (i.e. doesn't actually have its own factory), so it has to use specialist chipmaking "foundries"🎓 such as TSMC, which Nvidia CEO Jensen Huang called "the best foundry in the world," but noted his company is "not married to any one foundry."

CPU vs GPU - a mini-explainer

- A CPU (Central Processing Unit), the “brain” of a computer, is designed to perform many very complex general-purpose computing tasks using just a few processing cores that can handle a few threads at a time (serial processing.)

- A GPU (Graphics Processing Unit) is designed to do repetitive, parallel processing tasks incredibly quickly with thousands of smaller, more efficient processing cores, like rendering 3-D graphics and video processing (... and crypto mining.)

- GPUs are increasingly being used in IoT (Internet of Things) and AI (Artificial Intelligence). In IoT, GPUs are used for edge computing, which involves processing data locally on IoT devices such as real-time image recognition and natural language processing, and in AI, for super-fast training of deep neural networks.