@MiaMarkTwo@syzito.xyz @statsguy@mas.to NVIDIA has a speedrun of the problem that a lot of companies had. They started making a fairly niche product: a 3D accelerator for gaming. This was something that was a fairly limited (but high margin) market. They were able to take a lot of the graphics workstation market because their products were not quite as good but a lot cheaper than existing players there, but that had always been a tiny market. They were lucky because the costs were coming down and the time that they entered the market was about the time it was possible to create something that was just about cheap enough to go from the gamer market that 3dfx targeted to the wider every-computer-needs-one market. That gave them a load of early growth. The market likes growth. Then they released programmable shaders and the entire GPGPU movement started. That let them sell parallel compute engines that operated on regular data to a load of places that didn't care about graphics. Not the highest volume, but a larger market than 3D workstations ever were and similar margins. That gave them more growth. The market likes growth. When they were starting to reach market saturation, they released cards that were good at cryptocurrency 'mining'. Suddenly there was a path that let people buy NVIDIA GPUs and use them to create money (or, at least, things people were willing to exchange for money). And that gave them more growth, albeit driven by an obvious bubble. The market likes growth. But that bubble was going to burst. Augmented reality looked like a possible replacement but the people in that space were mostly building their own low-power chips (and it failed to take off). They needed a new bubble. Fortunately, someone combined the knowledge that you could represent layers in feed-forward neural networks as matrixes and then run them very quickly on a machine that was good at doing independent floating-point operations over regular data structures. And so you could run large feed-forward neural networks on GPUs. Bubble opportunity! So they hyped the hell out of that new use case. And that led to a load of new orders. And that's growth! The market likes growth. Unfortunately, none of their customers were actually making money from the GPUs used for running these neural networks. Unlike the previous bubble (where people were making money, though mostly from things that looked a lot like scams), this one needed to be fuelled by throwing more capital in. So they started doing all sorts of exciting deals with this kind of structure: NVIDIA invests in a company, marking it down as capital expenditure. The company turns around and promises to spend all of the money on NVIDIA GPUs.NVIDIA reports that money as revenue.The market sees increased revenue and knows that the value of a company is some multiple of its revenue, so the value of NVDA increases by more than the cost of the initial investment.NVIDIA sells enough shares to cover the investment. Apparently this is legal, but it absolutely shouldn't be. But each new investment loop like this generates more customers, which signals growth, and the market loves growth. The systemic problem is that the stock market redirects capital to things that have the potential for growth and this gives a huge incentive for any company to demonstrate growth. A company that has saturated its market and is consistently producing good products is less attractive to investors than one that is in a growing area. If you are in a market that is saturated, you need to either keep entering new markets (which, again, should be triggering antitrust laws because cross subsidies are illegal and about the only way you can successfully do this) or provide some rationale to the stock market of why your saturated market can grow. But I don't think that absolves NVIDIA. They have been doing a lot of deeply unethical things and I hope a future SEC (or European regulators, once they realise Trump's SEC is a waste of space) will determine that they were illegal. And I will absolutely remember the people who were celebrating Huang while he was doing all of this.