Agree @immamac. There is too much inherent value in the company. It's selling below book - which is wild. Random, mediocre fabs are priced above 2x book.
Moving material through the fab with external customers will be key for that portion of the business and the recovery. Production of tip of the spear process technology in high volumes/high yields is non-trivial, which is why there are so few players in that space.
The product side has seen reductions in market share due to competition from AMD and the Arm players - and MSFT has been aggressive in helping elevate the arm ecosystem. This year, we have QCOM, but next year, expect MediaTek and NVDA to deliver Arm products into the Windows Ecosystem. The performance is subpar, but the Perf/Watt is better. None of this was a surprise.
All players have been impacted by NVDA's presence in Data Center / HPC. They've been VERY aggressive in finding ways to map just about every high-thread-count or compute application into GPU. They did it with Crypto in ~2015 and that has since migrated to ASIC. Now it's AI inference and training. We are seeing some inference-focused development in the market (Groq, others) that *could* reduce NVDA's presence a bit. Moreover, as model optimization research improves, the overall compute needed to train and infer yesterday's models could go down by orders of magnitude. Does that result in higher fidelity models or the next "killer app?" Who knows? NVDA's main advantage in AI/HPC today: software ecosystem. AMD and Intel are 1/2-1 generation behind there.
The problem being addressed immediately - and the CEO mentioned it in his quarterly report - is that revenues have dropped ~20% since 2021 and staffing has increased by 15%. That's not a good combination and is being addressed by EOY, as per his public statements.