Reliable, high-quality AI inference is starting to command a premium.
In the long run, reliability itself will command a premium across all services.
#AI#Inference#AIInfrastructure
KLES plans to buy an optical packaging supplier with cash and shares. For private investors: who may buy the businesses we back, and when can a sale return cash to the fund? The deal is not complete.
3/3 In the AI age, I think knowing yourself matters even more. Our interests and experiences shape what we choose to create. Ask: What am I curious about? What am I good at? What keeps me learning? Then explore, test your ideas and change direction when needed.
1/3 Yesterday, as a teaching assistant in an innovation and entrepreneurship course at Zhejiang University, I spoke with first- and second-year students about knowing yourself and exploring your future. Career planning starts with understanding the person making the choice.
2/3 Is time spent understanding yourself wasted when competition is so strong? I think it helps us choose where to put our effort. Do we enjoy the work itself, or mainly the income and status it brings? Experience helps us find out. It is also OK to discover what does not fit.
Inelastic supply + three demand curves = price becomes the adjustment mechanism. Structural shortage forecasts start 2027.
The bottleneck isn't always the chip. Sometimes it's two layers down.
Everyone's watching the AI chip war — Nvidia licenses, the MATCH Act, ASML calling export controls "a bad idea."
Almost nobody is watching indium. That's a mistake.
Why now: displays (ITO targets) still take ~75–80% of demand, but HJT/perovskite solar and InP substrates for 800G/1.6T AI optical modules are stacking on top.
HydRegen announced an $8m Series A first close on 14 Oct, led by Genoa Ventures. Its catalysts use enzymes made through fermentation. The company says they work with existing factory equipment. Switching costs may matter as much as the catalyst's price.
With so much money flowing into AI, what about opportunities outside the headlines?
According to PitchBook–NVCA, AI companies accounted for 82.7% of US venture capital deal value in the first nine months of 2026.
This is US data. What opportunities are still out there for other
New opportunities are coming into focus in non-ferrous metals, AI, energy storage, medical innovation and space. In 2026, how can individual investors spot trends in a changing market? Which areas are worth watching? I welcome thoughtful discussion.
The 2026 OCP Summit program covered 800V DC, safety, protection and fault control for AI data centers. New systems must work safely with parts from different suppliers. Investors should watch test results, standards and system links.
AI racks need more power. I see 800V DC opportunities beyond SST: rack power boards, GaN devices and controls. Navitas announced a 24 kW board with NVIDIA, designed to convert 800V to 12V in one step. Next: real-load tests, customer approval and scale production.
On Oct. 14, Delta showed a prefabricated SST DC system at Energy Taiwan. Its modular design aims to speed up deployment and fit different loads. SST competition is also about delivering a full system. Investors should watch trials, safety, delivery time and repeat orders.