Kerman Kohli

Depending on when this article releases, semi-conductors and other AI/momentum related names have been knocked 30-40% from their all-time highs. Many are willing to call this a top of semis/AI/memory and cede victory for not participating. It is more likely than not, they are celebrating a little too early.

ARCx launched in 2020 as a token-governed DAO project. In 2021, ARCX was launched as a governance token. By 2022, the project was still evaluating product direction and product-market fit. At that time, ARCx offered USDC redemptions to holders who preferred to exit. For holders who chose to remain, no future direction was provided while the team continued to evaluate the appropriate operating focus.

One observation I’ve noticed in the High Performance Computing (HPC) build out for AI is just how integrated everything is. The chip design, chip packaging, data center build, infrastructure integration etc. Every part of the stack is so well thought out and tied together. Having everything in one place, together and co-located is why the system works better. Just like how machines learn from us, I believe something about the design of machines can teach us about ourselves.

Everyone thinks AI is a bubble and there is an overbuild of infrastructure. “Silicon Control” is an 8,000+ word essay that goes into detail about why that view is wrong and why we have a decade-long journey of being supply constrained. In the essay I go into: the case for infinite compute, how & why demand will scale from here, common misconceptions investors have about the supply chain and the implications of it all. The future is a silicon monopoly of less than 20 companies.

I’ve been thinking of this idea for a while now: the one of a techno-capitalist. In essence, the reasoning goes a little something like this: Financial markets are a reflection of world state, priced in. Not efficiently, but just priced. Increasing, financial markets are becoming dominated by technology and without technology our economy stalls (no growth). Currently, financial markets are run by purely finance people. They do not understand technology nor deem it important to do so.

It was August 5th, 2025 when I published my first article about compute more broadly speaking. That piece was called “We don’t have enough compute”. Link here: At the time, I hadn’t really spent lots of time thinking and studying about semiconductors but it became clear to me that our usage was going to be far higher than we predicted. The street had modelled in 2030 that at most we’ll need 10x the amount of compute usage by 2030. Compute is growing exponentially every year, not linearly.

RouteMesh is now in a spot where we have thousands of nodes in our system. Part of that means we can start to understand the software client landscape of these chains themselves. There’s not a lot of data around what client software runs on what chains... until now. When we add a new node in our system, we basically benchmark that node and extract as much information as we can around it.

It’s now been just under a year since I’ve been investing in semiconductors actively. My first article around “We don’t have enough compute” was basically spot in in August 2025. RAM shortages kicked in around September 2025 and semis have been parabolic since then. I had just started migrating my capital to semiconductors from crypto at the time. Virtually all the names I listed in my Machine Economy article have doubled since November last year.

I’ve been thinking a lot about what comes next in terms of crypto markets and where we go from here. The market generally feels lost and aimless, like a lost cause on its way to greatness. Most of the easy money has gone. Tokens are basically down only, which means yields also start collapsing across the board. Essentially all the ponzinomics have died out which means there’s nothing speculative really left.

As you may have noticed, I’ve been down the “you must own your own hardware” rabbit-hole in my writings. I am please to announce that I have fallen even deeper in this space. It’s my belief that compute is going to become increasingly harder to get your hands on and people will be increasingly compute constrained, especially if they’re reliant on the big labs for their compute.