25+
years in technology, from doctoral research to production silicon
Barcelona, ES
The constraints aren't a tax on ambition. They're the design brief.
One planet, rising electricity bills, a talent pipeline nobody is funding, and a handful of companies able to out-spend everyone else. I write about building the most advanced silicon and systems in the world with those designed in from the start — not bolted on at the end.
chip architecture × system co-design × AI infrastructure
~/cesc-guim/portrait.jpg
// Cesc Guim · founder & CEO, Openchip · Barcelona, ES
~/vision · design constraints, not afterthoughts
A human-centric way to build large-scale technology
I believe we can build the most advanced AI systems in the world — and still get this right for people, for the planet, and for the parts of the world without a seat at the table. That means treating five commitments as design constraints, not afterthoughts.
01
Built for, and by, humans
Human values set the requirements and the technology is built to meet them — not the reverse. Safe, secure and private by construction, and judged by the real action taken and the value delivered to a person.
Read the position →
02
Sustainable by design
We have one planet. Data-centre electricity demand is heading toward 950–1,300 TWh by the early 2030s — efficiency has to be an architectural decision, not a caveat.
Read the position →
03
Grows the next generation
Over a million additional semiconductor workers are needed globally by 2030. Progress that doesn't grow the talent underneath it is borrowing against its own future.
Read the position →
04
Doesn't quietly raise the cost of living
US electricity prices rose 6.9% in 2025 — more than double headline inflation. Compute demand that outruns the grid is a cost every household ends up paying.
Read the position →
05
Closes gaps, doesn't widen them
Open standards — RISC-V, OpenTitan, Caliptra, OCP — exist at every layer of the stack precisely so that capability isn't gated behind the few companies who can out-spend everyone else.
Read the position →
—
Aligned to European sentiment
Sovereignty, open standards and honest, sourced argument over assertion — a European way of building the same advanced technology, not a smaller version of it.
Read the position →
~/impact · citable output
Twenty years of silicon, measured
500+
patents filed
65+
peer-reviewed publications
5×
consecutive years recognised as a top corporate inventor, 2019–2023
~/journal · tail -f
Latest from the journal
From FLOPS to Intelligence per Watt
There is no working, adopted way to measure intelligence delivered per watt today. I think that gap — not any single architecture choice — is the one most worth closing first.
What Changed in Twelve Months
Capability, scale, energy, and the fraction of a model that actually does the work — the pace argument, told in hard, sourced numbers before any of it gets opinionated.
Design Space Exploration or Design Space Denial?
For a decade, AI silicon design converged on one comfortable target. Agentic and physical AI are breaking that convergence — and it's worth asking honestly whether we're still exploring the design space, or just getting good at exploring the same corner of it.
All 4 entries →~/method · on method
How this is written
I use AI as part of how I develop this material, and I would rather say so plainly than let anyone guess. The frameworks here — the three imperatives, the four fault lines, the five commitments — were built with AI assistance: gathering and structuring source material, pressure-testing an argument I was unsure of, checking a number against its source, and drafting.
What that assistance does not do is decide what I think. Every piece of this goes through me after the AI pass. I edit it, cut it, correct it and often rewrite it outright, because a draft that is merely coherent is not the same as one that is right — and the machine has no stake in whether a claim survives contact with a fab, a grid operator, or a room full of architects who have built the thing.
The positions on this site are mine. So are the judgements, and so are the mistakes. I mention the method because I write about building AI systems that are honest about what they are, and that argument would be hollow if I were not straightforward about using them myself.
~/speaking · in public
Recent keynotes and panels
15 SEP 2026
TALK
Dinar GrausTIC
GrausTIC · Barcelona
6–7 OCT 2026
SPEAKER
Spanish AI and semiconductor mission to Finland
ICEX–AESEMI · Finland
8 OCT 2026
PANEL
Securing the AI Supply Chain in a Divided World
World Summit AI 2026 · Amsterdam
~/personal · the rest of it
Outside the work
Family first, and then a great deal of running — which is where most of what I understand about long, uncertain projects actually comes from.
My family comes first, and they are the reason the rest of this has any shape at all. Everything below is what I do with what’s left.
I run long distances. Not as a hobby I fit around the work, but as the thing that taught me how to do the work: perseverance when the result is a long way off and nobody is watching, endurance as something you build deliberately over years rather than a quality you either have or don’t, and the specific discipline of starting something you are not certain you can finish.
A marathon is honest in a way few things are. The distance does not care about your plan, your intentions, or how good the first half felt. It only reflects the work you actually did in the months before, and it tells you the truth at exactly the point you would most like to be told something else. That is also a fair description of building silicon.
I think that is why athletics is how I make sense of the rest of it — including the argument on this site. The problems worth taking on are the ones where the outcome is distant, the feedback is slow, and the only way through is to keep turning up.
1:10
half marathon personal best
2:31
marathon personal best