Franck Ohrel

Founder/operator. 5 startups. 3 exits.
AI is not a model race. It is an operating system race.
CEO @HikariBlue · Austin/EU

[Franck Ohrel @FranckOhrel]

@StephaneMallard Excellent

[Franck Ohrel @FranckOhrel]

Argentina does not merely want to attract AI companies.
It wants to create a new category of non-human corporation: companies operated by AI agents or robots.

This is not a marginal issue.
It may be one of the first serious attempts to give legal form to the agentic economy.

Who signs?
Who owns?
Who pays the tax?
Who carries the liability?
Who answers when an autonomous agent causes harm?

Europe still looks at AI as an object to be regulated.
Javier Milei looks at it as an economic actor to be domiciled.
One may disagree with the political or regulatory risk.
But the structural shift must be understood: the next battle in AI will not be fought only over models.
It will be fought over legal status, taxation, liability, auditability, and governance systems.
Capitalism was already transformed once by the invention of the limited liability company.
The question now emerging is even more vertiginous:
What becomes of the company when the operator is no longer human, but software?
@HikariBlue

[Franck Ohrel @FranckOhrel]

The real arbitration Gilles is not
« Is AI going to replace developers? »But
« Who will be able to rebuild his operating model around AI before others? »
The winners will not be those who write the same code faster in the same organizations.
They will be the ones who redesign work systems: governance, architecture, product cycles, quality standards, responsibility, training, value measurement.

[Franck Ohrel @FranckOhrel]

macOS ATX
It would be almost fairer than Tahoe.
Apple continues to name macOS as if everything was still playing in California.
But part of technological America has already moved.
Austin is no longer a satellite.
It has become one of the new centers of gravity of American tech: capital, talents, energy, hardware, software, operations.

[Franck Ohrel @FranckOhrel]

The next AI moat is not model access.
It is operational trust.
Most organizations can now access world-class models.
Far fewer can explain:
why a decision was made,
which system influenced it,
who validated it,
what data shaped it,
and how it can be audited.
The scarce layer is shifting upward.
From intelligence generation to intelligence governance.

[Franck Ohrel @FranckOhrel]

Mais qui choisit encore la France ?
Nous sommes nombreux à l’aimer profondément.

Pas la France des slogans.
Pas la France des impôts empilés.
Pas la France des rues en feu, de la défiance et du renoncement.

La vraie France.

Celle qui a donné au monde une idée de la liberté, du goût, de l’excellence, de la science, de la littérature, de l’industrie et de l’État.

Voir ce pays glisser dans cet état est une tristesse immense.

Parce que la France n’était pas seulement un territoire.
C’était une promesse.

Un pays fabuleux.
Un pays capable de grandeur.
Un pays qui savait encore transformer l’intelligence en civilisation.

Aujourd’hui, la question n’est plus seulement : qui choisit la France ?

La vraie question est :

quelle France reste-t-il encore à choisir ?

[Elon Musk @elonmusk]

True

[Rand @rand_longevity]

you go from aging naturally to nobody having to age in a really short period

[Franck Ohrel @FranckOhrel]

@_h16 CQFD

[Franck Ohrel @FranckOhrel]

Most AI projects will not fail because of the model.
They will fail because the organization was never designed to absorb the change.

Poor governance.
Unclear ownership.
Fragmented workflows.
Misaligned incentives.

The technology is improving faster than companies are adapting.
That gap is becoming the real bottleneck.

[Franck Ohrel @FranckOhrel]

@elonmusk I’m so happy to live in Austin, Texas

[Franck Ohrel @FranckOhrel]

SoftBank's €45B AI infrastructure investment in France contains a deeper lesson.

AI is becoming a location problem.
Energy availability.
Grid connectivity.
Permitting speed.
Industrial policy.
Infrastructure financing.
The countries that solve these constraints fastest will capture disproportionate value.
The AI race is increasingly an infrastructure race.

[Franck Ohrel @FranckOhrel]

@OlivierBabeau Effrayant est le terme approprié

[Franck Ohrel @FranckOhrel]

TSMC openly saying that energy efficiency is starting to matter more than raw compute performance is a major signal.
The AI race is entering a new phase.
Not more intelligence at any cost.
More intelligence per watt.
More intelligence per dollar.
More intelligence per unit of infrastructure.
Energy is no longer a secondary constraint.
It is becoming part of the architecture itself.

[Franck Ohrel @FranckOhrel]

Everyone is looking at the $22,000.

The important number is $2,999.

When productive AI infrastructure becomes affordable to individuals, the economics of entrepreneurship change.

This is not just an AI story.

It is a capital formation story.
@nvidia

[Franck Ohrel @FranckOhrel]

@JeffBezos Courage to you and the whole team. Builders know the path is rarely linear.

[Franck Ohrel @FranckOhrel]

@Jason_A_Scharf Exactly.
Civilizations decline when they lose the ability to build faster than they lose the ability to talk.
The gap between the two is becoming visible again.

[Franck Ohrel @FranckOhrel]

The AI race will not be won by the countries with the best speeches about innovation.
It will be won by the countries able to produce:
- abundant energy,
- massive compute,
- fast permitting,
- industrial-scale capital deployment,
- elite engineering density.
This is why Austin matters.

Thx @Jason_A_Scharf #ATX

[Franck Ohrel @FranckOhrel]

The strategic divergence is becoming visible.
The United States increasingly treats AI as industrial infrastructure.
Europe still largely treats it as a regulatory framework.
Over time, infrastructure compounds faster than regulation.

[Franck Ohrel @FranckOhrel]

AI is no longer a software story.

It is becoming an infrastructure story.

Models need chips.
Chips need data centers.
Data centers need electricity.
Electricity needs permitting, capital, land, cooling and political will.

The AI race will not be won by prompts.

It will be won by operating capacity.