About
I’m Julien Simon.
30+ years in tech as a software engineer, CTO, and Chief Evangelist. 10 years in cloud and AI (AWS, Hugging Face, Arcee AI). Today, I’m an AI Operating Partner at Fortino Capital, a European PE firm, where I evaluate what’s real and what’s vapor, and help portfolio companies reinvent themselves. More info at julien.org and on my 500k-subscriber YouTube channel.
I’ve sat in the rooms where the slides get made. I know what the demos hide. And I’ve spent a lifetime studying world history and computing history — because the structural patterns that govern how technologies succeed, fail, get captured, and get misallocated are older than any of us. Railway manias. Telecommunications bubbles. The minicomputer-to-microcomputer transition. Japan’s semiconductor rise and software absence. The fiber overbuild. The cloud wars. The mechanisms rhyme, the actors change, and the analysts who don’t read history keep calling it unprecedented.
It’s hardly ever different this time.
What this is
The AI industry runs on two currencies: compute and narrative. The narrative is losing its peg. It always does — ask anyone who held fiber stocks in 2000 or believed the “new economy” had abolished business cycles.
Vendors announce “sovereign” cloud offerings that aren’t sovereign under the statute they claim to satisfy. Countries pledge billions for AI strategies that structurally cannot produce what they promise — for reasons that mirror the industrial policy failures of the 1980s and 1990s, if anyone bothered to check. Hyperscalers extend server depreciation schedules and call the accounting improvement “margin expansion.” Earnings calls say one thing. The 10-K says another. Nobody reads the 10-K.
I read the 10-K. And then I read the history.
The AI Realist is a long-form structural investigation — sourced from SEC filings, government statistical surveys, legislative text, court rulings, and regulatory deliberations. Not conference keynotes. Not LinkedIn threads. Not vendor whitepapers repackaged as analysis.
What I investigate
National AI ecosystems — the “Why Can’t It Build an LLM?” series. Countries with world-class engineering talent, massive budgets, and strategic urgency that still can’t produce frontier AI. The question isn’t “how far behind.” The question is what structural mechanism makes the system produce this outcome — and whether the constraint is fixable. Japan mastered semiconductors through the VLSI Project in the 1970s. The playbook that worked for hardware breaks on software, and the reasons it breaks tell you more about AI’s actual requirements than any capability benchmark. Published investigations: India, Japan, South Korea, Singapore.
Cloud and digital sovereignty — what “sovereign” actually means when you trace the legal pathway from statute to enforcement mechanism to data access. The CLOUD Act’s compelled disclosure provision. FISA Section 702. The entity structures that vendors market as protection and that don’t survive the Legal Pathway Trace. Empires have always projected legal authority beyond their borders — the mechanisms change, the dynamic doesn’t. Most “sovereign cloud” offerings are compliance theater. The law is specific about why.
The financial architecture of AI bets — capex sustainability, depreciation manipulation, the gap between announced commitments and actual capital expenditure, revenue attribution problems, and what the balance sheet reveals that the earnings call is designed to obscure. Every infrastructure buildout in history — canals, railroads, telegraph, fiber — has had a phase in which capital deployment outran revenue by years. Some of those bets paid off. Some produced spectacular write-downs. The difference was always visible in the financial structure before it was visible in the stock price. When a hyperscaler tells you margins improved, check whether they changed the depreciation schedule first.
AI infrastructure and tooling — what’s production-grade, what’s a demo, and what architectural choices actually mean for the people writing the checks. The gap between “works in a notebook” and “runs in production” is where most of the money burns.
The geopolitics of cloud and AI — export controls, compute access, chip supply chains, and the strategic competition that determines who builds frontier AI and who rents it. Technology has always been a vector of power projection — from the telegraph cable cuts of 1914 to semiconductor embargoes today. The question is never just who has the best model. It’s who controls the substrate.
And occasionally, whatever the EU Commission decided to inflict on the technology industry this week.
If you want to see what this looks like before you decide: Indians Rule Big Tech. Why Can’t India Build? maps the structural drain. Two Sovereign Clouds, One Legal Wall traces the legal pathway. Anthropic Named Names. The Timing Wasn’t Accidental. follows the geopolitics.
Sourcing
SEC filings, not earnings call summaries. Government statistical surveys, not Glassdoor. Legislative text, not law firm client alerts. Court rulings, not vendor compliance pages. When the best available source is a journalist's report rather than the primary document, the footnote says so. When a claim is the author's calculation, it's labeled. Vendor claims are unverified until independently confirmed. Every piece carries tens of detailed footnotes because the people who use this work — in investment committees, policy assessments, vendor evaluations, regulatory filings — need to trace the chain themselves.
I have opinions. They come from experience, history, engineering, and financial analysis, not ideology. You’ll agree with some. You won’t agree with others. Either way, you’ll get the reasoning and the evidence.
Welcome to the AI Realist.
