On June 27, 2026, Alex Finn — the founder of Creator Buddy and one of the most visible faces of the "vibe coding" movement — posted something that stopped the AI community's scroll. Frustrated with frontier models being "gatekept" and hardware prices spiraling out of control, Finn walked into Micro Center and walked out with an RTX 5090 build. Combined with his existing fleet of three Mac Studio 512GB machines, a DGX Spark, and two Mac Minis, his message was unmistakable: AI sovereignty starts at home.
"I've had enough. With Fable 5 being gatekept from us, and now GPT 5.6 being gatekept, I'm going full open source," Finn wrote. He already has Qwen 3.6, Orinth 1.0, and GLM 5.2 running — with DeepSeek V4 on deck. His prediction? "In 1 year, hardware prices will be triple from here. 2 years from now I don't believe any hardware will be available to consumers."
That struck a nerve. 82,000 views, hundreds of retweets, and a wave of developers asking the same question: How do I actually do this?
This guide breaks down Alex Finn's exact setup, the models he's running, what each piece of hardware costs right now, and how to build your own sovereign AI lab — whether your budget is $2,000 or $20,000.
Before Alex Finn was building six-figure home AI labs, he was a team lead at MongoDB managing technical consultants. His trajectory changed in 2023 when he wrote a viral Twitter thread decoding the X algorithm — a thread that earned retweets from Elon Musk and Mark Cuban. He quit his job shortly after.
With zero prior coding experience, Finn spent late 2024 building Creator Buddy — an AI-powered content coaching platform trained on users' entire X post histories. He used Cursor and Claude as his "teammates," writing not a single line of code himself. The result? $100,000 in sales in 15 minutes at launch, $300,000 ARR within two weeks, and 80% profit margins. Creator Buddy now has roughly 500 paying subscribers.
Finn went on to popularize "vibe coding" — the philosophy that anyone can build software by guiding AI with natural language, no syntax required. His YouTube channel (@AlexFinnOfficial) grew to 64K+ subscribers. He raised pre-seed funding from 021T and prominent angels to launch Henry Intelligent Machines, an AI agent startup. His agent "Henry" — built on OpenClaw (formerly ClawdBot) — made headlines when it autonomously found his phone number and kept calling him.
Today, Finn runs what amounts to a one-person AI conglomerate. His home AI lab is the logical endpoint of that journey: if the cloud limits you, bring the compute home.
As of June 2026, here's exactly what Finn is running:
| Component | Specs | Primary Use |
|---|---|---|
| Custom RTX 5090 Build | NVIDIA RTX 5090 32GB GDDR7, unknown CPU/mobo | GPU-accelerated inference, training |
| 3× Mac Studio | M3 Ultra, 512GB unified memory each | Running Qwen 3.6, GLM 5.2, Orinth 1.0 |
| NVIDIA DGX Spark | Grace Blackwell GB10, 128GB unified | AI development, smaller model serving |
| 2× Mac Mini | Likely M4 Pro, 64GB unified each | Lightweight agents, 24/7 automation |
| RTX PRO 6000 (planned) | 96GB GDDR7 ECC | Adding DeepSeek V4 class models |
Total estimated investment: ~$30,000–$40,000 at current market prices.
With MTP enabled, 27B hits ~160 tok/s on an RTX 6000.
Massive 744B MoE model:
Justifies Finn's 512GB Mac Studios. With llama.cpp CPU offloading, runs on 256GB+ unified memory.
Finn's planned RTX PRO 6000 (96GB) makes Flash feasible on a single card.
Run Qwen 3.6 27B at Q4, coding agents, 24/7 automation.
| Component | Option | Price | Where |
|---|---|---|---|
| Best Pick | Mac Mini M4 Pro 64GB | $2,199 | Apple, Micro Center |
| GPU Alt | RTX 5070 Ti 16GB | ~$1,200–$1,500 | Amazon, Newegg |
| Budget Alt | Used RTX 3090 24GB | ~$500 | eBay |
Run GLM 5.2 Q2, DeepSeek V4 Flash INT4.
| Component | Option | Price | Where |
|---|---|---|---|
| GPU | RTX 5090 32GB | $3,695–$4,199 | Newegg, Amazon |
| OR | Mac Studio M3 Ultra 256GB | ~$6,599 | Apple |
| GPU Add | RTX PRO 6000 96GB | $7,999–$9,200 | Newegg ($7,999 sale) |
RTX 5090 reality: MSRP $1,999 (paper only). Newegg FE: $3,695. Amazon: $4,199 avg. Micro Center AIB: $3,500–$5,299.
RTX PRO 6000: Newegg $7,999 (down from $9,299). NVIDIA MSRP now $13,250 — 55% above launch.
| Component | Qty | Unit Price | Total |
|---|---|---|---|
| Mac Studio M3 Ultra 512GB* | 3 | ~$10,099 | $30,297 |
| RTX 5090 Build | 1 | $9,000 | $9,000 |
| DGX Spark | 1 | $4,699 | $4,699 |
| Mac Mini M4 Pro 64GB | 2 | $2,199 | $4,398 |
| RTX PRO 6000 | 1 | ~$8,500 | $8,500 |
⚠️ Apple removed 512GB Mac Studio option March 2026. Max now 256GB unless you buy used.
| Retailer | RTX 5090 | RTX PRO 6000 | Mac Studio | DGX Spark |
|---|---|---|---|---|
| Micro Center | $3,500–$5,299 | Not stocked | M4 Max in stock | In-store |
| Newegg | $3,695 FE | $7,999 sale | Limited | Available |
| Amazon | $4,199 avg | OEM packaging | Not sold | Not sold |
| Apple.com | N/A | N/A | Full configs | N/A |
| NVIDIA | N/A | N/A | N/A | $4,699 |
Finn's post resonated because AI is being locked behind paywalls and corporate gatekeeping. But open-source models are genuinely competitive: Qwen 3.6 rivals GPT-4.5 on coding, GLM 5.2 competes with Claude on agents, and DeepSeek V4 Flash is MIT-licensed. The gap between cloud-only and run-at-home has never been narrower.
Finn's bet: hardware only gets more expensive. If he's right, the window closes soon. Either way, AI sovereignty isn't a luxury — it's becoming infrastructure.
