$10 million shipping container. That’s Tiny Corp’s Exabox, now open for pre-orders with a $100k deposit to snag your spot.
And here’s the pitch: drop this 20-foot beast on a concrete slab, plug in power, and boom — you’ve got a ready-to-roar AI training rig. No fussing with custom builds or data center drama. They claim it’ll crush Kimi-sized models in 10 weeks at 50% MFU, all while acting like one giant GPU thanks to Tinygrad magic.
Red for AMD fans, green for NVIDIA diehards. Both wired at 400 Gbps, scaling across multiples if your wallet’s deep enough. Specs aren’t locked yet — that’s the rub — but the hook is unbeatable efficiency: top FLOPS/$, GB/$, and GB/s/$ at launch.
“The whole box will be connected at at least 400 Gbps and is capable of training as one unit. At 50% MFU, it can do 3e24 (Kimi sized) training runs in 10 weeks. With tinygrad software, it will function as one big GPU, but it is made up of normal computers and you can also use PyTorch.”
Straight from Tiny Corp. Sounds slick, right? But let’s unpack the market play.
Exabox: Shipping Container Déjà Vu?
Sun Microsystems tried this in 2007 with Project Blackbox — servers in a box, plug-and-play for quick deploys. Flopped hard amid the financial crisis, but the idea stuck for edge computing. Tiny Corp revives it for the AI gold rush, where hyperscalers hoard NVIDIA clusters like dragons.
Difference? Blackbox was generic x86 drudgery. Exabox targets ML workloads, optimized for Tinygrad (their open-source PyTorch rival) and vanilla PyTorch too. Unique insight: this isn’t nostalgia — it’s a hedge against GPU shortages. As NVIDIA’s H100s vanish into OpenAI’s vaults, containerized rigs like this could let mid-tier labs punch above weight, echoing how container ships democratized global trade back in the ’50s.
Ship date? Q2-Q3 2027. Yeah, pre-order now for hardware that doesn’t exist yet. Manufacturing ramps next year.
Does $10M Buy the Best AI Bang?
Crunch the numbers. At claimed 3e24 FLOPS effective (via Tinygrad), that’s exascale territory in a footprint smaller than your average data center pod. Compare to NVIDIA’s DGX SuperPOD: similar power, but you’d pay through the nose for integration, cooling, networking.
Tiny Corp’s edge — if real — is the $/perf ratio. But un-finalized specs scream caution. What GPUs? How many? Power draw? A thick slab and juice, sure, but what’s the bill for 24/7 megawatts? And Tinygrad’s MFU claims — 50% on a cluster that big? PyTorch tops out lower on multi-node setups without heroic engineering.
Market dynamics scream opportunity. AI capex hit $100B+ last year; VCs pour billions into inference startups. Exabox slots in for orgs too small for Azure deals, too cash-flush for piecemeal racks. Bold prediction: if they hit specs, it’ll force NVIDIA to containerize — or watch indie players erode their moat.
But hype alert. $10M+ feels like enterprise sticker shock, especially post-2022 crypto winter when server farms gathered dust. Tiny Corp’s bootstrapped vibe (tinygrad roots) adds cred, yet scaling manufacturing? That’s no laptop fab.
Why Fork Over $100K Today?
Pre-order’s a refundable deposit — hold the line, no FOMO regret. Smart for labs planning 2027 ramps, dumb for speculators. Ties into Tinygrad’s open ethos: run your stack, scale smoothly.
Critique the spin: “Absolute best bang for buck” pre-specs? Classic vaporware whiff. Still, in a world where Grok-1 ate 300k H100s, turnkey exaFLOPS matter. Historical parallel bites back — Sun’s Blackbox presaged cloud, even if they tanked.
Exabox could spark a container AI renaissance, letting universities and startups train frontiers without begging Sam Altman for scraps.
Look, it’s risky. But data says AI hardware’s a squeeze — supply chains choke, prices soar. Tiny Corp bets on software smarts (Tinygrad) to leapfrog hardware parity.
🧬 Related Insights
- Read more: Lima v2.1 Adds macOS Support and AI Safety Guards—But Here’s What Actually Matters
- Read more: Why AI Mandates Fail: The Real Reason Engineers Resist (And How One Leader Got It Right)
Frequently Asked Questions
What is the Tiny Corp Exabox?
A $10M supercomputer in a 20-foot shipping container, optimized for AI training with AMD or NVIDIA GPUs, Tinygrad, and PyTorch support.
When does Exabox ship and how to pre-order?
First units Q2-Q3 2027. Pre-order with $100k refundable deposit at tinygrad.org.
Is Exabox better value than NVIDIA clusters?
Claims top FLOPS/$ efficiency, but specs pending — watch for launch benchmarks.
