The AI Hub · 2026-07-06 · AI Boutique Team · Infrastructure

The quiet rise of the local LLM.

“Can we run it on our own hardware?” went from a hobbyist question to the most common thing clients ask us in 2026. Here’s when the answer is yes — and what local models are genuinely good at today.

Three things changed at once. The open-weight models got genuinely good — Qwen is now the most-downloaded model family in the world, and MIT-licensed frontier-adjacent weights from DeepSeek and Zhipu ship the day they launch. The hardware caught up — NPU-equipped “AI PCs” and big-memory workstations run a quantised 70B-class model at conversational speed, and the tooling matured from weekend-project (Ollama) to production-grade (vLLM). And the regulators arrived: in banking, telecoms, health and anywhere the EU AI Act reaches, data residency has turned “the model comes to the data” from preference into requirement. Industry surveys now put a majority of enterprise inference on-premises or at the edge — up from around one-eighth in 2023.checked 2026-07-06

What runs on what — the 2026 hardware ladderchecked 2026-07-06
AI laptop / NPU
1–8B — assistants, redaction
Workstation, 1 GPU
8–32B — RAG, classification
Big-memory desktop / Mac
70B quantised — real work
Datacenter GPU server
70B+ full precision, MoE
The maths that matters: a 70B model needs ∼140 GB of memory at full precision, ∼35 GB quantised to INT4 — which is why it now fits on one cardchecked 2026-07-06

The economics: rent first, buy at volume

At low volume the cloud wins, full stop — you can’t amortise a GPU server against a few hundred thousand tokens a day, and API prices keep falling. The crossover comes with sustained volume: as a rule of thumb, a business consistently pushing around ten million tokens a day through a 70B-class model breaks even on its own hardware within 12–18 months, because API costs scale linearly with usage while local costs flatten once the box is paid for. Below that line, buy tokens. Above it — or the moment compliance forbids the data leaving the building — the calculation flips.checked 2026-07-06

Your situation
Verdict
Why
Light or spiky usage
Stay in the cloud
Hardware can’t amortise; APIs keep getting cheaper
∼10M+ tokens/day, steady
Run the numbers
Break-even on owned hardware in 12–18 months
Data can’t leave the building
Go local
Not a cost decision — a compliance one. Local is the answer

What local models are genuinely good at — today

Great fit
Ask-your-documents

RAG over contracts, policies and internal knowledge. The workhorse local use case — private by construction.

Great fit
PII-heavy pipelines

Redaction, classification, extraction on customer records — the data never crosses your firewall.

Great fit
Edge & low latency

Factory floors, field kit, air-gapped sites. Milliseconds instead of a cloud round-trip; works offline.

Good fit
Support triage

First-line drafting and routing at fixed cost — predictable volume is exactly what local pricing rewards.

Good fit
Code assist, private repos

Completion and review where the codebase is the crown jewels. Good, not frontier — know the gap.

Not yet
Frontier reasoning & long agents

Multi-hour autonomous runs and the hardest reasoning still belong to the big hosted models. Route accordingly.

Whether a local model belongs in your stack is — as ever — a measurement question: your volumes, your data rules, your hardware, your numbers. That’s a Reality Check, not a hunch.

Key conclusions
01

Local is a compliance answer first, a cost answer second. If your data can’t leave the building, the decision is already made — everything else is sizing.

02

Don’t buy hardware to feel sovereign. Below ∼10M steady tokens a day, an owned GPU server is an expense pretending to be a strategy. Rent until the volume argues otherwise.

03

Someone has to run the thing. A local model is a small production service: updates, monitoring, capacity, an owner. Budget the human, not just the box.

04

Hybrid is the grown-up architecture. Local for the private, high-volume, predictable work; hosted frontier for the hard 10%. The same routing logic that saves money on cheap APIs applies to your own silicon.

Book an AI Reality Check

Ready to make the AI you've already paid for pay you back?