ComputeLetter
A weekly plain-English briefing for small agencies and content studios that translates raw GPU and AI-infrastructure news (chip launches, model releases, price changes) into 'here is what this means for the AI tools you actually pay for, and what to do about it.'
Agency owners and small studio leads who use AI tools in client work and want a non-engineer take on industry shifts
- Monday morning briefing covering the prior week's chip, model, and pricing news
- 'What this means for your stack' recommendations with concrete tool switches to consider
- Per-tool cost trend charts over 6–12 months
- Vendor reliability scorecards based on price-change frequency and API uptime
AMD's MI400 launch is the first credible threat to Nvidia's AI pricing in years; agencies paying $500–5000/mo for AI compute need a translator between the technical news and their actual margins.
Real spend exists (agencies pay $27-408 CAD/mo per user for AI tools), but the addressable audience for a GPU-focused briefing is narrow and already served by free generalist AI newsletters like AI Weekly (44K subs).AI News — Today's Top Stories, Tracked & Ranked | AI Weekly ↗AI Pricing Canada 2026: 7 Tools from $0-$54 CAD/mo Compared ↗
Heavily crowded: Mercatus Weekly (12K subs, 8-min weekly digest with GPU price tables + token pricing), GPTCrunch, We Compare AI, The AI Minute (summarizes 47 newsletters), and 'Your Non-Technical AI Newsletter' on Substack all already occupy adjacent territory; the specific agency/non-engineer angle is a thin slice, not a defensible gap.Weekly GPU Market Report | Mercatus ↗AI Pricing & Model News — Weekly Newsletter | We Compare AI ↗The AI Minute | 47 AI Newsletters, Filtered to One ↗
Direct competitors (GPTCrunch, We Compare AI, AI Minute, Mercatus digest) appear free; Mercatus monetizes via paid compute products rather than the newsletter itself — no clear evidence yet that a paid weekly briefing on GPU news for agencies will convert, and agency owners tend to be price-sensitive.Weekly GPU Market Report | Mercatus ↗Weekly AI Digest — Stay Updated on AI Models | GPTCrunch ↗
AMD MI400 launch is confirmed for 2H 2026, so the 'why now' catalyst is real but ~6-12 months out, after which the specific Nvidia-vs-AMD pricing-arb story fades; structural AI infrastructure churn keeps the topic alive long-term but commoditizes the angle.AMD Positions Instinct MI400 Against NVIDIA Vera Rubin ↗AMD's next-generation AI chips set to power 2026 data center growth ↗
Trivially buildable solo: subscribe to a handful of feeds, run an LLM summarization pass, add agency-flavored commentary; no engineering or capital barrier — matches the source trend's 'LLVM tea leaves' parsing work at tiny scale.