Anish Acharya, general partner at Andreessen Horowitz investing from the AI apps fund, argues that AI has fundamentally changed startup strategy: founders can now build “narrow startups” — deeply specialized, high-priced products for relatively few customers — because models deliver 100x value leaps (“silver bullets”) that make product quality the primary growth driver, not distribution or marketing spend.
The Narrow Startup Thesis
Narrow startups build incredibly opinionated, deep products, charge very high prices (e.g., $200–300/month), and target a relatively small number of customers — 41,000 customers at $200/month yields a $100M run rate.
Precedent already exists: Google Ultra ($250/mo), Grok ($300/mo), OpenAI ($200/mo), Anthropic ($200/mo) — consumers flock organically and pay willingly because the products over-deliver on expectations.
Specialization is the new moat: with AI collapsing software creation costs, a small team can go so deep for a specific customer that a competitor would need years of roadmap to catch up.
Rich product ecosystems (not just a single feature) are another competitive vector — e.g., meeting recorders need spreadsheets, word processors, notes, diary apps to capture full value, which model labs are unlikely to prioritize.
Being multi-model (using Anthropic, OpenAI, Google together) is a structural advantage over single-model labs; products like AI code editors benefit from routing to the best model per task.
When products over-deliver — sometimes producing results better than the user imagined — they can charge for the compute-intensive reasoning that enables those outcomes; high COGS justifies high prices.
Why Product Wins Over Distribution Now
The “lie founders tell themselves” is that incremental improvements (lead bullets) equal a 100x leap (silver bullet); 100 lead bullets never equal one silver bullet.
Early AI products (ChatGPT, Midjourney) grew entirely organically — no CAC — because the value leap was so large customers pulled the product without incentives.
Willingness to pay blew past assumed ceilings: high COGS (especially video generation) forced real pricing, and customers kept paying even as prices rose.
In this era, there are no marketing problems, only product problems; if you need significant CAC, you haven’t delivered sufficiently on product.
Founders can now “go deep or go home” instead of “go big or go home” — AI models address the entire non-deterministic human experience (emotions, relationships, creativity, self-expression), not just intellectual tasks, and AI code generation lets tiny teams build at massive scale.
Building for Pull, Not TAM
Predicting TAM is a fool’s errand; Acharya’s first startup succeeded by building for a tiny, fast-growing platform (6M iPhones at App Store launch) he personally cared about, not by chasing a big market.
The useful prompt for founders: “What is the $1,000/month SKU of our product?” — what would it need to do, does it do it today, would people pay, have you tested it?
Customers paying dramatic prices is a stronger signal than any framework; conversely, a free product you must pay people to try is a warning sign.
Qualitative signals matter more than metrics: you can’t keep up with demand, customers pull the roadmap violently, you feel the pain points intuitively.
Retention and CAC are downstream of value; if you lose 90% of customers in year one, even “best-in-class” retention won’t save the business — think from first principles.
Founder Traps and Signals
Trap 1: Talking yourself into product-market fit — if you have to convince yourself, you don’t have it.
Trap 2: Hunting for metrics to justify PMF (calibrating “good” retention/CAC) instead of assessing business health from first principles.
Trap 3: The power user trap — a few ecstatic users don’t equal broad market fit unless you either (a) build for power users and capture their value (narrow startup model) or (b) build for mass market honestly.
Real PMF feels like the market pulling the product out of you, often violently; you simply cannot keep up with everything happening.
The Abundance Agenda: Why Now
This is a 3–7 year window, not a 20–50 year one: abundant capital and dramatic consumer interest in AI products converge now.
The best time to start a company in Acharya’s career — by a long shot — because you can build deeper, charge more, reach profitability with fewer customers, and let product pull do the work.
Core advice: be insanely ambitious on product, raise prices, adjust from customer feedback, ignore business frameworks, build for a small number of people, charge a lot, go insanely deep.
