Unique Research · Insight · September 20, 2024 · Source framing: Base10 Partners (TJ Nahigian & Luci Fonseca) via SaaStr
GenAI is a platform shift — models take text/image/audio/video/code and create new content. Base10 maps incumbents vs startups across platform / infra / apps and poses six founder questions.
Pace & capital
AI patents in 2021 were ~30× the level six years prior. VC: despite a hard raise climate, AI is hot — last year’s H1 alone >$4B into AI; this year already ~5× last year’s full-year AI capital.
Stack map
Platform / model: Google, Meta; OpenAI, Hugging Face, Cohere — build in-house and stitch third-party tools.
Infrastructure: tune, integrate, scale models economically (e.g. LangChain).
Applications: horizontal + vertical — Notion, Gorgias (incumbents); Jasper, Copy.ai, Harvey (startups). App-layer retention is hard: enterprise trial budgets thin; prosumer churn high; copycats muddy brand stories.
Incumbent plays
Google: invented Transformers, open-sourced path, then repositioned as AI company with Bard under OpenAI pressure. Meta: ads + feed/content models + years of monetization muscle. Microsoft: partner/outsource mastery — early OpenAI partnership ($1B then +$10B) with preferential access/rights. Nvidia: GPU demand explosion; recent revenue >+100%; top-five market cap. Earnings-call AI mentions up ~100× in two years. App layer: almost every company has an AI strategy; GitHub Copilot writing >half of some engineers’ code; ServiceNow embeds GenAI for summary / text-to-code. Many incumbents experiment in-house with open tools — know this if you sell to them.
Moats that matter now
Proprietary data, best workflows, deep customization that wins. Early in the AI platform shift, Base10 argues incumbents capture most value via distribution, data, and workflows.
Casetext: 11 years into law-firm software; growth stalled ~$10M ARR; shipped GPT-powered Co-counsel — $0→$9M ARR in nine months; acquired ~$650M cash. Hard for a greenfield startup to clear law-firm legal review.
Notion: first AI feature wave — pricing path toward ~$100M ARR AI; rolled to 30M users incl. growing enterprise; NotionAI sits in daily tools/workflows with private data; GenAI treated as foundational (like electricity), returning to “democratize tool creation.”
Gorgias: Shopify-centric CS ticketing; GenAI automation add-on priced at ~half core; tickets automated 7%→18% with 50% goal; 10k+ merchant distribution + training data.
Business-model shifts
ARPU / ACV / LTV up via AI add-ons (sub or per-transaction). Costs down (medical billing, QA) without matching price cuts → margin/LTV up. Retention rises where data + distribution + fit exist. Cloud/mobile history: Amazon (AWS) and Meta (mobile) captured more value than prior incumbents expected — similar pattern likely for GenAI; startups must be more strategic about positioning.
Six questions
For new GenAI company builders (3):
- Which niche/market/vertical can you absolutely dominate — not nibble — without head-on incumbent war?
- Retention is existential — how do you substantially improve the product for new GenAI expectations and PMF?
- How is GenAI in the core architecture, not a bolt-on? Best GenAI companies barely say “GenAI” — it drives everything. Build a long-term business.
For incumbent founders seeking GenAI value (3):
- Return to original mission — which 1–3 customer problems were previously tech-impossible? How will you concentrate resources?
- Walk every P&L / dashboard metric — how can GenAI improve cost, LTV, conversion, etc.?
- Innovation pace is extreme — how do culture and org keep experiment velocity matched to the platform shift?
— END —