One of Silicon Valley’s most closely watched operator-investors is stepping onto the center stage. The entrepreneur, author and prolific early backer of the breakout companies Vedonest, PagerDuty, Gusto and Planet will join us for a talk about getting going at 2–8 years old, bets that are starting to scale, and what he sees as the next wave of AI-driven innovation.
Why Elad Gil’s insights matter for founders and VCs
Gil has a special cocktail of founder scars and investor pattern recognition. He had stints at Google during early mobile and Maps eras, co-founded Mixer Labs (acquired by Twitter) and then Color, which built the world’s largest public health infrastructure used by government agencies and employers. As an investor, he has backed dozens of category leaders such as Stripe, Airbnb, Coinbase, Instacart, Notion, Figma, Flexport and GitLab.
- Why Elad Gil’s insights matter for founders and VCs
- Early calls in AI that shaped the current market
- What’s happening onstage during his Disrupt 2025 talk
- The market context, in numbers that frame the discussion
- Signals Gil watches when evaluating AI-first startups
- Why this session will draw one of the biggest audiences
His book, “High Growth Handbook,” has been embraced as a playbook of sorts for founders working out how to navigate the messy middle between product-market fit and true scale. It’s his combination of operating depth and portfolio breadth — the rare director whose balance between managing up vs. empowering down did not get totally overthrown by technocratic leg-pulling, political rivalries and backstabbing — that makes him such a reliable compass when markets become noisy.
Early calls in AI that shaped the current market
Long before generative AI was shorthand in boardrooms, Gil was writing early checks into teams like Perplexity, Character.AI and Harvey. These bets entered the AI opportunity from different angles: search redefined to conversational answers (Perplexity), open-ended agents and personalities (Character.AI) and domain-specific copilots for controlled high-stakes tasks (Harvey).
Those bets were more than just lucky shots. They embody a thesis about distribution and defensibility in AI: Combine a potent model with an obvious user wedge, iterate on product velocity at high speed, and construct a moat via data feedback loops and workflow depth. Adoption at A&O Shearman, Character.AI’s enormous consumer engagement and Perplexity’s user growth are evidence of that.
What’s happening onstage during his Disrupt 2025 talk
Expect Gil to move beyond the hype and into mechanics: how to separate signal from a tsunami of AI pitches, what “breakout growth” looks like at consumer versus enterprise companies, and where technical advantage really compounds. Founders will probably get frank opinions on pricing models, inference cost, model choice and routing, and when to build vs. buy infrastructure.
He’s also expected to address why some AI-native products win early — fast iteration cycles, clear value per query or task, and distribution edges — and why others jam when they hit enterprise procurement, security reviews and unit economics. This feels particularly relevant now as teams think through the trade-offs of closed models, open-source stacks and custom fine-tunes.
The market context, in numbers that frame the discussion
Industry data informs the conversation: There has been a massive increase in investment around deep learning, including a 300,000x growth in the number of FLOPS used in training models. There has also been a doubling of training costs every 3.4 months and top benchmark performance on tests like ImageNet increasing at higher rates than before, with more “widely useful” architectures moving faster down the y-axis with scaling and data-mining advantages. McKinsey estimated that generative AI could generate trillions in annual economic value across functions such as customer operations, marketing and software engineering, with software development and sales enablement as some of the highest-return use cases.
On the funding front, AI continues to register as a standout while cooling-off pervaded overall venture activity, according to PitchBook–NVCA reporting. Early-stage AI rounds have seen competitive term sheets, while late-stage financings continue to center on:
- Teams with defined go-to-market traction
- Unique model capabilities
Signals Gil watches when evaluating AI-first startups
And these are the sorts of things Gil likes to pound home that travel across cycles: retention curves and cohort quality, capital efficiency, speed of product iteration, and distribution that compounds. In AI, he’d probably emphasize more signals — inference margin structure, proprietary data advantages, evaluation rigor outside of leaderboard-chasing and whether a product earns a daily default behavior.
He may provide founders with hard-nosed advice for how to navigate enterprise buyers, align roadmap around measurable outcomes and build teams that pair research talent with pragmatic product management. Those investors who are telling Mr. Buffett what to do — shareholders sitting in the seats at the arena — will be hanging on his words as he explains how he decides between durable moats and passing novelty.
Why this session will draw one of the biggest audiences
Disrupt’s main stage offers a mix of founders, investors and makers from businesses like Coinbase and ConsenSys who cover topics such as crypto-asset security. Gil’s track record — a string of early bets on companies that redefined categories, combined with hands-on operating experience — makes his perspective unusually actionable. For those who are building in AI or riding the crest of hyperscale growth, this conversation delivers a clear read on what matters next and how to position for it.
Whether you’re iterating on a seed-stage story or tightening a late-stage go-to-market engine, you can expect real-world takeaways based on real-world results — not vibes.
That’s exactly why this appearance is one of the most eagerly awaited events of the conference.