The Real Change
AI has lowered the marginal cost of building software. This is real and significant. A function that previously required ten engineers for six months can now be prototyped by two engineers in three weeks. The compression extends to product design, testing, documentation, and deployment infrastructure. The cost of building something that works has fallen dramatically.
What it has not changed is the cost of building a durable company. Enterprise sales cycles haven't compressed. Regulatory compliance hasn't been automated away. The human trust required to become embedded in critical workflows still takes years to earn. The difference between a prototype and a production system that a Fortune 500 will bet its operations on is still enormous.
The Widening Gap
The opportunity in the AI era is not that building is cheap. It's that the gap between “cheap to experiment” and “expensive to differentiate” has widened — and that gap is where venture returns live. As more companies can build more quickly, competitive density in any application category increases rapidly. The advantage that early AI-native startups derived from technical capability compresses as those capabilities become accessible to everyone.
What doesn't compress: years of domain-specific training data. Compliance certifications that took 18 months to earn. Workflow integrations that required hundreds of hours of customer implementation. Trust relationships with enterprise buyers who have already decided to run their operations through your system.
The Core Insight
The AI era didn't change the fundamentals of venture. It amplified them. Durable companies are more durable. Fragile companies are more fragile. Know which one you're building or backing.
What This Means for Investment Decisions
For investors, the AI era creates a specific pattern to look for: companies that are using cheap experimentation to move fast to a defensible position, rather than companies that are mistaking cheap experimentation for defensibility itself.
The test: if a well-funded competitor got API access to the same models today, how long would it take them to replicate your core product's performance for your best customer? If the answer is months, you have a product, not a business. If the answer is years — because you have proprietary data, embedded workflows, compliance infrastructure, and earned trust — you have something worth backing.
The J-Curve in the AI Era
AI hasn't materially shortened fund lifecycles or compressed the J-curve. The DPI (distributions to paid-in capital) timeline — the actual cash return to LPs — still typically requires 8-12 years from first investment to final distribution. Companies take time to scale, find product-market fit, build moats, and reach exits. AI tools help them build faster, but enterprise adoption, regulatory approval, and trust-building remain time-constrained.
Fund managers who expect AI to compress their fund timelines are likely to be disappointed. Those who use AI to identify better companies faster, diligence more efficiently, and support their portfolio more effectively will have structural advantages — but the underlying business reality of company-building timelines hasn't fundamentally changed.