a16z's Anish Acharya: SaaS Isn't Dead, We're Not in a Bubble, and the Real AI Winners Are Just Getting Started

The “SaaS apocalypse” has become the laziest take in tech. Anish Acharya, General Partner at a16z leading consumer and fintech investing, spends 80 minutes with Harry Stebbings methodically dismantling it, while offering a far more nuanced map of where value is actually being created and destroyed in the AI era.

The Conversation

Acharya brings a rare combination of perspectives: two-time founder (Snowball acquired by Credit Karma, SocialDeck acquired by Google), operator who scaled Credit Karma’s U.S. Card business to over 100 million members, and now a16z GP investing at Series A across consumer and fintech. The conversation covers the SaaS disruption narrative, AI bubble fears, the open vs. closed source debate, a16z’s internal culture, and where the next wave of AI-native companies will emerge.

The “Vibe Code Everything” Fallacy

Acharya’s opening salvo is blunt: the general story that we’re going to vibe code everything is flat wrong.

“You have this innovation bazooka with these models. Why would you point it at rebuilding payroll or ERP or CRM?”

His argument: the whole SaaS market being “oversold software” conflates two very different phenomena. Yes, AI will disrupt some software categories. But the idea that every SaaS product can be replaced by a vibe-coded alternative misunderstands what makes enterprise software sticky in the first place.

The real disruption vector isn’t vibe coding; it’s AI agents breaking the lock-in of legacy providers. When the cost of transitioning from one SaaS provider to another drops dramatically (because AI handles the systems integration), that changes the competitive dynamics far more than building crude replacements from scratch.

Incumbents vs. Startups: The Distribution War

On who actually wins the AI transition, Acharya offers a framework worth studying:

Developer tools look like Cloud, not Uber/Lyft. The developer tool market won’t converge to one winner the way ride-sharing did. Instead, it’ll look like cloud infrastructure: multiple large players coexisting because developers have genuinely different needs and preferences. This means the market can support Cursor, Replit, and others simultaneously without a winner-take-all dynamic.

Incumbents have distribution, startups have innovation speed. But distribution alone isn’t enough. The companies that will win are those that can combine distribution with genuine AI-native product thinking, not just bolting AI features onto existing products.

The switching cost collapse is the real story. When AI can handle the painful systems integration work that previously locked enterprises into their existing vendors, the entire competitive landscape shifts. Companies that won through lock-in rather than product superiority are most vulnerable.

The Chatbox Is Not the Future

One of Acharya’s more counterintuitive claims: the chatbox interface that dominates current AI products is not the long-term winning paradigm.

Browse-based interfaces (where AI surfaces and organizes information proactively) are still preferable for most use cases. The chat paradigm works for exploration and one-off queries, but for repeated workflows, users want structured interfaces that leverage AI under the hood rather than conversational interactions.

Power users are 10x more valuable in the AI era. As AI amplifies individual productivity, the gap between a power user and a casual user widens dramatically. This has profound implications for product design and pricing: optimizing for power users becomes even more critical.

Why We’re Definitively Not in an AI Bubble

Acharya’s bubble argument is data-grounded rather than vibes-based:

The difference between the current AI moment and previous bubbles (dot-com, crypto) is that AI companies are generating real, rapidly growing revenue. The underlying technology is delivering measurable productivity gains, not just speculative potential. When you look at the actual revenue trajectories of AI companies versus their valuations, the ratios are stretched but not disconnected from reality the way they were in 2000 or 2021.

The Market Map: Dozens of Winners in Legal and Support

For sectors like legal and customer support, Acharya predicts not one dominant winner but dozens. His reasoning: these industries are fragmented by nature, with different subsegments requiring specialized domain knowledge. A legal AI tool for patent litigation has almost nothing in common with one for M&A due diligence. This fragmentation creates space for many $100M+ companies rather than a single $10B platform.

Do Margins Even Matter Anymore?

On the hotly debated question of whether AI company margins matter, Acharya’s position is pragmatic: margins matter, but the timeline for when they need to be “good” is longer than traditional SaaS. AI companies burn more on compute but also grow faster. The key metric isn’t gross margin in isolation; it’s the trajectory of margin improvement as the company scales and optimizes its model usage.

Lessons from Marc Andreessen: Being Right Supersedes Process

Acharya shares a revealing insight about a16z’s internal culture: Marc Andreessen’s core belief is that the quality of being right supersedes process.

“Mark talked about this: when you’re small, the VC sort of gives you power. The VC basically takes your brand, which is not big, and they lend you their brand.”

This isn’t just about contrarian thinking; it’s about having the conviction to act on non-consensus views. Andreessen’s approach: do the deep work to form a view, then back it fully. Half-measures and consensus-seeking produce mediocre returns.

The “Triple, Triple, Double, Double” Question

On whether the classic SaaS growth benchmark (T2D3) is dead, Acharya’s nuanced take: the framework itself isn’t dead, but the composition of what drives that growth is changing. AI companies can hit those growth rates, but the revenue composition looks different: more usage-based, less seat-based, with different retention dynamics. The companies hitting those numbers today are doing it with fundamentally different go-to-market motions than SaaS companies did five years ago.

Open vs. Closed Source: A False Binary

Acharya rejects the framing of open vs. closed source as a binary choice. The reality is a spectrum: Meta’s approach to open-sourcing Llama serves a strategic purpose (commoditizing the model layer to increase demand for Meta’s other products), while OpenAI’s closed approach serves a different strategic purpose (capturing more of the value chain). Both can be right simultaneously for their respective business models.

The more interesting question isn’t open vs. closed but whether open-source models will be good enough for the marginal use case. If they are, closed-source providers need to offer dramatically better performance or unique capabilities to justify the premium.

Is Kingmaking Real?

On whether top-tier VCs can make or break companies through sheer distribution power, Acharya’s answer is refreshingly honest: investors can be a catalyst, but cannot take a product that would not otherwise be the winner and anoint them the winner.

He points to Deel’s Alex Bouaziz as the gold standard of founder-VC leverage. Bouaziz knows how to maximally extract value from a16z’s network, calling partners at 7 a.m. on Sundays for introductions, even on deals as small as $1,000. But the reason this works is because Deel already has the best product; a16z’s brand and network amplify an existing advantage rather than creating one from nothing.

a16z’s 100/100 Rule

Perhaps the most revealing cultural insight: a16z operates on what Acharya calls the 100/100 rule.

“We have to see 100% of the deals in our domain and we win 100% of the deals that we go after.”

They don’t believe in luck. Missing a great company because you never saw it is unacceptable. And when they do engage, they expect to win. The enforcement mechanism is equally intense: every two years, a16z conducts 360-degree reviews where they interview every founder a GP works with. If founders say you’re telling the truth, showing up, doing the work, and being responsive, you’re doing a great job regardless of returns. If not, “you’re looking for a job elsewhere.”

The AI-Native Category Thesis

Acharya’s most forward-looking framework: we’re about to enter the era of AI-native categories.

His timeline of the current cycle:

• Late 2022: ChatGPT launches
• 2023: Obviously good ideas get started (legal AI, code AI, creative tools)
• Late 2024: Reasoning models (o1, DeepSeek) start working, making previously non-functional ideas suddenly viable
• 2025: Those early leaders scale
• 2026: New AI-native categories emerge that were inconceivable two years ago

OpenClaw and Moltbook are just the beginning. The operative question for founders now: knowing what we all know about AI capabilities today, what company would you build? The answer shouldn’t be “a better version of existing SaaS”; it should be something that couldn’t have existed before.

Interestingly, Acharya notes the early leaders from 2023-24 have maintained their dominance (Harvey, Gamma, Runway), which surprised him. In the mobile era, the early leaders (Friendster) were displaced by later entrants (Facebook). This time, first-mover advantage has proven more durable, possibly because AI products accumulate data advantages faster.

“People are looking at the point and they say that we’re overhyping it. But they’re not looking at the slope, which is being underhyped.”

His Biggest Investing Mistake

Acharya’s most candid admission: his worst investments came from being too casual about product-market fit, particularly in 2021. The specific trap: a founder had a credible theory that matched Acharya’s own theory for why the product would reach PMF. But having a shared theory isn’t the same as having evidence of traction.

The lesson: it’s easy to overestimate the path from zero to one. Investing with self-deception (“let’s just assume it’s working when it’s not quite working”) is the cardinal sin. If you’re intellectually honest about a belief that something will work, that’s a fine seed bet. But conflating your thesis with the company’s reality is where you get burned.

Some Thoughts

This conversation is worth the time for anyone trying to understand the nuanced reality behind the SaaS disruption narrative.

• The “SaaS is dead” and “AI bubble” narratives are both lazy. The truth is surgical: some categories are being disrupted, others are being strengthened, and a whole new set of categories is being born. Understanding which is which is the entire game.
• The switching cost collapse thesis is underappreciated. When AI eliminates the pain of migrating between vendors, winners shift from “best lock-in” to “best product.” This is a massive wealth transfer waiting to happen.
• Acharya’s “point vs. slope” distinction on Moltbook applies broadly: any given AI demo might be overhyped, but the direction it points to is consistently underhyped.
• a16z’s 360-review system (measuring GPs by founder satisfaction rather than returns) explains a lot about why they’ve maintained deal flow in an era where capital is commoditized. Culture as competitive moat.
• The AI-native category thesis is the most actionable insight: stop trying to build better versions of existing software. Ask what’s newly possible, not what’s newly improvable.