萝莉视频

Public market software multiples are hovering at decade lows as investors price in the long-term risk of AI disruption. Meanwhile, private market valuations for AI startups continue to hit record highs. Striking a balance between these two conflicting signals is the central challenge for today’s growth equity investors.

To understand how institutional capital is navigating this gap, 萝莉视频 News recently interviewed , a partner at . Ranum has spent nearly 15 years at the firm, where he focuses on B2B enterprise software, security and industrial infrastructure. Prior to joining Sapphire, he spent 12 years as a product management and strategy executive at .

His recent investments include core infrastructure plays such as and , as well as the industrial AI platform .

In this e-mail interview, Ranum breaks down how the definition of net revenue retention is shifting, why he believes 2026 will see a historic run of major tech IPOs, and where real enterprise demand is materializing on the factory floor.

This interview has been edited for clarity and brevity.

萝莉视频 News: You鈥檝e been at Sapphire for 15 years. Right now, public market software multiples are at decade lows as Wall Street worries about AI disruption, while private AI valuations are hitting record highs. As a growth investor caught in the middle, how are you valuing companies today? Are traditional growth metrics like net revenue retention still the gold standard, or has the math completely changed?聽

Ranum: The gap between public and private market signals right now is unlike anything I’ve seen. I think it creates a real opportunity for investors who can make sense of it. Public software multiples have come down hard, while private AI valuations are hitting record highs. Those two things can’t both be right indefinitely, but the fundamentals underneath are holding up. Gross margins, free cash flow, and NDR have actually improved. The market is broadly pricing in disruption risk, but the companies that are genuinely building enterprise value are still being built.

What that means for how I evaluate companies is that I’m spending more time on whether something is genuinely embedded in how enterprises work, not just whether the numbers look good today. NRR still matters. It tells you whether customers are finding real value. But it’s a lagging indicator. What tells me more is whether switching away from a product would meaningfully disrupt operations. If the answer is yes, that’s a more durable signal than any retention metric.

The current regulatory environment has essentially frozen large-scale tech M&A, and the IPO market is sluggish. If the traditional exit pathways are bottlenecked, how does that change the way you underwrite a Series B or C bet? Do companies just have to stay private and build to massive scale longer than they used to?聽

Ranum: I鈥檇 push back a bit on the framing that M&A is frozen. Software M&A activity actually picked up meaningfully in 2025, with deal value rising 40% year over year to $334 billion across 678 transactions. We saw that in our own portfolio with over half a dozen acquisitions in the past six months. What鈥檚 changed is the pricing. The valuations are being reset, but the deals are getting done.

On IPOs, I believe 2026 is shaping up to be a historic year, with having gone public, having filed, and reportedly set to file soon. If they follow through, we’re looking at some of the largest IPOs ever over the next several months. That’s a remarkable moment. Below that tier, though, the picture is more nuanced. Companies that meet today’s higher bar will wait for more favorable conditions, likely into 2027 or beyond. That means you have to build accordingly, focusing on margin alongside revenue, so you have real optionality when the time comes. The secondary market also helps, giving companies and their investors more flexibility as they wait.

You used to love investing in what you called 鈥渂oring software,鈥� or tools that quietly automated mundane enterprise tasks. Today, every software company claims to be an AI company. In 2026, does traditional SaaS even exist as a viable investment category anymore, or is a software startup inherently unbackable if it isn鈥檛 AI-native from day one?

Ranum: I don鈥檛 think the narrative is AI vs. SaaS. Instead, it’s AI plus SaaS. The companies that are struggling aren’t struggling because they’re SaaS businesses. They’re struggling because investors are in a 鈥渟how me鈥� era, and they don’t have clear answers yet.

Show me the free cash flow. Show me the path to profitability. Show me how AI is actually helping you win. You can’t get a stock bump anymore just by claiming you’re integrating AI. The market wants evidence of monetization.

The way I think about it is whether a company is building something that fundamentally changes how work gets done, or just layering AI on top of a workflow that a human is still doing. We used to back systems of record and workflow companies where the human was doing all the work. Now we’re in a position where the system itself can come in and actually do some of those tasks. That’s a different category of value entirely, and it changes what we look for. The bar has moved, but the opportunity is very real for the companies that can clear it.

Your core thesis is that the LLM stack is fracturing into distinct, standalone billion-dollar layers, such as orchestration (LangChain) and identity (WorkOS). But we鈥檙e seeing a massive border war. Big model providers like OpenAI are building their own tools, and data giants like are buying up security tools. How do standalone startups protect their turf when giants encroach from both sides?

Ranum: Both fracturing and consolidation are happening simultaneously, and I think that’s actually the right way to think about it. The moat isn’t about being first in a category. It’s about becoming genuinely embedded in how enterprises work. The companies I’m most excited about are the ones capturing orchestrated workflows in which the enterprise’s actual processes run through the product. That makes them very hard to displace, regardless of what the giants are building around them.

Because of your background at SAP, you know how enterprise buyers think. Right now, CFOs are looking at massive AI pilot bills and demanding to see actual ROI. When a startup is pitching an enterprise on a software governance or security tool, how do they defend that line item to a cynical CFO before the enterprise has even fully figured out its core AI strategy?聽

Ranum: What we consistently hear from buyers is that trust has become what actually separates the market. Security, governance, compliance, and auditability aren’t nice-to-haves anymore. They’re what make an AI deployment defensible when the CFO or the board asks hard questions.

And cost predictability is right alongside that. We’re in an era of greater focus on ROI, and enterprises want to know what this will cost them at scale before they commit. The vendors that can answer that question clearly are winning deals over the ones that can’t.

It feels like Silicon Valley is obsessed with the glamour of humanoid robots right now. Meanwhile, Sapphire鈥檚 big bets in this space, like Tractian, focus on practical, unglamorous industrial AI and predictive maintenance. Are humanoid robots an expensive venture capital distraction right now? Where is the actual, contract-signing enterprise demand on the factory floor today?聽

Ranum: The near-term ROI story is in constrained, high-value industrial settings such as packing, picking, inspection, and maintenance. These environments have clear labor economics, manageable deployment risk, and real buying cycles. That’s where the contracts are getting signed today.

Our portfolio company Tractian is a good example of what that looks like in practice. Unplanned downtime costs the world’s 500 largest companies roughly 11% of their revenue annually, which is a massive, measurable problem.

Tractian addresses it directly by combining sensor hardware with AI that detects early warning signs of equipment failure. The value proposition is concrete before you sign the contract, and the platform gets smarter the longer you use it. That’s the kind of embedded, compounding value we look for.

The humanoid era will come, but the gradient approach beats the all-or-nothing bet for near-term value creation. Start with specific, well-defined tasks where the payoff is obvious and work from there. The market is ready for that today.

Heavy industry and manufacturing are notoriously slow to change. A startup can’t just plug a modern AI API into a 30-year-old machine on a factory floor. For founders trying to build in the industrial tech space, is the winning strategy to build entirely new autonomous hardware, or is the bigger venture opportunity in retrofitting the world’s existing infrastructure with smart software?聽

Ranum: I believe the winning strategy is smart software layered on top of existing infrastructure rather than replacing it. Factories aren’t going to rip out 30-year-old machines because a startup has a better alternative. That’s just not how it works. The opportunity is in making those machines intelligent.

That said, the hardware-plus-software combination really does matter. You can’t get the data without the sensors. But the durable value is in the software layer that keeps learning over time. That’s where I鈥檓 focused.

In pure software, a buggy AI agent might mean a broken spreadsheet or a weird email draft 鈥� annoying, but fixable. In robotics and industrial tech, a mistake means a factory line shutting down or a broken multimillion-dollar asset. From a venture perspective, how much harder is it to scale a robotics startup when the cost of product failure is so high in the physical world?聽

Ranum: I’d actually reframe the question. The cost of failure in physical environments is what makes the value proposition defensible. When the downside of getting it wrong is measurable, the upside of getting it right is equally concrete. You can walk into a sales conversation and show a customer exactly what prevention is worth before they sign anything. That’s a different conversation than selling software, where ROI takes quarters to show up.

From a scaling perspective, the key is discipline about where you deploy first.

Related 萝莉视频 queries:

Illustration: