Sapphire’s Ranum says AI startups face a proof gap

Sapphire Ventures partner Anders Ranum says growth investors are caught between depressed public software multiples and record private AI valuations, forcing a harder test of which startups can turn AI demand into durable enterprise value. In an interview with Crunchbase News, Ranum said traditional metrics still matter, but buyers and investors are asking for clearer evidence of cash flow, profitability and defensible deployment.

Ranum has been at Sapphire Ventures for nearly 15 years, focusing on B2B enterprise software, security and industrial infrastructure. Before Sapphire, he spent 12 years in product management and strategy at SAP. His recent investments include LangChain, WorkOS and industrial AI company Tractian.

Public software valuations have been pressured by investor concern that AI could weaken incumbent business models, while private AI companies are still drawing high prices. Ranum told Crunchbase News that gross margins, free cash flow and net dollar retention have improved across parts of software, even as the market prices in disruption risk. He said he still watches net revenue retention, but treats it as backward-looking. A stronger signal, in his view, is whether removing a product would disrupt a customer’s operations.

M&A is active, but prices have reset

Ranum pushed back on the idea that large technology M&A has stopped. He cited 2025 software M&A deal value of $334 billion across 678 transactions, up 40% year over year, and said Sapphire saw more than half a dozen portfolio company acquisitions over the past six months. His read is that deals are happening, while valuation expectations are being repriced.

On IPOs, Ranum said 2026 could bring a major wave of tech listings. He pointed to SpaceX having gone public, Anthropic having filed and OpenAI reportedly preparing to file. He said companies below that top tier may wait until 2027 or later for better conditions, which raises the bar for margin discipline and gives secondary transactions a larger role while companies remain private.

AI branding is no longer enough

Ranum framed the current software market as a period in which investors want proof rather than positioning. He said companies need to show free cash flow, a credible route to profitability and evidence that AI is improving their competitive position. A vendor does not get rewarded, in his telling, for saying it has added AI without showing monetization.

That does not mean traditional SaaS is uninvestable, according to Ranum. He said the distinction is less about AI replacing SaaS and more about software systems taking on tasks that humans previously performed inside workflows. Companies that only add AI features to existing processes face a different standard than companies whose products execute parts of the work themselves.

Ranum also said the AI infrastructure stack is both fragmenting and consolidating. Model providers and data platforms are expanding into adjacent tools, but he argued that startups can defend themselves if they become embedded in enterprise workflows. His test is whether the customer’s actual processes run through the product, making replacement costly.

Industrial AI demand is more specific

In industrial AI, Ranum said near-term demand is showing up in defined factory tasks such as packing, picking, inspection and maintenance. He contrasted those use cases with broader humanoid robotics bets, saying the immediate enterprise case is clearer where labor costs, deployment risk and buying cycles are more constrained.

He cited Tractian as Sapphire’s example of that thesis. According to Ranum, unplanned downtime costs the world’s 500 largest companies about 11% of revenue each year. Tractian combines sensors and AI software to detect signs of equipment failure, a hardware-plus-software model Ranum said is necessary because factories are unlikely to replace older machines wholesale.

For industrial startups, Ranum’s argument is that retrofitting existing infrastructure with sensors and learning software is more practical than asking manufacturers to rip out legacy equipment. He also said the high cost of failure in physical environments can strengthen the sales case, because the value of preventing downtime can be quantified before a contract is signed.

This story draws on original reporting from Crunchbase News.