Sapphire Ventures: AI Startups Facing Valuation Crisis

The 'Show Me' Era: A Challenge for AI Startups

In the current context, software multiples in the public market are at historically low levels, a situation that reflects investor concerns about potential disruption from artificial intelligence. At the same time, valuations of AI startups in the private market continue to soar, reaching record highs. This dichotomy creates a major challenge for growth investors, who must navigate these contradictory signals. Anders Ranum, a partner at Sapphire Ventures, is well-positioned to comment on this situation, having spent nearly 15 years focusing on B2B enterprise software, security, and industrial infrastructure. Before joining Sapphire, he gained valuable experience at SAP for 12 years.

Ranum has recently invested in essential infrastructure companies such as LangChain and WorkOS, as well as the industrial AI platform Tractian. In an email interview with Crunchbase News, he shared his thoughts on the evolution of net revenue retention, his forecasts for tech IPOs in 2026, and the areas where enterprise demand is truly manifesting.

Valuing Companies in a Changing Market

Public software multiples have significantly dropped, while private AI valuations are reaching new heights. Ranum sees this gap as an opportunity for investors who can understand this complex dynamic. Fundamentals such as gross margins, free cash flow, and net revenue retention have improved, despite the perceived risk of disruption. To evaluate companies, he focuses on the actual integration of products into business operations, rather than current figures. Net revenue retention remains a key indicator, but he is more interested in the impact of a product change on operations. If a product change significantly disrupts operations, it indicates sustainable value.

Ranum emphasizes that net revenue retention (NRR) is still a crucial indicator for assessing the value a company brings to its customers. However, he considers this metric to be lagging and prefers to focus on evaluating companies' dependence on their products. If a product is so integrated that a change would disrupt operations, that represents a stronger signal of sustainable value than any retention metric.

Impact of Regulatory Changes on Investment Strategies

The current regulatory environment has slowed large-scale tech mergers and acquisitions, and the IPO market is also sluggish. However, Ranum notes that M&A activity in the software sector increased in 2025, with a 40% rise in transaction value, reaching $334 billion across 678 deals. He anticipates that 2026 will be a historic year for IPOs, with companies like SpaceX going public, Anthropic having filed its paperwork, and OpenAI expected to file soon. Companies must prepare to remain private longer, focusing on margins and using the secondary market for more flexibility.

Ranum challenges the notion that M&A activity is frozen, highlighting that transactions continue to materialize, although valuations are being readjusted. He sees 2026 as potentially historic for IPOs, with tech giants ready to go public. However, he acknowledges that for companies below this level, the situation is more complex. They need to focus on building solid margins and wait for more favorable market conditions, likely until 2027 or beyond.

AI and SaaS: A New Investment Dynamic

Ranum does not view AI and SaaS as opposing categories but rather as complementary. Companies must show tangible evidence of AI monetization to attract investors. It is about building solutions that fundamentally transform work, rather than simply adding an AI layer to existing processes. Companies that succeed in meaningfully integrating AI have a real opportunity to stand out. Investors are in an era of "show me," and companies must prove their ability to generate free cash flow and achieve profitability.

Ranum explains that the current market demands concrete evidence of how AI contributes to profitability and free cash flow. Companies can no longer merely claim to integrate AI to see their valuations rise. They must demonstrate how AI fundamentally changes the way work is done, automating tasks that were once managed by humans. This ability to transform operations is what distinguishes successful companies in today's landscape.

The LLM Stack Fracture and Startup Protection

Ranum observes a fracture in the LLM stack into several autonomous multi-billion dollar layers, such as orchestration and identity. Large companies like OpenAI are building their own tools, while giants like Databricks are acquiring security tools. Startups must deeply integrate into business processes to protect themselves from competition from the giants. Protection does not lie in being first in a category, but in the actual integration of products into the functioning of businesses.

Ranum emphasizes that startup protection does not depend on being the first to market but on the ability to deeply integrate into the business processes of client companies. Startups that succeed in capturing orchestrated workflows, where actual processes flow through their products, become very difficult to replace, even when giants like OpenAI develop competing tools.

Trust and Enterprise Demand

Trust has become a key factor for enterprise buyers. Startups must demonstrate the security, governance, and compliance of their AI solutions to convince CFOs. Cost predictability is also crucial for securing contracts. Providers who can clearly meet these requirements win contracts over those who cannot.

Ranum explains that trust is now a determining criterion for companies when choosing AI solutions. Aspects of security, governance, and compliance are no longer options but essential requirements. Companies must also be able to forecast costs at scale to convince CFOs of the viability of their solutions. Providers who can offer this transparency and security have a significant competitive advantage.

Industrial AI and On-the-Ground Demand

While Silicon Valley focuses on humanoid robots, Ranum sees real demand for practical industrial AI, such as that developed by Tractian. Constrained industrial environments offer short-term ROI opportunities, with clear purchasing cycles. Unexpected downtime costs the 500 largest global companies about 11% of their annual revenue, representing a massive and measurable problem. Tractian directly addresses this by combining sensor hardware with AI that detects early signs of equipment failure.

Ranum highlights the example of Tractian, which illustrates how industrial AI can provide concrete and measurable solutions. By combining hardware sensors with artificial intelligence capable of detecting early signs of failure, Tractian enables companies to reduce unexpected downtime, which represents a huge cost for large enterprises. This practical, ROI-focused approach is what attracts contracts today.

Strategies for Innovation in Industry

For startups in industrial technology, the winning strategy is to use smart software to modernize existing infrastructure rather than replace it. The combination of hardware and software is essential for obtaining data and creating sustainable value. Factories are not going to get rid of 30-year-old machines simply because a startup has a better alternative. The opportunity lies in the intelligence of these machines.

Ranum firmly believes that modernizing existing infrastructure with smart software is the way forward for industrial startups. Rather than replacing aging machines, companies can add layers of intelligence that enhance performance and extend the lifespan of equipment. This approach allows them to leverage existing investments while integrating cutting-edge technologies.

The Challenges of Failure in Physical Technology

In robotics and industrial technology, the cost of failure is high. Startups must navigate carefully to scale, minimizing the risks of costly mistakes that could lead to production stoppages or asset failures. In software, a faulty AI agent might mean a broken Excel sheet or a strange email draft—annoying but fixable. In robotics and industrial technology, an error means halting a production line or failing a multi-million dollar asset. From a venture capital perspective, it is more challenging to scale a robotics startup when the cost of product failure is so high in the physical world.

Ranum emphasizes that in the realm of physical technology, the consequences of an error can be far more severe than in the software sector. A failure in a robotic or industrial system can lead to significant financial losses and production interruptions. Therefore, startups must be particularly vigilant in their development and scaling, focusing on reliability and safety to avoid costly failures.