Investors are shifting their evaluation criteria for AI startups, moving away from marketing pitches toward rigorous technical audits. According to Softline, venture capital firms now demand proof of proprietary code and dedicated computing infrastructure, rejecting companies that merely function as wrappers for existing large language models.
The issue of 'agent-washing' has reached a significant scale, with Gartner estimating that $234 billion in corporate SaaS spending is at risk due to vendors mislabeling basic automation scripts as autonomous AI agents. This discrepancy often leads to project cancellations once clients realize the systems lack genuine autonomy.
Venture ecosystems in regions like Kazakhstan and Indonesia have introduced mandatory architectural audits for startups. Companies must now demonstrate ownership of intellectual property and physical infrastructure for model training and inference to secure funding, as investors prioritize long-term operational stability over buzzwords.
The market is entering a phase of consolidation, as the failure rate for AI projects globally reaches approximately 80 percent. In Russia, only about 6 percent of organizations report measurable efficiency from AI implementations. Additionally, investors are increasingly scrutinizing the energy consumption of data centers to ensure long-term viability.
