AI pricing needs to come down to increase adoption, says Jefferies' Brent Thill

Brent Thill, speaking from the Amazon conference in Las Vegas, engaged with the CNBC hosts about a recent report concerning Microsoft's AI sales. He posited that the report "misses the point" by focusing on quotas when the larger narrative suggests a booming, albeit nascent, market. Thill observed firsthand at the conference, and through Jefferies’ own research, that "AI demand is accelerating," citing robust numbers from companies like Snowflake and Salesforce. He underlined that Microsoft itself has indicated that demand is outstripping supply, with a 50-plus percent RPO backlog, signifying a strong appetite for AI capabilities within the enterprise.

However, this robust demand is not translating into immediate, unfettered adoption due to a fundamental economic barrier. Thill highlighted that "the software industry needs to go generally to a lower pricing model to get agents live." This insight suggests that initial pricing strategies for AI tools, particularly for emerging AI agents, have been set too high. Companies, in their rush to capitalize on the AI wave, may have overestimated the immediate willingness of enterprises to absorb premium costs for technologies still finding their footing.

Indeed, AI applications for the enterprise are, as Thill aptly put it, "less than a year old." He likened AI agents to "infants in a crib" with "diapers on and a bottle," a vivid metaphor emphasizing their developmental stage. This immaturity implies that while the potential is immense, the immediate, tangible ROI for businesses might not yet justify the steep price tags. Microsoft’s own internal "running joke" about products being "rough" initially, then "better," and finally "nailed" in a one-to-two-year cycle, resonates with this perspective. The widespread rollout of these AI agents, particularly for enterprise, is still a few years away, with Thill being "very bullish" on a 2026-2027 timeframe for the true enterprise AI boom.

The eventual commoditization of AI agent builders is also a looming reality. Thill acknowledged that while "we are not there yet," it is an inevitable outcome, drawing parallels to the internet, cloud, and SaaS booms. The proliferation of agent builders, with "dozens on the showroom floor" at industry conferences, points to an impending competitive landscape that will naturally drive down prices. This dynamic will force companies to differentiate beyond raw AI capabilities, focusing instead on integration, specialized applications, and proprietary data.

Yet, this commoditization will not impact all players equally. The "big platform companies" such as Amazon, Google, and Microsoft possess an inherent, enduring advantage. Thill succinctly stated, "they have the infrastructure. The agents have to have a home and they have to be fed and watered." These hyperscalers provide the foundational cloud infrastructure and extensive ecosystems necessary for AI agents to function, scale, and integrate effectively within complex enterprise environments. Their ability to manage the underlying compute, storage, and networking, coupled with their existing customer relationships and data pools, creates a powerful moat against commoditization at the infrastructure layer.

From an investment perspective, Thill contended that many software companies are currently undervalued if the enterprise AI revolution truly takes hold in 2026-2027. He noted the stark contrast between the semiconductor sector, which has seen significant gains, and the software sector, which has lagged. This disparity suggests that the market may not yet be fully pricing in the long-term revenue influx and acceleration that enterprise AI could bring, presenting a potential opportunity for investors who understand the patient, cyclical nature of technology adoption.