Quick Facts
- Developers at Developer Week conference questioned AI’s productivity promises amid security and cost concerns
- Memory costs now represent 65-70% of data center server bills, with DRAM pricing set to rise 275-300% through 2027
- Nearly 2,000 exposed AI servers lack proper authentication controls, according to security researchers
Enterprise developers are pushing back on AI adoption promises, citing fundamental issues with security, memory costs and interoperability that challenge the technology’s readiness for production environments.
At the Developer Week conference in San Jose on Thursday and Friday, engineers and executives gathered to assess AI’s impact on enterprise operations. The mood reflected growing skepticism about AI’s practical benefits.
“If AI is supposed to be a revolutionary productivity tool, then why am I still doing most of the work?” asked Tony Loehr, a solutions engineer at Cline Bot Inc., capturing widespread developer frustration.
The memory crisis represents the most immediate challenge. AI servers demand eight times more memory than traditional machines, with memory now accounting for 65% to 70% of a data center server’s bill of materials. Hyperscalers absorb roughly $100 billion annually in incremental memory costs.
DRAM pricing compounds the problem. Prices are projected to rise 275% to 300% from 2025 through 2027, with conventional DRAM approaching $1.20 to $1.30 per gigabit. By 2026, memory alone could reach $550 billion to $570 billion—roughly 55% to 58% of the entire semiconductor market.
Security concerns add another layer of complexity. Security professionals from Red Hat and IANS Research documented authentication failures in AI systems. Nearly 2,000 Model Context Protocol servers exposed on the web lack proper authentication and access controls.
“AI projects don’t fail first on capability; they fail first on trust,” said theCUBE Research principal analyst Scott Hebner. “Until enterprises can verify and defend outcomes, autonomy stays trapped in low-stakes use cases.”
Infrastructure constraints create additional bottlenecks. Cisco President Jeetu Patel identified three major adoption barriers: “The first is infrastructure. There’s just not enough power, compute, and network bandwidth in the world. Infrastructure is oxygen for AI.”
IBM’s Nazrul Islam, chief architect and CTO for AI and integration platforms, pointed to interoperability as the core issue. “The problem is not the model, the problem is not the agent, the problem is the interactions. We’re missing the interoperability and not the intelligence.”
The National Institute of Standards and Technology launched the AI Agent Standards Initiative to address emerging interoperability, identity and security challenges. Issues around trust, authentication and safe integration have become pressing as organizations experiment with agent-based systems.
Cost management concerns extend beyond memory. ThoughtSpot VP Anjali Kumari described an “AI readiness gap” driven by “rigid tools and unpredictable cloud costs.”
Despite the challenges, IBM research found that 42% of enterprise-scale companies with more than 1,000 employees have actively deployed AI. Workers report AI saves 40 to 60 minutes per day when it works properly.
Read more: The AI trust gap: Developers grapple with issues around security, memory, cost and interoperability