CoreWeave CEO Mike Intrator defended his company's aggressive expansion into AI infrastructure at Goldman Sachs' Communacopia + Technology Conference, attributing public skepticism to the pace of technological disruption rather than fundamental concerns about the business model.
"The speed of change is frightening," Intrator told CNBC, framing resistance to AI adoption as a natural human response to rapid innovation rather than justified caution about the sector's trajectory. The comments arrive as CoreWeave, a private GPU cloud infrastructure provider, positions itself as a critical enabler of large language model deployment and training across enterprise and consumer segments.
CoreWeave operates at the intersection of two overlapping trends: explosive demand for compute infrastructure to power AI models and widespread apprehension about AI's societal and economic consequences. The infrastructure play differs materially from consumer-facing AI applications. CoreWeave sells compute capacity, networking, and storage optimized for GPU workloads to model developers, inference providers, and enterprises building internal AI systems. The company functions as a landlord to the landlords in the AI stack.
Intrator's framing matters. By attributing pushback to discomfort with speed rather than to legitimate questions about AI deployment, capital allocation, or technology risk, he sidesteps scrutiny of whether current AI infrastructure spending reflects actual demand or speculative fervor. CoreWeave has raised substantial capital and is privately valued at billions of dollars. Its business thesis hinges on persistent and growing demand for GPU compute capacity. That thesis holds only if enterprises and AI developers continue deploying models at current or accelerating rates.
The semiconductor and AI infrastructure sectors face distinct but related risks. NVIDIA, the dominant supplier of GPUs, commands extraordinary valuations built on the assumption that AI spending will compound for years. If enterprise adoption slows, if models plateau in capability, or if inference efficiency improves faster than expected, demand for new GPU capacity could weaken. CoreWeave's revenue model depends directly on this assumption holding.
The CEO's comments also reflect pressure from multiple directions. Public markets have begun pricing in realistic risk scenarios for AI spending. Some large enterprises have publicly scaled back AI project ambitions or delayed deployments. Meanwhile, generalist skeptics and domain experts have raised concerns about training costs, inference efficiency, environmental impact, and the concentration of compute capability among a small number of providers. None of these concerns disappears because technological change moves quickly.
CoreWeave has attempted to differentiate by offering specialized infrastructure for specific AI workloads, competing on availability and customization rather than on NVIDIA's total dominance of GPU supply. That strategy faces headwinds: NVIDIA controls the chips themselves, and hyperscalers including Amazon Web Services, Google Cloud Platform, and Microsoft Azure integrate GPU capacity directly into their platforms, reducing CoreWeave's addressable market.
Intrator's appeal to the psychological shock of rapid innovation is emotionally resonant but economically incomplete. Investors in CoreWeave, its customers, and competitors in the GPU cloud space should focus on demand elasticity, actual workload growth, and competitive positioning rather than on whether society finds change unsettling.