CUDA's Architect Is Nvidia's Real Succession Story — Not the CEO Search

Nvidia's moat isn't silicon — it's a 20-year software ecosystem built by Ian Buck. With Jensen Huang gone, the man who made GPUs universal now guards the company's irreplaceable competitive advantage.

Jensen Huang is out at Nvidia following revelations of extramarital affairs, one allegedly involving a Chinese intelligence operative. The board named CFO Colette Kress as interim CEO. But the succession question that actually matters for Nvidia's $3-trillion-plus market position isn't who sits in the corner office — it's who maintains the software platform that locks in 90% of AI compute.

That person is Ian Buck, 48, vice president of hyperscale and HPC at Nvidia, and the creator of CUDA — Compute Unified Device Architecture. Buck started building CUDA in 2004 with two people and two libraries. Today it supports over 900 libraries and AI models, each corresponding to an industry running on Nvidia hardware. The platform's power comes from millions of developers who have spent two decades building on it, creating switching costs no competitor has overcome. Buck's path was improbable. He studied computer graphics at Princeton — a field Witt's biography describes as carrying a stigma of immaturity. At Stanford, Buck began connecting multiple GPUs to push them beyond graphics rendering. His experimental language Brook caught Nvidia's attention. Alongside hardware architect John Nickolls, who died of cancer in 2011, Buck built the architecture that let developers use standard C to harness GPU parallel computing for any domain. Nvidia lost money on CUDA for over a decade before the AI boom validated the bet. The moat CUDA creates is both technical and sociological. Every CUDA library represents accumulated expertise in a specific domain — weather forecasting, medical imaging, protein modeling. AMD is explicitly trying to close the software gap. A coalition called "Beyond CUDA" held an event at Nvidia's own GTC conference. Startups are raising funds to crack the ecosystem. None have succeeded. Job postings requiring CUDA skills grew from 2025 to 2026, according to workforce analytics firm Lightcast, and demand for human CUDA engineers continues rising even as coding agents proliferate. Buck's value extends beyond code. Colleagues describe him as capable of switching between deeply technical exposition and accessible explanation mid-sentence. Semiconductor legend Lip-Bu Tan, now Intel's CEO, identified Buck as one of only three true partners Huang relied on, alongside GPU engineering VP Jonah Alben and Kress. Multiple industry analysts named Buck as a succession candidate well before the scandal broke. The competitive dynamics are shifting fast. DeepSeek's efficient AI models spooked markets last year — but the Chinese team achieved those results precisely by exploiting CUDA at a deep level. Buck himself called them "one of the best CUDA development teams outside." This is the paradox of Nvidia's position: even competitors who appear to threaten the company often reinforce CUDA's centrality by building on it. The post-Huang risk is talent flight. Nvidia has historically maintained extremely low turnover, but several analysts have expressed concern that employees may now be more likely to leave. Buck's role as a recruiter for the specific breed of developer who thrives in GPU-level optimization becomes critical. The question isn't whether CUDA's technical lead holds — it's whether the human ecosystem around it stays intact when the gravitational center of the company just walked out the door.