Startups scaling 'lean' after funding

Daniel Nadler, founder and CEO of OpenEvidence, used a panel at NVIDIA’s GTC to predict that future “tech giants” could run with fewer than 100 employees; OpenEvidence was valued at $12 billion after a $250 million round earlier this year. Forbes’ AI 50 reporting shows the median headcount at leading AI startups fell to about 89 employees from roughly 150 the prior year, illustrating the lean staffing trend among high‑value AI firms. Venture and portfolio-level compute pools have already emerged as a practical response: Andreessen Horowitz’s “Oxygen” program gives portfolio companies access to private clusters of NVIDIA H100 GPUs to meet model training and inference needs. NVIDIA reinforced the centralized‑compute trajectory by announcing a strategic expansion with CoreWeave and a $2 billion investment to accelerate more than 5 gigawatts of AI factory capacity through 2030. At GTC, NVIDIA CEO Jensen Huang projected at least $1 trillion in demand for Blackwell, Vera Rubin and related AI chips through 2027, underscoring why startups and VCs are shifting budget from headcount to concentrated GPU infrastructure. CoreWeave’s prior disclosure of a roughly $6.3 billion arrangement tied to NVIDIA inventory and subsequent contract backlogs helped lift the specialized “neocloud” market, which industry coverage says is winning multi‑billion deals by offering GPU‑optimized clusters to AI builders.