Nvidia CEO Jensen Huang surprised investors with a 'half a trillion' forecast. It'll come up at earnings (www.cnbc.com)

Nvidia CEO Jensen Huang stunned investors in October by saying the company already has roughly $500 billion in orders for 2025–2026, a backlog that he said includes current Blackwell GPUs, next‑generation Rubin GPUs, networking gear and related parts. That disclosure — which will be a focal point when Nvidia reports Q3 earnings on Wednesday — implies materially higher 2026 revenue than Wall Street expected: analysts at Wolfe Research and others estimate data‑center sales could be tens of billions of dollars above prior forecasts. FactSet consensus for Q3 is $54.8B in sales and $1.25 EPS, with guidance watchers looking for a January quarter outlook near $61.9B; Nvidia’s stock has nonetheless traded slightly lower since Huang’s announcement amid investor debate over sustainability. For the AI/ML community the takeaway is structural: Nvidia still dominates AI accelerators (over 90% market share of AI GPUs), and its sales backlog plus large hyperscaler capex suggest continued demand for high‑end training and inference hardware even as cloud providers and labs develop custom ASICs (Google TPUs, Amazon Tranium, OpenAI/Broadcom chips). Strategic deals — up to $10B equity and GPU commitments with OpenAI, $5B with Intel to improve interoperability, and investments in Nokia and startups — further lock in GPU footprint across data centers and telco infrastructure. Earnings commentary will be watched for clarity on 2026 visibility, China export limits (H20 constraints) and how Nvidia expects competition and supply dynamics to evolve.