{"slug": "nvidia-s-673b-forecast-reveals-ai-compute-demand-still", "title": "Nvidia's $673B forecast reveals AI compute demand still", "summary": "Nvidia's $673 billion forecast reveals that AI compute demand remains robust, with inference revenue now growing faster than training capex, according to the company's Q1 earnings call. The supply constraint has shifted to CoWoS-L packaging capacity, with TSMC's Arizona fab not expected to relieve it until 2027, while Samsung's HBM3E yields lag SK hynix by 15-20 percentage points. Nvidia's software moat deepened with CUDA 12.6's FP8 kernel fusion cutting inference latency 2.3x on Llama-3-70B versus Hopper, and networking revenue growing 2.5x year-over-year.", "body_md": "# Nvidia's $673B forecast reveals AI compute demand still\n\nWhat's genuinely new isn't the top-line figure. It's the mix shift Jensen teased on the call. Training clusters still command the headlines, but inference revenue — specifically token-generation workloads on Hopper and Blackwell — is now growing faster than training capex. Enterprise inference, not hyperscaler pre-training, became the marginal dollar driver in Q1. That flips the procurement model: you're no longer selling DGX pods to three buyers; you're selling HGX platforms to every SaaS vendor adding a copilot, every telco running RAN-inference, every automaker deploying vision-language models at the edge.\n\nThe supply constraint moved upstream. CoWoS-L packaging capacity is the new bottleneck, not wafer starts. TSMC's Arizona fab won't meaningfully relieve this until 2027 at earliest. Meanwhile Samsung's 4nm yield on HBM3E stacks still lags SK hynix by 15-20 percentage points, which means Nvidia's allocation leverage over AMD and Intel just widened. If you're planning a cluster build for H2 2025, your BOM is effectively locked to Blackwell Ultra availability — and lead times are already quoting 40 weeks.\n\nSoftware moat deepened quietly. CUDA 12.6's FP8 kernel fusion for Blackwell cuts inference latency 2.3x on Llama-3-70B versus Hopper, but the real lock-in is NIM microservices. Enterprises deploying via NIM don't just get optimized kernels — they get versioned, supported containers with SLA-backed security patches. Try replicating that stack on ROCm today; the engineering cost exceeds the GPU delta for any team under 50 people.\n\nNetworking revenue growing 2.5x YoY tells its own story. Spectrum-X and NVLink Switch aren't accessories — they're the only way to scale past 32K GPUs without melting the fabric. Ethernet with RoCE v2 still drops packets at 400Gbps under all-to-all traffic patterns; InfiniBand doesn't. That's why every 100K-GPU cluster RFP now specifies NVLink domain architecture.\n\nThe bear case: China export controls just erased $12B/quarter in H20 revenue. The bull case: sovereign AI builds in Saudi Arabia, UAE, Japan, and EU are each budgeting $5-15B over three years — and they're buying full-stack, not just silicon.\n\nMy read: $673B is conservative if Blackwell Ultra yields hit 70% by Q4. If they don't, the number holds but the timeline stretches. Either way, the compute demand curve isn't bending.\n\n[Trump's chip tax proposal might actually cripple the AI hardware 1h ago](/en/news/7935/)\n\n[Nvidia is building a massive political machine to protect its AI 1h ago](/en/news/7931/)\n\n[Jensen Huang's take on AGI makes it sound like we've already 7h ago](/en/news/7896/)\n\n[Why we need to stop treating air conditioning as a luxury 8h ago](/en/news/7886/)\n\n[Nvidia's Vera CPU just crushed the AMD EPYC 9655P in kernel 16h ago](/en/news/7850/)\n\n[Nvidia's massive $500B financing move raises serious questions 1d ago](/en/news/7795/)\n\n[Next Trump's chip tax proposal might actually cripple the AI hardware →](/en/news/7935/)\n\n## All Replies （9）\n\nHmm, that's a real bottleneck concern. Without memory chips, even the best GPU is just an expensive paperweight. Have you looked into alternative suppliers like Micron or SK Hynix? Or maybe reconditioned chips from old hardware?", "url": "https://wpnews.pro/news/nvidia-s-673b-forecast-reveals-ai-compute-demand-still", "canonical_source": "https://promptcube3.com/en/news/7938/", "published_at": "2026-08-27 20:54:37+00:00", "updated_at": "2026-08-27 21:19:19.788934+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-infrastructure", "ai-chips", "ai-products"], "entities": ["Nvidia", "Jensen Huang", "TSMC", "Samsung", "SK hynix", "AMD", "Intel", "CUDA 12.6"], "alternates": {"html": "https://wpnews.pro/news/nvidia-s-673b-forecast-reveals-ai-compute-demand-still", "markdown": "https://wpnews.pro/news/nvidia-s-673b-forecast-reveals-ai-compute-demand-still.md", "text": "https://wpnews.pro/news/nvidia-s-673b-forecast-reveals-ai-compute-demand-still.txt", "jsonld": "https://wpnews.pro/news/nvidia-s-673b-forecast-reveals-ai-compute-demand-still.jsonld"}}