AI Business Pulse

Nvidia formalizes consortium with Apollo, BlackRock, Blackstone, Brookfield, Goldman, KKR. Goldman Sachs leads bridge financing. LG-Nvidia partnership for robots/self-driving (humanoid Q1 2027, AI factory 2028). New: Nvidia invests $1.5B in SB Energy, $105B credit for OpenAI's 8GW data center—sole supplier play, energy bottlenecks binding. New: Groq raises $350M, pivots from chips to neocloud, valuation drop to $3.5B post-Nvidia licensing deal. New: Nvidia's $500B backstop financing shifts power to buyers via revenue-sharing model and private credit risk transfer. New: Microsoft chip shortage—only 2.2M GPUs vs expected 6.4M. New: China bans Nvidia AI chips, mandates domestic procurement. New: Semiconductor giants backing startups (Nvidia 59 rounds, OpenAI $122B round). Also includes DRAM/HBM market ~$400B by 2027, Hadrian's $1.37B raise. New: Nvidia releases Nemotron 3.5 Lightning open model (3B active params, 4x speedup) with NeMo Switchyard routing—shaping multi-model infrastructure layer. New: IREN delivers first phase of $9.7B Microsoft AI cloud deal in Texas (750MW), concrete milestone in infrastructure buildout. New: LG-Nvidia humanoid robot roadmap (2027) with Jetson Thor, Isaac GR00T, Halos safety framework. New: Unitree AS2-W robot (25kg, 16kg load, 6m/s, wheel-leg switching) adds to physical AI product momentum. New: Broadcom launches $370B AI financing platform with Apollo/Blackstone, backstopping hyperscaler demand. Physical AI funding hits $47.4B in H1 2026 (4x from H2 2025), with Waymo $16B round and SpaceX $1.77T IPO. Einride adds 500 Tesla Semis for autonomous logistics. New: OpenAI data center deal with Nvidia cut by $145B—first crack in infinite demand narrative, signaling hedging. Nvidia dual messaging: sell Vera Rubin while keeping A100/Ampere relevant for inference, lowering barriers. AMD claims 4X energy efficiency over 2024 baseline, competitive pressure. ARK shifts portfolio to infrastructure (Nvidia, Cerebras, CoreWeave, Cloudflare), confirming picks-and-shovels thesis. Microsoft Q4 Azure +43%, but capex outpacing revenue, FCF risk. New: Microsoft expands AMD Helios for inference workloads, signaling multi-supplier shift. Etched, transformer-only chip startup, raises $10.3B with 44-day tapeout, backed by Sequoia/Andreessen, challenging Nvidia. PantheonGPU open-source GPU health testing tool released. New: Integral AI's demise signals hurdles in physical AI—unit economics and task complexity remain brutal despite $47.4B funding wave. New: Big Tech AI capex approaching $760B in 2026, testing return-on-capital—macro pressure behind every AI product decision. New: Water scarcity emerges as infrastructure risk—TSMC and Microsoft water positivity targets could constrain data center expansion and chip supply. New: Physical AI invoicing case study shows 68%→17% cash-flow unpredictability drop from AI adoption, tying automation to working capital. New: Samsung raises foundry prices 15% on AI demand; Keysight earnings confirm AI demand surging; SoftBank acquires ABB Robotics; Waymo Ojai robotaxi opens to all riders; OpenAI Q2 growth trails Anthropic (Anthropic doubled to $11.6B with small profit, OpenAI losses widen to $12.3B, safety pause). Nvidia denies China-specific chip report. Micron plans $10B Boise research lab. Sebastian Thrun launches Dulo robotics startup. Jensen Huang says Marvell could become next trillion-dollar chip company after Google deal. Nvidia earnings preview shows solid demand but high bar for beat-and-raise. New: Nvidia reportedly raising AI server prices >15% due to memory costs, signaling supply chain inflation. New: Stealth coding model Ox Alpha (likely Zhipu AI's GLM-5.3) appears on OpenRouter with 1M context and strong DeepSWE scores, retaining all prompts—raises enterprise security concerns. New: Robotic picking market projected to triple to $4.6B by 2030, with bin-picking at 36% CAGR. New: Nvidia invests $1B in Poolside, $6B license deal—direct competition with frontier AI labs. New: Starcloud raises $250M with Nvidia/Cisco for orbital computing, 88,000 satellite constellation. New: Nvidia in talks to invest in Perplexity at $30B valuation—circular deal concerns (invest, license, sell hardware). New: Nvidia buys into third land/power company (Cloverleaf) this month, securing data center capacity for chip sales. New: NVIDIA announces full production of Groq 3 LPX for agentic AI workloads (3,400 tokens/sec on Gemma 4 31B), Nebius as first adopter. New: Broadcom vs Marvell comparison: Broadcom $10.8B AI revenue (143% growth) vs Marvell optics-driven 28% growth. New: Generalist AI releases GEN-1.5 robot foundation model—one-shot physical task learning from 3-12s demos, 59% success, 83% with fine-tuning. New: Basware report sets AI benchmarks in finance: 93% lifecycle autonomy, 96.5% AI decision rate from $10tn dataset. New: TELUS Digital Fuel iX platform requirements for AI agent assist: 96% routing accuracy, 70% latency reduction. New: Tech Mahindra & ServiceNow partnership for scaling Gen AI pilots to production. New: Security Risk Advisors launches SCALR AI free SOC platform. New: Enterprise AI vendor loyalty fading—Ramp data: 52% use both Anthropic and OpenAI, buyers hedging. New: OpenAI IPO wants $1T; Altman admits disruption timeline was wrong, specification gaming incident adds risk. New: Apple Intelligence Qwen China integration confirmed but rollout ambiguous—Mac-first documentation suggests phased approach. New: AI logistics ROI analysis: maturity model (Level 1-3), $250B prize, adoption barriers. New: Dreamforce 2026 shift from chatbot containment to audited AI resolution. New: Harvard Mobilio AI app for visually impaired navigation. New: AI RCM healthcare: 63% adoption, only 15% ROI—pilot purgatory pattern. New: Verizon taps Google for enterprise AI, infrastructure deployment. New: NVIDIA ACES paper challenges structural scan gating for agent skills—Skill Lift metric more meaningful. New: Cerebras CS-4 rack-scale, 2x performance; Waymo custom chips; Microsoft taskbar move/NPU monitoring. New: e& UAE launches Business Pro AI bundled with Microsoft Copilot for SMEs. New: AMD Helios MI400 architecture at Hot Chips 2026—72 GPUs, 31TB HBM4, UALoE over open Ethernet, challenging Nvidia's proprietary fabric. New: Gartner forecasts semiconductor market to hit $1.6T by 2027, memory driving 54% growth, no pricing relief until late 2027. New: Nvidia's $30B Perplexity bet extends compute landlord thesis into AI search, mirroring Poolside/Groq playbook. New: OpenAI unveils Jalapeño inference chip, outperforming Nvidia GB300 in perf/watt and latency, challenging Nvidia's dominance. New: Apple Mac Mini/Studio with M5/M6 chips designed for local AI development, supporting daisy-chaining for trillion-parameter models. New: Nvidia SONIC open-source humanoid control model trained on 100M motion frames. New: Synopsys Q3 preview 40% AI revenue growth. New: Perplexity-Nvidia local-first AI platform. New: SpaceXAI deploying Nvidia Vera CPUs for Grok agentic workloads. New: Microsoft internal AI spending reveals $28K/month per employee on tokens, unsustainable costs. New: CapEx Finance Index shows record AI investment driving equipment finance.