The AI Loop #37: NVIDIA’s $500B AI Push, Google’s New Gemini, and OpenAI Hits Hyperdrive Welcome to The AI Loop #37. It’s been another huge week in AI, with major investments, faster models, and new technology pushing the industry forward at a pace that still feels hard to believe. The race is accelerating across every layer, from the infrastructure powering the biggest models to the tools making frontier AI faster, cheaper, and more practical to use in the real world. And at Perle, we have some major news of our own with a new Samsung partnership. Let’s get into it! Perle Labs Update Perle Labs 🤝 Samsung Samsung is leveraging Perle Labs’ human-verified data infrastructure to support enterprise-scale AI development. As Samsung expands AI across connected devices, health, manufacturing, networks, and enterprise systems, Perle will support the structured data workflows needed to train, evaluate, and improve those systems. It’s a major milestone for Perle and a powerful validation of what we’re building: trusted data infrastructure for AI operating at real-world scale. 👉 Read the full announcement 👈 Top AI Stories NVIDIA and Wall Street Plan a $500B AI Buildout NVIDIA is joining forces with six financial giants, including BlackRock, Goldman Sachs, and KKR, to unlock more than $500 billion for AI infrastructure. The plan would make NVIDIA-powered AI factories investable like other major infrastructure assets, with outside investors supplying most of the capital. Half a trillion dollars is a staggering number, even by AI standards. The race for compute is becoming one of the largest infrastructure buildouts in history. Sources: NVIDIA | Reuters Google Drops Another Major Gemini Model Google launched Gemini 3.7 Flash, its newest model built for coding, tool use, and long-running AI agents. Google says it brings stronger planning, better instruction-following, and fewer retries, all while remaining fast and affordable enough for high-volume workloads. The wild part? It arrived just three weeks after Gemini 3.6 Flash. The model race is now moving so quickly that even major releases barely get a moment to breathe. Sources: Google DeepMind | Reuters OpenAI Pushes GPT-5.6 Sol to 750 Tokens per Second OpenAI just put a frontier model into overdrive. Its new Ultrafast API tier can run GPT-5.6 Sol up to 14 times faster, generating as many as 750 output tokens per second with infrastructure from Cerebras. That kind of speed could transform voice AI, coding, financial research, customer support, and security. Frontier intelligence is getting much closer to operating in real time. Source: OpenAI That’s a Wrap AI is getting faster, the infrastructure behind it is getting bigger, and the gap between research and real-world deployment keeps shrinking. For Perle, our work with Samsung marks an important step forward as we continue building the trusted data infrastructure this next era of AI will depend on. That’s all for now! Until next week 👋

The AI Loop #37: NVIDIA’s $500B AI Push, Google’s New Gemini, and OpenAI Hits Hyperdrive Welcome to The AI Loop #37. It’s been another huge week in AI, with major investments, faster models, and new technology pushing the industry forward at a pace that still feels hard to believe. The race is accelerating across every layer, from the infrastructure powering the biggest models to the tools making frontier AI faster, cheaper, and more practical to use in the real world. And at Perle, we have some major news of our own with a new Samsung partnership. Let’s get into it! Perle Labs Update Perle Labs 🤝 Samsung Samsung is leveraging Perle Labs’ human-verified data infrastructure to support enterprise-scale AI development. As Samsung expands AI across connected devices, health, manufacturing, networks, and enterprise systems, Perle will support the structured data workflows needed to train, evaluate, and improve those systems. It’s a major milestone for Perle and a powerful validation of what we’re building: trusted data infrastructure for AI operating at real-world scale. 👉 Read the full announcement 👈 Top AI Stories NVIDIA and Wall Street Plan a $500B AI Buildout NVIDIA is joining forces with six financial giants, including BlackRock, Goldman Sachs, and KKR, to unlock more than $500 billion for AI infrastructure. The plan would make NVIDIA-powered AI factories investable like other major infrastructure assets, with outside investors supplying most of the capital. Half a trillion dollars is a staggering number, even by AI standards. The race for compute is becoming one of the largest infrastructure buildouts in history. Sources: NVIDIA | Reuters Google Drops Another Major Gemini Model Google launched Gemini 3.7 Flash, its newest model built for coding, tool use, and long-running AI agents. Google says it brings stronger planning, better instruction-following, and fewer retries, all while remaining fast and affordable enough for high-volume workloads. The wild part? It arrived just three weeks after Gemini 3.6 Flash. The model race is now moving so quickly that even major releases barely get a moment to breathe. Sources: Google DeepMind | Reuters OpenAI Pushes GPT-5.6 Sol to 750 Tokens per Second OpenAI just put a frontier model into overdrive. Its new Ultrafast API tier can run GPT-5.6 Sol up to 14 times faster, generating as many as 750 output tokens per second with infrastructure from Cerebras. That kind of speed could transform voice AI, coding, financial research, customer support, and security. Frontier intelligence is getting much closer to operating in real time. Source: OpenAI That’s a Wrap AI is getting faster, the infrastructure behind it is getting bigger, and the gap between research and real-world deployment keeps shrinking. For Perle, our work with Samsung marks an important step forward as we continue building the trusted data infrastructure this next era of AI will depend on. That’s all for now! Until next week 👋

Source:https://x.com/PerleLabs/status/2088350786912969118