TickerTain
$SPCX····$NVDA····$MSFT····$GOOGL····$ANTHROPIC·$TSLA····$CRWV····$NBIS····$HIMS····$TSM····$ASML····$FIGURE·$LYFT····$UBER····$SNOW····$META····$ELEVENLABS·$CURSOR·$COGNITION·$OPENAI·$AAPL····$RKLB····$XAI·$AMZN····$AMD····$000660.KS····$INTC····
← newsroom
AI Infrastructure · Three-way NVIDIA-Cloud-Customer partnership enables 9-month 10 PB/day migration
now playing · AI Infrastructure
AI Infrastructuretailwindscore 9/10prudhvi vatala
Zero-code GPU acceleration of Spark workloads delivers 3-3.6x performance gains at petabyte scale
Nvidia Spark RAPIDS enables drop-in GPU acceleration for PySpark workloads (joins, repartitions, shuffles) with 3x+ improvement on join-heavy jobs and 2x on unions, requiring zero code chan…
AI Infrastructuretailwindscore 9/10prudhvi vatala
NVIDIA Spark RAPIDS delivers 76% cost cut at 10 PB/day with zero code changes
Snap migrated 10+ petabyte/day Spark workloads to NVIDIA Spark RAPIDS on Google Cloud, achieving 76% job cost reduction, 62% fewer cores, 80% memory reduction, and eliminating 120 TB of dis…
AI Infrastructuretailwindscore 8/10prudhvi vatala
Snap shares idle inference GPUs for batch workloads via creative capacity sharing
Snap built a platform to borrow idle GPU capacity from online inference serving (idle 1-5 AM Pacific) for batch experimentation pipelines, implementing preemption to prioritize user-facing…
Data Center Infrastructuretailwindscore 8/10prudhvi vatala
Time-zone-based GPU sharing between online inference and batch processing unlocks stranded capacity
Snap exploits diurnal usage patterns (1-5am Pacific idle inference GPUs) to run batch experimentation pipelines, building a custom Kubernetes-based platform with preemption that prioritizes…
Memory & Storagetailwindscore 8/10prudhvi vatala
GPU high-bandwidth memory eliminates 120TB disk spill and cuts memory footprint 80% in Spark pipelines
GPU-native parallelism and high-bandwidth memory address the biggest scaling bottleneck in data pipelines — disk spill from memory pressure — delivering 80% memory reduction and removing 12…
Data Center Infrastructuretailwindscore 7/10prudhvi vatala
GPU capacity sharing platform with preemption maximizes utilization across teams
Snap built a ground-up data platform allowing any team to leverage idle GPU capacity with automatic preemption for priority user-facing workloads, creating a template for enterprise GPU res…
Neoclouds & Cloud Computingtailwindscore 7/10prudhvi vatala
Three-way partnership model (Snap, Nvidia, Google Cloud) accelerates GPU migration from prototype to production in 9 months
Deep technical collaboration between end-user, silicon vendor, and cloud provider — including Nvidia Aether for cross-environment Spark tuning — compresses GPU adoption timelines for comple…
AI Infrastructuretailwindscore 7/10prudhvi vatala
Three-way NVIDIA-Cloud-Customer partnership enables 9-month 10 PB/day migration
Snap credits a three-way partnership between NVIDIA, Google Cloud, and Snap for migrating 10+ petabyte/day production pipelines from prototype to full production in 8-9 months, highlighting…