Building a Sustainable Future
summary
Claims big carbon numbers — training one AI model emits about 5x an American car’s lifetime CO2, cloud emissions now exceed airlines, and global computer electricity use may rise ~20% by 2030. Benchmarks show server cuts of 60→7 (Wayfair), 360→20 (PayPal), and 5,000→600 (Criteo) while keeping SLAs. Architecture uses hybrid memory architecture (HMA) with index-in-RAM and SSD data, runs on Graviton for ~50% lower cost/carbon, adds JSON and graph (Gremlin/TinkerPop) and vectors (HNSW) in labs, and offers a 60-day free trial.
AI-written summary · about
chapters (14)
- 0:00Sustainable IT and AI impact
- 1:14Electricity use by computers and AI carbon
- 3:40Digital paradox in fast-growing economies
- 5:29Database choice for sustainability and TCO
- 7:14Server reduction with Aerospike at scale
- 9:03PayPal fraud detection with Aerospike graph
- 11:53Aerospike on ARM for efficiency
- 12:55Hybrid memory for fewer servers
- 14:19Aerospike data models: JSON, graph, vectors
- 16:46Vector search demo without GPU
- 18:09Future of AI: continuous learning
- 19:55Aerospike Cloud beta launch
- 20:59Summary: sustainable future with efficient data
- 22:04Q&A: legacy data in predictive analytics
related talks
20mBuilding GenAI That Doesn't Cost the Earth
01h 02mJoint Keynote Address: Scaling AI Sustainably: An Uncharted Territory
46mMaking Software & Data Architectures Sustainable
41mWriting Greener Software Even When You Are Stuck On-Prem
55mCode that Won't Heat the Planet: Engineering Sustainable Software for a Better Future
34mPath to production: An environmentally sustainable approach
17mJavaScript Saves the World
31mThe state of Green Washing