How to limit your cloud impact, from bare metal to AI
summary
Claims that cloud moves cut emissions need context: global digital emissions sit near 4% with data centers ~1% and hardware ~75% of that, PUE averages 1.55 versus 1.37 for Scaleway and best 1.16, water use examples show Meta ~0.18 W metric and Microsoft 0.49 and a four‑DC deployment versus 29 Google DCs uses ~800× less water per site. Practical tools discussed include IPMI, RAPL and lm‑sensors, reuse targets aim for 880,000 reconditioned servers by 2025, and a built environmental impact calculator (bare metal first, monthly reports on the bill) appears in public beta.
AI-written summary · about
chapters (19)
- 0:00Introduction
- 1:08Why cloud impact matters
- 2:31Three pillars of digital emissions
- 3:12Data center efficiency
- 5:26Water usage in data centers
- 6:32Cooling systems cost
- 7:15Sustainable data center options
- 8:41Electricity carbon intensity
- 9:25Hardware emissions lifecycle
- 11:34Tools for measuring hardware
- 12:18Extend hardware lifespan
- 13:45Green coding basics
- 14:47Carbon-aware computing
- 16:33Environmental impact calculator
- 18:39AI’s environmental footprint
- 20:09AI energy and water costs
- 21:52Reducing AI’s impact
- 22:35Practical AI emissions steps
- 25:05Key takeaways for sustainable IT
related talks
26mWhy sustainability makes sense for tech
41mWriting Greener Software Even When You Are Stuck On-Prem- 26mWhy sustainability makes sense for tech
12mHow to cut through the noise and look at the physics behind cloud sustainability
32mGreenOps in the cloud
55mToward a carbon-aware Cloud
34mEmbracing Sustainability in Cloud-Native Ecosystems
35mSustainable Platform Engineering: Having a Green Thumb Without Shooting Up