Twitter says it has signed a multiyear deal to use AWS to power Twitter timelines, shifting away from its own infrastructure
Context & Ripple Effects
Twitter built and ran its own serving infrastructure for most of its life, so handing the timeline — its hottest, highest-traffic code path — to AWS is a structural retreat, not an experiment. The move landed at the front of a wave: within weeks Twitter also shifted its offline analytics, data processing, and ML workloads to Google Cloud, splitting its stack across two hyperscalers rather than one.
First-order effects
- AWS gains a marquee consumer-scale workload that validates it can carry real-time serving for a top social platform, while Twitter sheds capacity planning for its core product and pays AWS on usage instead.
Second-order effects
- Google Cloud answered within two months by taking the analytics and ML half of Twitter's estate, and Meta later formalized AWS as its long-term strategic cloud provider — hyperscalers now compete directly for workloads big platforms previously kept in-house.
Third-order effects
- The 2023 episode where Twitter resumed paying Google Cloud after a public standoff over bills shows the endgame of this pattern: platforms lose negotiating leverage once their serving layer runs on rented infrastructure, and multi-vendor splits become the standard hedge against single-cloud dependency.
The trend: Large consumer platforms are dissolving self-operated infrastructure into hyperscale clouds, deliberately splitting workloads across AWS and Google to balance capability against pricing leverage.