Google Drops Cloud Computing Prices By Up To 30 Percent, Launches Preemptible Instances
Frederic Lardinois / TechCrunch :
Context & Ripple Effects
This move extends a discounting arc Google started earlier in 2015 with Cloud Storage Nearline for cold data — a pattern of undercutting on the least-differentiated layers of the stack. Preemptible instances formalize a second lever beyond list-price cuts: selling idle capacity cheap to workloads that tolerate interruption.
The mechanism proved durable. Three years later Google extended the same model to accelerators with Preemptible GPUs at a 50% discount under a 24-hour cap, and its 2017 Standard Tier networking option traded premium fiber routing for cheaper public-internet transit — the same trade price against polish.
First-order effects
- Customers running batch, rendering, or other interruption-tolerant jobs can now buy compute at up to 30% below list or accept preemption in exchange for steep discounts — directly attacking the price floor Amazon and Microsoft charge for equivalent VMs.
Second-order effects
- Rivals are forced into an explicit answer on spot/interruptible pricing rather than leaving the discount tier uncontested, and buyers gain leverage to renegotiate existing commitments by holding a credible walk-away option.
Third-order effects
- If the pattern holds, cloud compute hardens into a two-tier market — firm capacity priced at a premium and interruptible surplus sold near marginal cost — a logic Google later extended from pricing to lock-in itself when it eliminated exit fees for customers leaving for rivals.
The trend: Cloud infrastructure is repricing around capacity utilization — discounts for interruptible workloads, cheap tiers for tolerant traffic, and eventually zero switching costs — turning raw compute into an increasingly commoditized market.