Replit cuts autoscale compute prices 81% to keep app hosting in-house

Replit will cut its autoscale compute price by 81% on August 1st, reducing the cost from $3.20 to $0.60 per million compute units, as the AI software builder aims to keep users on its platform after they generate an app. The price for 1 million requests will drop from $1.20 to $0.40, outbound data transfer will cost $0.05 per GiB (half the current rate), and app-storage charges will halve to $0.015 per GiB each month. Replit said apps running at scale will cost over 50% less, though the $1 monthly base fee for autoscale deployments and database compute and storage prices remain unchanged.

Replit, founded by Amjad Masad https://replit.com/blog/author/amjad-masad and Haya Odeh https://ld.replit.com/about , will cut its autoscale compute price by 81% as the AI software builder tries to keep users on its platform after they generate an app. Replit said in a thread on X https://x.com/Replit/status/2080322911311667300 that the new hosting rates will take effect beginning August 1st. The largest reductions apply to usage-based deployments. According to Replit's detailed pricing update https://docs.replit.com/billing/aug-cloud-billing-updates , autoscale compute will fall from $3.20 to $0.60 per million compute units, while the price for 1 million requests will decline from $1.20 to $0.40. Scheduled compute will receive the same 81% reduction as autoscale compute. Outbound data transfer will cost $0.05 per GiB, half the current rate. Replit is also halving app-storage charges to $0.015 per GiB each month. Basic and advanced storage operations will each become about one-third cheaper. The reductions are broad, though they do not apply evenly across Replit Cloud. The $1 monthly base fee for autoscale deployments will remain in place, and database compute and storage prices will not change. A legacy reserved VM configuration with one vCPU and 4 GiB of RAM will rise from $0.042 to $0.0486 per hour, a roughly 16% increase. Replit is adding larger reserved instances with eight or 16 dedicated vCPUs. Replit said apps running at scale will cost over 50% less and claimed its new rates match or beat industry benchmarks. Its documentation does not identify the competing services or workload assumptions behind that comparison. Actual savings will depend on whether an app's bill is concentrated in compute, requests, storage, database usage or an always-on virtual machine. Replit wants to own the path from prompt to production Masad, a former Facebook engineer and founding engineer at Codecademy, has spent years turning Replit from a browser-based coding environment into a system that writes, publishes and operates software. Odeh, Replit's co-founder responsible for design, helped shape the product around users who may have little conventional programming experience. Hosting is central to that strategy. Replit Agent can generate an application, while Replit Cloud supplies autoscaling infrastructure, authentication, a database, analytics and deployment. Lower infrastructure prices reduce the incentive for a user to export a generated project and configure hosting elsewhere. Replit's publishing documentation https://docs.replit.com/billing/deployment-pricing separates workloads into autoscale, scheduled, static and reserved VM deployments. Autoscale instances start when requests arrive and stop charging for compute when an app is idle. Reserved VMs run continuously for production services that need predictable resources, while scheduled deployments handle recurring jobs. That arrangement lets Replit sell app creation and operation through the same credit system. Publishing charges are deducted from each user's monthly credits before Replit bills for additional usage. Replit's current subscription page https://replit.com/pricing lists $25 in monthly credits for Core subscribers and $100 for Pro subscribers. Cutting unit prices therefore increases how much hosting those existing allowances can cover without changing the headline subscription prices. The pricing change follows a period of rapid expansion funded by outside capital. In March 2026, Masad said Replit had raised $400 million at a $9 billion valuation https://replit.com/blog/replit-raises-400-million-dollars . Replit also said users from 85% of Fortune 500 companies were building with its products and projected $1 billion in run-rate revenue by the end of 2026. Those figures are Replit's own and were not accompanied by audited financial results. Reducing cloud rates gives Replit another way to convert AI experimentation into recurring production usage. A generated prototype creates intermittent Agent revenue. An application that stays deployed can consume compute, requests, storage and database capacity for months. Replit is accepting lower unit prices to make that longer relationship easier to justify. The new rates will arrive with each billing renewal Users will not all switch to the lower prices on August 1st. Replit says each account will move to the new schedule when its first billing cycle beginning on or after August 1st starts. The timing will therefore depend on the account's renewal date. The change will apply automatically to Replit Core, Pro, Teams and Enterprise accounts. Existing deployments will continue running without configuration changes, according to Replit's documentation.