Comprehensive guide to every Google Cloud free tier service in 2026. 20+ free products, $300 trial credit, and hidden costs explained. BigQuery 1 TiB, Cloud Run 2M req/mo, e2-micro VM, and more.

GCP Free Tier Complete Guide 2026

Published 2026-03-27 · Reviewed 2026-10-31 · Figures compiled 2026-03-27, last checked 2026-10-31 · 94 of 143 figures in the tables below come from our records for 1,498 developer tools · 28 Google and Firebase entries tracked

Always Free

$300

Trial Credit

1 TiB

BigQuery Free

Cloud Run Req/Mo

Three ways to start free. Google offers the Free Tier, the $300 Free Trial and product-specific free offers such as the Gemini API free tier in Google AI Studio. Google's Free Tier table lists 28 products, including Cloud Run (2 million requests a month) and BigQuery (1 TiB of querying a month). It has no end date, but Google can change or remove its limits with 30 days' notice. Most limits are monthly; App Engine's limits and Firestore's operation limits are daily. The $300 Free Trial runs for 90 days and needs a credit card or other payment method; the trial billing account is not billed. Only new customers get it: Google gives the Free Trial to accounts that have never been paying users of Google Cloud, Google Maps Platform or Firebase and have not signed up for the Free Trial before. An app built in Google AI Studio can be deployed to Google Cloud (Cloud Run, with Firestore, Firebase Authentication and Cloud SQL for PostgreSQL) using the Starter Tier without a credit card or billing account. The BigQuery sandbox also needs no credit card or billing account and includes 1 TiB of queries a month; its tables expire after 60 days. Accounts opened after 2026-03-02 cannot spend the credit on the Gemini API. The Gemini API free tier has per-minute and per-day limits; the Cloud AI APIs in the Free Tier have monthly limits.

The free VM: Always Free includes one e2-micro VM a month in us-west1, us-central1 or us-east1, with 30 GB of standard persistent disk.

The hidden costs: the e2-micro VM is free only in us-west1, us-central1 and us-east1. Outbound data transfer beyond each product's free allowance is billed. Firestore's free operations are counted per day. Load balancers are billed (forwarding rules and data processed).

Comparing clouds? See our Cloud Free Tier Comparison for a side-by-side matrix of AWS vs GCP vs Azure vs DigitalOcean.

Always Free Products

These products remain free indefinitely within their monthly (or daily) quotas. No expiration, no trial period.

Service	Free Limits	Category
Compute Engine (e2-micro)	1 e2-micro VM/month (us-west1, us-central1, us-east1), 30 GB standard persistent disk Source ↗ Google Cloud's Free Tier requires a billing account. (From docs.cloud.google.com/free/docs/free-cloud-features, read 2026-10-07.) A new Google Cloud user gets the billing account by signing up for the Free Trial, which requires a credit card or other payment method. Google uses it to verify identity and reduce fraud. (From docs.cloud.google.com/free/docs/free-cloud-features, read 2026-10-23.) Usage beyond the Free Tier's monthly limits is billed at standard rates. (From docs.cloud.google.com/free/docs/free-cloud-features, read 2026-10-07.) The e2-micro instance limit is by time, not by instance: all e2-micro instances together are free for the number of hours in the month. (From docs.cloud.google.com/free/docs/free-cloud-features, read 2026-10-04.) An external IPv4 address in use on a VM, including the free e2-micro, costs $0.005 an hour. The free tier covers one hour a month. A VM running all month pays about $3.65 for it. External IPv6 addresses are free. (From cloud.google.com/vpc/network-pricing, read 2026-10-11.) The free tier covers 30 GB-months of standard persistent disk. A disk created in the Google Cloud console defaults to balanced persistent disk, which the free tier does not cover. Choose Standard persistent disk, or create the VM with the gcloud CLI, which defaults to standard. (From docs.cloud.google.com/compute/docs/disks/persistent-disks, read 2026-10-11.) If you do not upgrade to a Paid billing account by the end of the 90-day trial or after spending the $300 credit, your Free Trial billing account closes and all its resources stop, including a free e2-micro VM. They are permanently deleted after a 30-day grace period without an upgrade. After upgrading, the Free Tier still applies. (From docs.cloud.google.com/free/docs/free-cloud-features, read 2026-11-02.) The Free Tier has no end date. Google may change or eliminate its usage limits with 30 days' advance notice. (From docs.cloud.google.com/free/docs/free-cloud-features, read 2026-11-02.) The free 1 GB a month covers traffic from a VM in us-west1, us-central1 or us-east1 to visitors anywhere except China and Australia, including Europe, the rest of Asia, South America and Africa. (From docs.cloud.google.com/free/docs/free-cloud-features, read 2026-11-05.) An e2-micro VM has 2 vCPUs that together sustain 25% of CPU time. It has 1 GB of memory and can burst to 100% for short periods. (From docs.cloud.google.com/compute/docs/general-purpose-machines, read 2026-11-05.) While a billing account is a Free Trial account, you cannot add GPUs, use Google Cloud Marketplace, request a quota increase or create Windows Server VMs. Upgrading to a Paid billing account removes these restrictions. (From docs.cloud.google.com/free/docs/free-cloud-features, read 2026-11-05.) After upgrading to a Paid billing account, the remaining $300 credit can still be used within the original 90 days from sign-up. (From docs.cloud.google.com/free/docs/free-cloud-features, read 2026-11-05.)	Compute
Cloud Run	2M requests/month, 360K GiB-seconds memory, 180K vCPU-seconds Source ↗	Compute
Cloud Run functions (1st gen)	2M invocations, 400K GB-seconds, 200K GHz-seconds, 5 GB egress/month. Current Cloud Run functions are billed on Cloud Run pricing. Source ↗	Compute
App Engine	28 instance-hours/day (F1 instances), 1 GB egress/day Source ↗	Compute
GKE	$74.40/month credit per billing account. Covers the cluster management fee for one Autopilot or zonal Standard cluster; not compute, networking or other resources. Source ↗	Containers
Cloud Shell	5 GB persistent home directory, web-based terminal with built-in tools Source ↗	Dev Tools
BigQuery	1 TiB queries/month, 10 GiB storage Source ↗	Analytics
Firestore	1 GiB storage, 50K reads/day, 20K writes/day, 20K deletes/day Source ↗	Database
Cloud Storage	5 GB-months Standard (us-west1, us-central1, us-east1), 5K Class A ops, 50K Class B ops, 100 GB egress a month to anywhere except China and Australia Source ↗	Storage
Pub/Sub	10 GiB messages/month Source ↗	Messaging
Datastream	100 GiB of change data capture a month per billing account (AlloyDB or Spanner to BigQuery) Source ↗	Data
Cloud Build	2,500 build-minutes/month (e2-standard-2) Source ↗	CI/CD
Cloud Deploy	First active delivery pipeline per billing account Source ↗	CI/CD
Artifact Registry	500 MB storage Source ↗	Containers
Cloud Logging	50 GiB logs ingestion/month, 30-day retention Source ↗	Logging
Cloud Monitoring	All non-chargeable Google Cloud metrics; the first 1 million time series read through the Monitoring API per billing account Source ↗	Monitoring
Workload Manager	5,000 resource evaluations a month Source ↗	Management
Secret Manager	6 active secret versions, 10K access operations/month Source ↗	Security
Cloud Source Repositories	Not available to new customers since June 17, 2024. Existing customers keep up to 5 project-users, 50 GB storage and 50 GB egress/month free. Source ↗	Source Control
Cloud Vision API	1,000 units/month (label detection, face detection, OCR, etc.) Source ↗	AI/ML
Cloud Natural Language API	5,000 units/month (sentiment analysis, entity extraction) Source ↗	AI/ML
Cloud Speech-to-Text	60 minutes/month per account on the V1 API. The V2 API has no free minutes. Source ↗	AI/ML
Cloud Video Intelligence API	1,000 units/month (label detection, shot detection) Source ↗	AI/ML
Cloud Translation API	500,000 characters/month (applied as a $10 monthly credit) Source ↗	AI/ML
Workflows	5,000 internal steps and 2,000 external HTTP calls per month Source ↗	Integration
Application Integration	400 executions/month, first 2 connection nodes for Google services; up to 20 GiB of data processed per month Source ↗	Integration
reCAPTCHA	10,000 assessments/month free on the Essentials and Premium tiers, counted per organization. Enterprise tier: $1 per 1,000 assessments on a 12-month commitment. Source ↗	Security
Cloud KMS	100 active key versions and 10,000 cryptographic operations per month, for keys created with Cloud KMS Autokey Source ↗	Security
Security Command Center	Standard tier Source ↗	Security
Web Risk	100,000 uris.search calls a month Source ↗	Security
Cloud Scheduler	3 jobs per billing account per month Source ↗	Scheduling
Firebase Auth Source ↗	Included at no cost on the Spark plan, except SMS phone sign-in. The 50K-MAU tier requires Identity Platform; Spark projects that upgrade are limited to 3,000 DAU. SAML/OIDC: 50 MAUs on Blaze, 2 DAU on Spark.	Auth
Firebase Hosting Source ↗	10 GB storage, 360 MB/day transfer, custom domain + SSL	Hosting
Firebase Realtime Database Source ↗	1 GB storage, 10 GB/month transfer, 100 simultaneous connections	Database

Cloud Run and Firestore together cover a small serverless backend within Free Tier limits. Cloud Run gives 2 million requests a month and Firestore 1 GiB of storage. Firebase Authentication, except SMS phone sign-in, is included at no cost on the Spark plan; its 50,000-MAU no-cost tier needs Identity Platform, and Spark projects that upgrade are limited to 3,000 daily active users. Add Cloud Build (2,500 build-minutes a month) for CI/CD and Cloud Logging (50 GiB a month) for observability.

Free Trial & AI/ML Tools

The $300 trial is credit for Google Cloud products over 90 days; accounts opened after 2026-03-02 cannot spend it on the Gemini API. During the trial you cannot add GPUs to VM instances, use Google Cloud Marketplace, request a quota increase or create Windows Server VMs, and the credit does not pay for partner generative AI models offered as a managed API (model as a service). Source ↗ The Gemini API has its own free tier, with limits shown per project in AI Studio.

Service	Free Limits	Category
$300 Free Trial Credit	$300 credit for 90 days. Credit card or other payment method required. Accounts opened after 2026-03-02 cannot spend it on the Gemini API. During the trial you cannot add GPUs to VM instances, use Google Cloud Marketplace, request a quota increase or create Windows Server VMs, and the credit does not pay for partner generative AI models offered as a managed API (model as a service). Source ↗	Trial
★ Gemini API (AI Studio)	Free tier on the Gemini 3.x Flash and Flash-Lite models; limits shown per project in AI Studio	AI/ML
Google Colab Unconfirmed	Free notebooks run for at most 12 hours. GPU access is heavily restricted and GPU types vary.	AI/ML
AlloyDB Omni	Free to download and use for development, testing, prototyping and demos; production or data-processing use is paid ($40 per vCPU a month).	Database
Data Studio (formerly Looker Studio)	Free for report creators and viewers. Connects to 1,400+ data sources. Data Studio Pro: $9/user/project/month.	Analytics

Gemini Code Assist's free tier for individuals has ended. Google's deprecation notice says that starting June 18, 2026, the IDE extensions stopped serving requests for the Gemini Code Assist for individuals, Google AI Pro and Google AI Ultra tiers, and that this also applies to Gemini CLI. Gemini CLI still runs with a Gemini API key.

Best Picks by Use Case

Common developer use cases and which GCP free services to combine. The stacks below can run within free tier limits, but the always-on server also pays for its external IPv4 address.

Recommended Free Stacks

Hobby API backend

Cloud Run (2M req/mo) + Firestore (1 GiB) + Firebase Auth + Cloud Logging. Serverless, auto-scales to zero, always free. Best for REST/GraphQL APIs with moderate traffic.

Always-on server

e2-micro VM (us-central1) + 30 GB persistent disk (choose Standard, the console defaults to balanced, which is billed) + Cloud Monitoring. Great for bots, cron jobs, small web servers. Add Cloud Shell for web-based SSH. An external IPv4 address costs about $3.65 a month, while external IPv6 is free.

Data analytics

BigQuery (1 TiB of queries and 10 GiB of storage a month) + Data Studio (formerly Looker Studio, free for report creators and viewers). Queries past the first 1 TiB a month are billed, including queries on public datasets. Use Cloud Run functions to load data on a schedule.

AI/ML prototyping

Gemini API free tier (AI Studio) + Cloud Run functions + Cloud Storage. Or use Google Colab's free notebooks, where GPU access is heavily restricted. The Vision, Natural Language and Speech-to-Text APIs have monthly free quotas. See our AI/ML tools guide for more options.

Static site with Firebase

Firebase Hosting (10 GB storage, custom domain + SSL) + Firebase Auth + Firestore. Or use Cloudflare Pages (unlimited bandwidth, always free) as an alternative. See our Firebase alternatives comparison.

Hidden Costs & Gotchas

GCP's free tier is generous, but these costs catch developers off guard.

Egress charges $0.12–0.23/GiB

Premium Tier is the default. From us-central1 it costs $0.12 per GiB (up to 1,024 GiB) to North America, Europe and Asia (excluding Korea and Indonesia), and up to $0.23 elsewhere. Compute Engine's Free Tier covers 1 GB a month from North America to every destination except China and Australia, on Premium Tier only. Cloud Storage's Free Tier covers 100 GB a month from buckets in us-west1, us-central1 or us-east1 to every destination except China and Australia, including Europe. Standard Tier's first 200 GiB a month are free, per account. Cross-region transfer is also billed.

e2-micro region restriction ~$7/mo wrong region

The free e2-micro VM is only free in us-west1, us-central1, and us-east1. Launching in any other region (including Europe, Asia) will incur standard Compute Engine charges.

External IPv4 address ~$3.65/mo always on

An external IPv4 address in use on a VM, including the free e2-micro, costs $0.005 an hour. The free tier covers one hour a month. A VM running all month pays about $3.65 for it. External IPv6 addresses are free. An ephemeral address is charged only while the VM runs.

Console disk default ~$3/mo for 30 GB

The free tier covers 30 GB-months of standard persistent disk. A disk created in the Google Cloud console defaults to balanced persistent disk, which the free tier does not cover (about $0.10 per GiB a month in us-central1, about $3 for 30 GiB). Choose Standard persistent disk, or create the VM with the gcloud CLI, which defaults to standard.

Cloud Storage class operations $0.05/10K Class A

Class A operations (writes, lists) cost more than Class B (reads). The free tier covers 5K Class A and 50K Class B/month. Heavy write workloads (logging, backups) can exceed this quickly.

Firestore daily limits $0.03–0.06/100K reads

The free quota covers the default database only; named databases get none. Free: 1 GiB of storage, plus 50,000 reads, 20,000 writes and 20,000 deletes a day. Past the quota, reads cost $0.03 per 100,000 in us-central1 and $0.06 in the nam5, nam7 and eur3 multi-regions.

Free trial credit expiration Not billed

The $300 credit lasts 90 days. When it is spent or 90 days pass, the Free Trial billing account closes unless you upgrade to a Paid billing account: resources stop, and data in services such as Compute Engine is marked for deletion. After a 30-day grace period without an upgrade, the resources are permanently deleted. The trial account is not billed, but a billing account set up through AI Studio is billed automatically once its credits are used up or expire.

Load balancer costs $18+/mo

The first 5 forwarding rules cost $0.025 an hour (about $18.25 for a 730-hour month). Data processed costs $0.008 per GiB in us-central1 and up to $0.0128 per GiB in other regions. Cloud Run and App Engine include their own load balancing for free.

Persistent disk snapshots $0.05/GB/mo

The Free Tier includes 30 GB-months of standard persistent disk and no snapshot storage. Standard snapshot storage costs $0.05 per GB a month in us-central1. Snapshots are incremental and billed on the compressed size of the data changed since the previous snapshot.

BigQuery streaming inserts $0.05/GB streamed

BigQuery queries are free up to 1 TiB/month, but streaming inserts cost $0.01/200 MB. Batch loading is free. If your app streams data in real-time, costs can add up.

Pro tip: Set up budget alerts immediately. Go to Billing → Budgets & alerts → Create Budget → set target to $1. You'll get notified at 50%, 90%, and 100% of the threshold. Also enable Recommender for cost optimization suggestions. A Google Cloud Free Trial billing account is not billed. New AWS accounts on the Free plan are not charged either; they close after 6 months or when their credits run out, unless upgraded to a Paid plan.

GCP vs AWS vs Others

Here's how they compare for developers building side projects or MVPs. See our AWS Free Tier guide for the full AWS breakdown.

Provider	Key Free Tier	Best For
AWS	Always Free: Lambda 1M req/mo, DynamoDB 25 GB	Enterprise workloads, serverless (Lambda)
AWS's Free Tier FAQ says most new customers need no payment method to sign up. AWS may ask for one to verify identity. (From aws.amazon.com/free/free-tier-faqs, read 2026-10-23.) The sign-up without a card is Sign up for AWS (new), launched September 16, 2026, but AWS's docs say it is still being released to a limited number of customers, so a new customer may not be able to use it yet. (From docs.aws.amazon.com/accounts/latest/reference/sign-in-new.html, read 2026-10-23.) Anyone who has or has had an AWS account is not eligible for AWS's Free plan or Free Tier credits. They must give payment information and go on the Paid plan. (From aws.amazon.com/resources/create-account, read 2026-10-23.) On the paid plan of Sign up for AWS (new), each project can have a monthly spend limit starting at $20. AWS pauses the project when spend reaches it. Advanced sign-up accounts have no spend limits. (From aws.amazon.com/blogs/aws/aws-reimagines-the-getting-started-experience, read 2026-10-23.)
Azure Source ↗	12 months: 750 hours each of B2pts v2 and B2ats v2 VMs. Always free: SQL Database (100K vCore seconds/mo, 32 GB), Cosmos DB (1K RU/s, 25 GB). $200 credit for 30 days, then pay-as-you-go to keep the free services.	Microsoft stack teams, enterprise
Cloudflare Source ↗	Workers 100K req/day, R2 10 GB (zero egress), D1 5 GB, Pages unlimited	Edge computing, static sites, storage
On the Free plan an account's Workers get 100,000 requests a day, shared with Pages Functions and reset at midnight UTC. Past that, a route set to fail closed returns Cloudflare's error 1027, and a route set to fail open skips the Worker as if none were configured. (From developers.cloudflare.com/workers/platform/limits, read 2026-10-06.) On the Free plan a request may use 10 ms of CPU time; one that uses more fails with error 1102. Time spent waiting on fetch calls, KV reads or database queries does not count. (From developers.cloudflare.com/workers/platform/limits, read 2026-10-06.) Cloudflare's terms prohibit processing or collecting credit card information on any site using its free services. (From cloudflare.com/terms, read 2026-10-06.) Cloudflare's subscription agreement's list of restrictions does not include commercial use. The only restriction specific to free services bars processing or collecting credit card information on the site. (From cloudflare.com/terms, read 2026-10-29.) Cloudflare's subscription agreement bars signing up for the services on behalf of a third party unless Cloudflare permits it in writing. (From cloudflare.com/terms, read 2026-10-29.) Cloudflare may change a Worker's workers.dev subdomain for any reason. Custom domains are not affected. (From cloudflare.com/service-specific-terms-developer-platform, read 2026-10-06.) Until December 1, 2026, Workers Logs on the Free plan allows 200,000 log events written per day with 3-day retention; from that date it allows 0.5 GB of log and trace ingestion per day for the whole account with seven-day retention. The Paid plan includes 20 million log events per month, then $0.60 per additional million, with 7-day retention. (From developers.cloudflare.com/workers/platform/pricing, read 2026-10-17.) From December 1, 2026, all Cloudflare logs and traces use the same pricing model. These share one account-level ingestion and storage allowance: Workers Logs, Workers Traces, Cloudflare Traces, Containers logs, R2 Data Access Logs, AI Gateway logs and Issues. Free includes 0.5 GB of ingestion per day and seven-day retention; ingestion stops at the daily limit and resumes at 00:00 UTC. Paid includes 50 GB of ingestion and 10 GB-month of storage per billing cycle, then $0.25 per GB of ingestion and $0.10 per GB-month of storage. (From developers.cloudflare.com/observability/pricing, read 2026-10-17.) Cloudflare Workers' Free plan needs no credit card. (From cloudflare.com/products/workers, read 2026-10-22.)
Railway Source ↗	$1 of free credit a month after 30-day $5 trial	Quick prototypes, hobby projects Sign up via our referral link and get $20 in credits What you have to do to get it The $20 is credited when the new account upgrades to a paid plan (Hobby or Pro), not at signup Get $20 in credits → We may earn a commission if you sign up through this link. See our affiliate disclosure.
After the 30-day trial, the Free plan allows 1 project, 3 services per project and no custom domains. Cron jobs are available during the trial only. (From railway.com/pricing, read 2026-10-12.) Railway's trial, with a one-time $5 credit, needs no credit card. (From railway.com/pricing, read 2026-10-22.) After 30 days passes or $5 is spent, the free trial reverts to the Free plan, which provides $1 of free credit per month. The credit does not roll over month to month. (From docs.railway.com/pricing/free-trial, read 2026-11-03.) Run `ssh railway.new` to get a free Linux VM with no account or credit card. The VM has 2 vCPU, 2 GB RAM, and a preview URL. Claude Code, Codex, OpenCode, and the Railway Agent are preinstalled. Railway identifies you by your SSH key. Each VM can be used for 60 minutes. You can get 3 VMs per IP address per day. (From docs.railway.com/pricing/free-trial, read 2026-11-02.) Open the claim link from the welcome message and sign up within 24 hours to keep the VM. The VM, files, and URL move into your account. Unclaimed VMs are deleted with their files. Railway caps free VMs at high demand. (From docs.railway.com/pricing/free-trial, read 2026-11-02.) Railway Sandboxes are available on every plan. Usage draws from the plan's included credit. Free plan limits are 10 sandboxes per environment, 2 vCPU and 2 GB each, and an idle timeout of at most 5 minutes. (From docs.railway.com/sandboxes, read 2026-11-03.) SMTP is only available on Railway's Pro plan and above. SMTP is disabled on the Free, Trial, and Hobby plans. Those plans must use transactional email services with HTTPS APIs. (From docs.railway.com/networking/outbound-networking, read 2026-11-07.) Railway's Hobby plan has a $5 monthly minimum, including $5 of usage credits. If usage is $5 or less per billing period, no extra charge applies. If usage exceeds $5, Railway charges the total usage. Included usage resets each billing cycle and does not accumulate. (From railway.com/pricing, read 2026-11-10.) The Limited Trial for trial accounts not verified through GitHub has restricted outbound network access and only a limited set of ports. Verification is an automated process at railway.com/verify. Railway does not respond to requests for verification. The Hobby plan unlocks the full experience. (From docs.railway.com/pricing/free-trial, read 2026-11-07.) Railway deletes stateful volumes created by Trial accounts 30 days after the expiration of the trial credits. An upgrade retains the data. (From docs.railway.com/pricing/free-trial, read 2026-11-07.)
Render Source ↗	Free web services (512 MB RAM), free PostgreSQL (30 days)	Small web apps, side projects
Render's docs say not to use Free instances for production applications. (From render.com/docs/free, read 2026-10-05.) Free web services cannot send outbound traffic on ports 25, 465, or 587, commonly used for SMTP. (From render.com/docs/free, read 2026-10-05.) A free Postgres database expires 30 days after creation and becomes inaccessible. You then have a 14-day grace period to upgrade to a paid plan before Render deletes it with all its data. (From render.com/docs/free, read 2026-10-05.) Free Postgres databases have no backups of any kind. (From render.com/docs/free, read 2026-10-05.) A free web service loses any changes to its local filesystem, such as uploaded images or a SQLite database, every time it redeploys, restarts or spins down, and it cannot attach a persistent disk. (From render.com/docs/free, read 2026-10-05.) If a workspace uses all 750 free instance hours in a month, Render suspends all of its free web services until the start of the next month. (From render.com/docs/free, read 2026-10-12.) Render's free web services need no credit card. (From render.com/articles/platforms-with-a-real-free-tier-for-developers-in-2026, read 2026-10-22.) Render Sandboxes runs isolated microVMs for agents and untrusted code. It is available on every plan including Hobby, billed at $0.12 per active CPU hour plus $0.038 per hour for memory for the default size. (From render.com/pricing, read 2026-10-29.) A Hobby workspace allows 100 concurrent sandboxes by default. There is no maximum session length on any plan, including Hobby. (From render.com/pricing, read 2026-11-07.) Render Sandboxes is in public beta. During the beta, sandboxes have no public URL and no private network to other Render services. (From render.com/docs/sandboxes, read 2026-10-29.) Render offers Free instances only for static sites, web services, Postgres and Key Value. Background workers and cron jobs need a paid compute plan. (From render.com/docs/free, read 2026-11-05.)
Supabase Source ↗	500 MB database, 50K MAU auth, 1 GB storage	Firebase alternative with SQL
Supabase makes daily backups only for Pro, Team and Enterprise projects, so a free project has none, and Supabase recommends exporting it yourself with the CLI's db dump command. (From supabase.com/docs/guides/platform/backups, read 2026-10-05.) A paused project can be restored from the dashboard for 1 year. After that, only its backup file and Storage objects can be downloaded. (From supabase.com/docs/guides/platform/upgrading, read 2026-10-05.) Supabase notifies you when you exceed the Free Plan quota. Continued overuse brings restrictions that can include a read-only database, 402 responses to every API request, or paused projects. They lift when the quota refills next billing cycle or on upgrade to Pro. (From supabase.com/docs/guides/platform/billing-faq, read 2026-10-05.) Supabase's Free Plan needs no credit card; paid plans require one on file. (From supabase.com/docs/guides/platform/get-set-up-for-billing, read 2026-10-22.)
Vercel Source ↗	Hobby plan (personal, non-commercial use only): 100 deployments/day, 100 GB Fast Data Transfer/month, 1M function invocations/month.	Frontend apps and Jamstack sites
Vercel restricts Hobby to non-commercial personal use: a deployment used for anyone's financial gain, including a paid employee or consultant writing the code, needs a Pro or Enterprise plan. (From vercel.com/docs/limits/fair-use-guidelines, read 2026-10-04.) In most cases, exceeding Hobby plan usage limits requires waiting 30 days before using the feature again. (From vercel.com/docs/plans/hobby, read 2026-10-07.) On Hobby, Vercel's terms let it use your content to train its AI models and share it with third parties for their model training; you can opt out in your team settings. (From vercel.com/legal/terms, read 2026-10-07.) Hobby teams get 10 GB of Deployment Storage; going over it can block new deployments until you delete some. (From vercel.com/changelog/hobby-projects-now-retain-fewer-deployments-to-free-up-storage, read 2026-10-07.) After 30 days, Vercel deletes a Hobby project's old deployments except its 3 most recent production deployments, its 3 most recent deployments of any type and its current production deployment. A team over its 10 GB of Deployment Storage loses those other deployments immediately. (From vercel.com/changelog/hobby-projects-now-retain-fewer-deployments-to-free-up-storage, read 2026-10-12.) Flat Rate CDN is available on Pro plans. Pro includes the lowest tier at no extra cost: 1M CDN requests and 1 TB of data transfer a month. Paid tiers are $20 a month for 10M requests and 50 TB, $100 for 50M and 50 TB, and $300 for 150M and 50 TB. A Pro team turns it on in Billing settings. Usage above a tier moves the team to a higher tier at the start of the next cycle. Flat Rate CDN is not for bulk file distribution or download services, or large-scale media or file delivery that uses most bandwidth. (From vercel.com/docs/pricing/flat-rate-cdn, read 2026-10-16.) Vercel's Hobby plan needs no credit card. (From vercel.com/docs/plans/pro-plan/trials, read 2026-10-22.)
Neon Source ↗	100 CU-hours and 1 GB of storage per project, auth up to 60k MAU, branching	Serverless Postgres with branching and scale-to-zero
Neon's Free plan needs no credit card. (From neon.com/pricing, read 2026-10-22.) An agent can create a Neon Postgres project without a Neon account. Unclaimed projects expire in 72 hours and are limited to 100 MB storage and 1 GB transfer. A person claims the project to keep it. (From neon.com/docs/reference/claimable-neon, read 2026-10-22.) On the Free plan, scale to zero can't be disabled: computes suspend after 5 minutes of inactivity. (From neon.com/docs/introduction/plans, read 2026-10-04.) Neon doesn't pause Free projects for inactivity: an idle database starts again on the next query, with no restore step. (From neon.com/pricing, read 2026-11-01.) Running out of CU-hours or network transfer suspends the compute until the next billing period or an upgrade. (From neon.com/docs/introduction/plans, read 2026-10-04.) Neon's Free plan provides 100 CU-hours per project per month. One CU allocates about 4 GB of RAM, and 100 CU-hours is enough to run a 0.25 CU compute (about 1 GB of RAM) for 400 hours, about 17 days of continuous running. All computes in a project count toward usage, including those for each branch and any read replicas. (From neon.com/docs/introduction/plans, read 2026-11-03.) At the storage cap, inserts, updates and deletes fail until space is freed or the plan is upgraded. (From neon.com/docs/introduction/plans, read 2026-10-04.) Neon's Azure East US 2, West US 3, and Germany West Central regions no longer accept new projects as of April 7, 2026; Free-plan projects there inactive for 90+ days may be deleted after October 5, 2026. (From neon.com/docs/import/azure-regions-deprecation, read 2026-10-07.) Platforms that create databases for their users can apply for the Agent Plan. You must have a Neon Scale plan with a credit card on file and get Neon's approval. Neon sponsors the databases for your free-tier users. Paid usage gets up to $25,000 in credits and compute at $0.106 per compute-unit hour. (From neon.com/docs/introduction/agent-plan.md, read 2026-10-11.) Neon gives $20 in free credits to try paid features, claimed through a promo link in the Neon Console. (From neon.com/pricing.md, read 2026-10-11.) Neon's pricing page offers startups up to $100k in credits. The startup page describes the Databricks Startup Program, up to $200K in credits for Neon and Databricks for qualifying startups. (From neon.com/startups, read 2026-10-11.) Neon's open source programme: Neon reviews each application and decides per project whether to support it with a monthly cash sponsorship through GitHub Sponsors, Neon credits, or both, and how much. Cash sponsorships come from a pool of at least $200,000 a year and run for six-month terms that do not renew automatically. Supported projects can receive up to $5,000 in Neon credits and up to $5,000 in Neon AI Gateway credits. It requires an OSI-approved license and a GitHub Sponsors profile. (From neon.com/programs/open-source.md, read 2026-10-29.) After the free plan, Neon's paid plans bill only for usage, with no monthly minimum: Launch charges $0.106 per CU-hour of compute and Scale $0.222, both charge $0.35 per GB-month of storage, extra branches cost $1.50 per branch-month, and network transfer beyond 500 GB per project costs $0.10 per GB. (From neon.com/pricing.md, read 2026-10-15.) Neon Realtime has been in public beta since 2026-10-15. It provides live updates from SQL queries without polling and is free during the beta. Compute and egress are billed under the plan; beta runs in four AWS regions: us-east-1, us-east-2, eu-central-1 and ap-southeast-1. (From neon.com/docs/realtime/overview, read 2026-10-16.) Neon archives branches older than 14 days and not accessed for 24 hours. Unarchiving happens automatically when a branch is connected to or queried. Branches with large amounts of data may connect and query slower while unarchiving. (From neon.com/docs/guides/branch-archiving, read 2026-10-29.)

The trade-off: GCP's Always Free tier includes an e2-micro VM, 2M Cloud Run requests a month and 1 TiB of BigQuery queries a month. For a complete free-tier stack combining the best of multiple providers, see our Free Startup Stack guide.

Recent GCP & Google Changes

GCP and Google pricing changes we've tracked. See the full timeline for all 513 tracked changes across all providers.

Date	Service	Change	Impact
effective Oct 23, 2026	Google Gemini API	The Gemini API keeps running. The Deep Research agent preview deep-research-pro-preview-12-2025 can be called until October 23, 2026, when Google shuts it down. Google names deep-research-preview-04-2026 as the replacement. Source ↗	LOW
effective Oct 22, 2026	Google Gemini API	The Gemini API keeps running. Gemini Omni Flash Preview (gemini-omni-flash-preview) and the Veo 3.1 preview models (veo-3.1-generate-preview, veo-3.1-fast-generate-preview and veo-3.1-lite-generate-preview) can be called until October 22, 2026, when Google shuts them down. Google names gemini-omni-1.1-flash as the replacement for all four. Source ↗	LOW
effective Oct 20, 2026	Google Gemini API	On Gemini Enterprise Agent Platform (formerly Vertex AI), Google retired gemini-2.5-pro, gemini-2.5-flash and gemini-2.5-flash-lite on October 20, 2026. Google names gemini-3.8-flash as a replacement for all three. On the Gemini API, Google shuts down gemini-2.5-pro and gemini-2.5-flash on December 1, 2026. gemini-2.5-flash-lite has no shutdown date on the Gemini API. Source ↗	MEDIUM
effective Sep 18, 2026	Google Gemini API	Since 2026-09-18 Google serves the Gemini 2.5 models only to users who have used them before, and points new projects to 3.5 Flash-Lite or 3.8 Flash. Source ↗	MEDIUM
effective Sep 18, 2026	Google	On 2026-09-18 Google stopped renewing every remaining standalone Google Developer Program Premium subscription, Workspace accounts included. Monthly Google Developer Program Premium subscribers keep Premium benefits until 2026-10-18 and annual subscribers until 2027-09-18; their membership then returns to the standard tier. Google recommends AI Pro or AI Ultra for consumers, and Gemini Enterprise for enterprises. Source ↗	MEDIUM
effective Sep 1, 2026	Firebase	From September 1, 2026, Firebase limits Remote Config on the no-cost Spark plan to 100,000 fetch requests a day per project. A Spark project over the limit keeps full service for 30 days from the first day it goes over, and after that requests above the limit are throttled unless it upgrades to Blaze. Blaze has the same 100,000 a day at no cost and then $0.06 per 10,000 requests. Firebase's pricing page had listed Remote Config as no-cost before. This applies to projects that enable Remote Config on or after September 1, 2026. Older Spark projects follow on December 1, 2026. Source ↗	MEDIUM
effective Aug 18, 2026	Google Content API for Shopping	Google Content API for Shopping was sunset on 2026-08-18, and its requests have returned progressive errors since 2026-09-01. Google names the Merchant API, generally available since July 2025, as its successor; developers who need more time can apply for extended access. Source ↗	HIGH
effective Jun 30, 2026	Google Tenor API	Google Tenor API shut down on June 30, 2026, and new API key sign-ups stopped on January 13, 2026. Only the API ended: Tenor.com, the GIF Keyboard apps and Gboard still serve Tenor GIFs. Developers who used the API need another GIF API. Source ↗	HIGH
effective Jun 18, 2026	Google Gemini Code Assist	Google stopped serving Gemini Code Assist for individuals, its free tier, on 2026-06-18. The IDE extensions and Gemini CLI no longer accept Login with Google for the individuals, Google AI Pro and Google AI Ultra tiers. Google stopped selling new Standard and Enterprise subscriptions on 2026-10-09. Existing subscribers can add licenses until 2027-01-31. Annual subscriptions can auto-renew until 2027-01-31; monthly subscriptions can auto-renew until 2027-12-31. Agentic coding assistance is now available through Antigravity, included in Gemini Enterprise editions. Source ↗	HIGH
effective Jun 18, 2026	Gemini CLI	Gemini CLI stopped accepting Login with Google for personal accounts on 2026-06-18, which ended the personal-account free tier of 60 requests a minute and 1,000 a day. It still runs with a Gemini API key, Vertex AI, or a Gemini Code Assist Standard or Enterprise licence. Source ↗	MEDIUM
effective Jun 1, 2026	Google Gemini API	Gemini 2.0 Flash and 2.0 Flash-Lite were shut down on 2026-06-01, as Google announced on 2026-02-18. Google names gemini-3.6-flash as the replacement for 2.0 Flash and gemini-3.1-flash-lite for 2.0 Flash-Lite. Source ↗	HIGH
effective May 31, 2026	Google Maps	Google Maps Platform client ID authentication reached its announced end date on May 31, 2026 and is being phased out: Google periodically blocks client ID requests ahead of a final termination date it has not published. API keys replace it. Source ↗	HIGH

Upcoming Changes

These are announced changes that have not taken effect yet, listed soonest first.

Date	Service	Change	Impact
effective Nov 18, 2026	Google Antigravity	Google Antigravity's Free and Google AI Plus plans keep running. Until November 18, 2026, they include Claude Sonnet 4.6, Claude Opus 4.6 and GPT-OSS-120b; Google removes these models on that date. After that date, Antigravity's models page lists two reasoning models for the Free and Google AI Plus plans: Gemini 3.8 Flash and Gemini 4 Hydrogen. Source ↗	LOW
effective Nov 19, 2026	Google Gemini API	The Gemini API keeps serving gemini-3.6-flash with no shutdown date. On Gemini Enterprise Agent Platform (formerly Vertex AI), gemini-3.6-flash can be called until November 19, 2026, when Google retires it. Google names gemini-3.8-flash as the replacement. Source ↗	LOW
effective Dec 1, 2026	Google Gemini API	Google shuts down Gemini 2.5 Pro (gemini-2.5-pro) and Gemini 2.5 Flash (gemini-2.5-flash) on the Gemini API on December 1, 2026. Both can be used until that date. After the shutdown, requests that still name gemini-2.5-pro are answered by a different model, gemini-3.8-flash, billed at Gemini 3.8 Flash rates, and requests that still name gemini-2.5-flash are answered by gemini-3.5-flash-lite, billed at Gemini 3.5 Flash-Lite rates. Google deprecated both on 2026-11-02 and lists breaking API changes in a migration guide for each replacement. Gemini 2.5 Flash-Lite (gemini-2.5-flash-lite) is not deprecated. The Gemini API keeps running. Source ↗	MEDIUM
effective Dec 1, 2026	Google Gemini API	Gemini 4 Argon on the Gemini API has introductory pricing through November 30, 2026. Standard input is $2.00 per million tokens through November 30, 2026, then $4.00 from December 1, 2026. Output is $10.00 then $20.00. Batch and Flex input is $1.00 then $2.00, output $5.00 then $10.00. Priority input is $3.60 then $7.20, output $18.00 then $36.00. It is not available on the Free tier. Source ↗	MEDIUM
effective Dec 1, 2026	Firebase	Firebase Spark projects that enabled Remote Config before September 1, 2026 can use it under the old terms until December 1, 2026. From December 1, 2026, the 100,000-fetches-a-day limit applies to them, with 30 days of full service from the first day they go over, then throttling unless they upgrade to Blaze. A Spark project that upgrades to Blaze before November 15, 2026 is not billed for Remote Config until February 1, 2027. Blaze projects from before September 1, 2026 start Remote Config billing on February 1, 2027. Source ↗	MEDIUM
effective Dec 1, 2026	Google Gemini API	Gemini 4 Carbon (gemini-4-carbon), generally available on the Gemini API since November 10, 2026, has introductory pricing through November 30, 2026. Standard input is $2.00 per million tokens through November 30, 2026, then $4.00 from December 1, 2026. Output is $10.00 then $20.00. Batch and Flex input is $1.00 then $2.00, output $5.00 then $10.00. Priority input is $3.60 then $7.20, output $18.00 then $36.00. It is not available on the Free tier. Source ↗	MEDIUM
effective Dec 3, 2026	Google Gemini API	Google shuts down Gemini Omni 1.1 Flash (gemini-omni-1.1-flash) on the Gemini API on December 3, 2026. It can be used until that date. After the shutdown, requests that still name gemini-omni-1.1-flash are answered by a different model, gemini-omni-1.2-flash, billed at Gemini Omni 1.2 Flash rates. Google released Gemini Omni 1.2 Flash on 2026-11-05. The Gemini API keeps running. Source ↗	LOW
effective Dec 13, 2026	Google Gemini API	gemini-live-2.5-flash-native-audio can be called on Gemini Enterprise Agent Platform (formerly Vertex AI) until December 13, 2026, when Google retires it. Google names gemini-3.8-live as the replacement. Source ↗	LOW
effective Dec 18, 2026	Google Gemini API	The Gemini API keeps running. The text-to-speech preview models gemini-2.5-flash-preview-tts, gemini-2.5-pro-preview-tts and gemini-3.1-flash-tts-preview can be called until December 18, 2026, when Google shuts them down. Google names gemini-3.8-flash-tts or gemini-3.8-flash-lite-tts as the replacement. Google moved this shutdown from an earlier date in November. Source ↗	LOW
effective Dec 31, 2026	Google Jules	Jules, Google's asynchronous coding agent, has been read-only since November 11, 2026, and winds down on December 31, 2026. Google announced this on 2026-10-29, giving users until November 11, 2026 to finish their sessions. Code, branches and pull requests Jules created stay in GitHub; settings, memories and session histories can be exported to a local folder. Source ↗	HIGH
effective Jan 1, 2027	Google Gemini API	Gemini 3.8 Flash and Gemini 3.6 Flash on the Gemini API have introductory pricing through December 31, 2026 on Google AI Studio and Gemini Enterprise Agent Platform. Standard input is $0.75 per million tokens through December 31, 2026, then $1.50 from January 1, 2027. Output is $3.75 then $7.50. Batch and Flex input is $0.375 then $0.75, output $1.875 then $3.75. Priority input is $1.35 then $2.70, output $6.75 then $13.50. The Free tier stays free of charge. Source ↗	MEDIUM
effective Jan 1, 2027	Google Programmable Search Engine	Existing customers can call Google's Custom Search JSON API until January 1, 2027, when Google discontinues it. It is closed to new customers. Programmable Search Engine engines set to search the entire web, or more than 50 domains, can do so until the same date. Google names Vertex AI Search as the alternative for up to 50 domains. Source ↗	HIGH
effective Jan 28, 2027	Google Gemini API	gemini-3.7-flash can be called on Gemini Enterprise Agent Platform (formerly Vertex AI) until January 28, 2027, when Google retires it. Google names gemini-3.8-flash as the replacement. On the Gemini API it is not shutting down and Google routes its requests to gemini-3.8-flash. Source ↗	LOW
effective Mar 15, 2027	Google Gemini API	The Gemini API keeps running. Gemini 2.5 Flash Image (gemini-2.5-flash-image) can be called until March 15, 2027, when Google shuts it down. Google names gemini-3.1-flash-lite-image as the replacement. Source ↗	LOW
effective Mar 31, 2027	Firebase	Firebase keeps running. The Firebase Extensions service can be used until March 31, 2027. Installed extensions keep running after that date. After March 31, 2027, they cannot be updated, reconfigured or uninstalled, and their configuration cannot be downloaded. Google's replacement is function kits in Cloud Functions for Firebase (2nd gen). Source ↗	MEDIUM
effective Mar 31, 2027	Google Gemini API	The Gemini API keeps running. Live API preview model gemini-2.5-flash-native-audio-preview-12-2025 can be called until March 31, 2027, when Google shuts it down. Google names gemini-3.8-live as the replacement. The other Live API preview, gemini-3.1-flash-live-preview, has no shutdown date. Google had dated both in November and moved the dates. Source ↗	LOW
effective Apr 1, 2027	Google Gemini API	text-embedding-005, text-embedding-004, text-multilingual-embedding-002 and multimodalembedding@001 can be called on Gemini Enterprise Agent Platform (formerly Vertex AI) until April 1, 2027, when Google retires them. Google’s table names no replacement for them. Source ↗	MEDIUM
effective May 7, 2027	Google Gemini API	The Gemini API keeps running. Gemini 3.1 Flash-Lite (gemini-3.1-flash-lite) can be called until May 7, 2027, when Google shuts it down. Google names gemini-3.5-flash-lite as the replacement. Source ↗	LOW
effective Jun 30, 2027	Google Cloud Deployment Manager	Google Cloud Deployment Manager is turned down after June 30, 2027. Support ended April 1, 2026. Since June 30, 2026, new users cannot enable the Deployment Manager API or create a first deployment. Existing deployments can use Deployment Manager until June 30, 2027. After that date, Deployment Manager can no longer be used. Resources it created continue to run and are managed with standard Google Cloud tools. Google recommends Infrastructure Manager. Source ↗	MEDIUM
effective Sep 14, 2027	Google Cloud Video Intelligence API	Google Cloud Video Intelligence API shuts down on September 14, 2027. It has been deprecated and unsupported since September 14, 2026, and can be used until its shutdown. Google recommends migrating to its Gemini family of models. Source ↗	MEDIUM
effective Sep 30, 2027	Firebase	Firebase keeps running. Firebase Test Lab can be used until September 30, 2027. After that date, no tests can run. Google's replacement is the Developer Device Platform on Google Cloud. The Developer Device Platform requires billing. The no-cost Spark plan includes 10 virtual-device tests a day and 5 physical-device tests a day. Source ↗	MEDIUM
effective May 14, 2028	Google Gemini API	The Gemini API keeps running. The embedding model gemini-embedding-001 can be called until May 14, 2028, when Google shuts it down. Google names gemini-embedding-2 as the replacement. Source ↗	LOW

Data Source

Powered by AgentDeals. The tables on this page were compiled by hand from official GCP pricing pages and have not been re-checked since. We track 28 Google Cloud-related entries and 46 GCP/Google pricing changes. The pricing changes we track are updated continuously; the tables above are not.

Query GCP data programmatically via our MCP tools — search for GCP services, compare with alternatives, or track pricing changes from your AI coding assistant.

A source check reads each cited page for that service's name and a price. The figures in the tables above are from our own records, except for the services whose entry below dates them to a read of the vendor's own page.

AWS — We read that on 2026-11-05 from aws.amazon.com/free, and our check recorded: the page names AWS as "aws" and states "$100"
Azure — We read that on 2026-10-28 from azure.microsoft.com/en-us/pricing/purchase-options/azure-account, and our check recorded: the page names Azure as "azure" and states "$200". Our stored terms were last confirmed on 2026-10-28.
Cloudflare Workers — We read that on 2026-11-02 from developers.cloudflare.com/workers/platform/pricing, and our check recorded: the page names Cloudflare Workers as "cloudflare workers" and states "$5" and "$0.30" and "$0.02" and "$5.00". Our stored terms were last confirmed on 2026-11-02.
Firebase — We read that on 2026-10-26 from firebase.google.com/pricing, and our check recorded: the page names Firebase as "firebase" and states "10 GiB/month" and "20K writes/day" and "50K reads/day" and "20K deletes/day". Our stored terms were last confirmed on 2026-10-03.
Google Colab — the page we cite for this offer states that it is free, in words, and states no allowance we can read
Neon — We read that on 2026-10-28 from neon.com/pricing, and our check recorded: the page names Neon as "neon" and states "$0". Our stored terms were last confirmed on 2026-10-28.
Railway — We read that on 2026-11-02 from railway.com/pricing, and our check recorded: the page names Railway as "railway" and states "$ 0" and "$5" and "$1" and "$ 5". Our stored terms were last confirmed on 2026-11-02.
Render — We read that on 2026-10-24 from render.com/pricing, and our check recorded: the page names Render as "render" and states "$ 0" and "$ 25" and "$0" and "$25". Our stored terms were last confirmed on 2026-10-02.
Supabase — We read that on 2026-11-02 from supabase.com/pricing, and our check recorded: the page names Supabase as "supabase" and states "$ 0" and "$ 25" and "$10" and "$ 10". Our stored terms were last confirmed on 2026-11-02.
Vercel — We read that on 2026-11-07 from vercel.com/pricing, and our check recorded: the page names Vercel as "vercel" and states "$0" and "$20" and "100 GB / month" and "1M allowed requests / month". Our stored terms were last confirmed on 2026-11-07.

Related Guides

Explore all 1,498 developer tool deals → Browse the full index or connect via MCP