Google Cloud Review 2026

What you’re actually signing up for

Google Cloud is Google’s public cloud, run from Mountain View, California. It launched in 2008 with App Engine and now sells more than 100 products, from Compute Engine and BigQuery to Vertex AI, where Gemini models are metered SKUs like anything else. For you the unit that matters is the billing account. Credits, tiers and program grants attach to it, and every project linked to it draws on the same balance.

Google courts startups harder than any other hyperscaler. The startup page leads with the line that nearly all generative AI unicorns run on it, and that pitch shapes the credit programs. Grants are larger for companies that build on Gemini, and the AI-first track adds $150,000 in year one on top of the standard Scale tier. Read the programs with that bias in mind.

How the $300 trial really behaves

Signing up at cloud.google.com/free creates a Free Trial billing account preloaded with $300, valid for 90 days. You need a card or another payment method, and Google places a $0 to $1 authorization that isn’t a charge. You must never have paid for Google Cloud, Google Maps Platform or Firebase, and government entities in the US and Mexico are excluded unless Google agrees in writing.

While you’re on the trial you can’t attach GPUs, create Windows Server VMs, use Marketplace, provision VMware Engine or request quota increases, and you’re capped at eight concurrent Compute Engine cores. Eight cores doesn’t go far. Clicking Activate on the Welcome page lifts those limits, keeps your remaining credit until the original expiry date and starts billing anything beyond it to your card. Credit left on day 90 is gone.

What stays free after the trial

The Always Free tier is separate from the trial and has no end date. It includes one non-preemptible e2-micro VM per month in us-west1, us-central1 or us-east1, 30 GB-months of standard persistent disk, 5 GB-months of regional Cloud Storage, 1 TiB of BigQuery querying with 10 GiB of storage, 2 million Cloud Run requests, 2 million Cloud Run functions invocations, 1 GiB of Firestore and 10 GiB of Pub/Sub messages. That’s a small web app with a database, for free, for as long as Google keeps the list.

Limits count per billing account per month, don’t roll over and can change with 30 days of notice. The marketing page and the docs disagree on Cloud Build, with the docs listing 2,500 build-minutes a month, so trust the docs. Quota increases are a separate mechanism from credits and are unavailable until you activate, which trips up anyone who wants more than a handful of vCPUs during the trial.

What it costs once the credit is gone

Pricing is pay as you go per SKU, with discounts for commitment and interruptibility. Spot VMs run up to 91% off on-demand, resource-based committed use discounts reach 55% for most machine series and 70% for memory-optimized ones, and flexible commitments pay 28% for one year or 46% for three on series such as C3 and N2. Interruptible work is where the real money is; commitments are for the steady floor.

For Gemini, Vertex AI lists Gemini 3.8 Flash at $0.75 per million input tokens and $3.75 per million output tokens through December 31, 2026, rising to $1.50 and $7.50 on January 1, 2027. Gemini 4 Argon is $2.00 and $10.00 through November 30, 2026, then $4.00 and $20.00. Non-global endpoints add 10%. You can control billing with alerts-only budgets, and since October 13, 2026 with spend cap budgets on a limited set of services.

Credits for startups, researchers and students

The Google for Startups Cloud Program has two tiers. Start, for pre-funded companies founded within 24 months with a working MVP, grants $2,000 plus 12 months of Workspace Business Plus and $200 in Google Skills credits. Scale, for pre-seed to Series A companies with institutional equity funding, founded within five years and holding no more than $5,000 in prior credits, covers 100% of usage up to $100,000 in year one and 20% up to $100,000 in year two.

AI-first Scale companies get year one raised to $250,000, which is where the $350,000 headline comes from. The seed-to-Series A tier adds $12,000 in Enhanced Support credits, $600 a month of Maps Platform for a year and $500 in Google Skills credits. Program credits don’t cover Marketplace or third-party models. Researchers can apply for $5,000 through the education site, and students can request 200 Google Skills credits that expire after one year.

Where your Gemini spend should live

The reason to put AI spend on Google Cloud rather than AI Studio is that Vertex AI usage is an ordinary Google Cloud SKU. Program credits and the Welcome credit absorb it, enterprise controls apply, and the same billing account carries BigQuery and Cloud Run. Model Garden lists over 200 models, but only Google models such as Gemini and Gemma are covered by program credits. Partner models bill directly, credits or not.

Regions matter for both cost and credits. The Always Free VM and storage only count in us-west1, us-central1 and us-east1. Vertex AI global endpoints are 10% cheaper than regional ones. Since January 27, 2026, Google AI Pro and Ultra subscribers also get $10 or $100 a month in Google Cloud credits usable on Vertex AI, Cloud Run and the Gemini API, which is enough to keep a small agent in production without touching your card.

What can go wrong with your account

The trial carries no SLA, no indemnity and no technical support. If your card expires mid-trial the account may be suspended until you fix it. When the trial lapses without activation, resources stop and you get 30 days to contact support, migrate data from services other than Compute Engine, or upgrade. After that the data is deleted.

Paid support is a separate subscription, which is why the Scale tier bundles $12,000 of Enhanced Support credits. Quota increase requests only work on an activated account and can take days for GPUs. Keep a budget alert and a disabling action in place from day one. An alerts-only budget warns you, but it doesn’t stop the spend.

Where we land on it

For a solo developer, the Always Free tier plus the $300 trial is a fair deal as long as you treat the trial as a 90-day clock with hardware limits. For a funded startup building on Gemini, the program is the most generous of the big three in year one, and the credits flow through the same billing account as the rest of your stack. We give it 4.3: strong credits and clear docs, held back by trial restrictions and the AI Studio exclusion.