AWS t2.micro vs t3.micro: Why t3 Wins (And When t2 Still Makes Sense)
Specs Comparison
Both t2.micro and t3.micro are entry-level burstable instances, but t3.micro delivers noticeably more for a similar price:
| Spec | t2.micro | t3.micro |
|---|---|---|
| vCPUs | 1 | 2 |
| RAM | 1 GiB | 1 GiB |
| Processor | Intel Xeon (variable gen) | Intel Xeon Platinum 8000 |
| Baseline CPU | 10% of 1 vCPU | 10% of each vCPU (20% total) |
| CPU Credits/hr | 6 | 12 |
| Network | Low to Moderate | Up to 5 Gbps |
| EBS Bandwidth | Standard | Up to 2,085 Mbps |
| Hypervisor | Xen | Nitro |
| Free Tier | Yes | Yes |
t3.micro gives you 2 vCPUs with 20% total baseline vs t2.micro's 1 vCPU with 10% baseline. That means t3.micro can handle twice the sustained CPU load before it starts consuming burst credits. Combined with newer Nitro-based infrastructure and faster networking, t3.micro is the objectively better spec sheet.
Pricing Comparison
Counterintuitively, the newer and more powerful t3.micro is actually cheaper than t2.micro in most regions:
| Region | t2.micro (On-Demand) | t3.micro (On-Demand) | Difference |
|---|---|---|---|
| us-east-1 | $0.0116/hr | $0.0104/hr | t3 is 10% cheaper |
| us-west-2 | $0.0116/hr | $0.0104/hr | t3 is 10% cheaper |
| eu-west-1 | $0.0126/hr | $0.0114/hr | t3 is 10% cheaper |
| ap-southeast-1 | $0.0132/hr | $0.0118/hr | t3 is 11% cheaper |
Monthly Cost (Running 24/7, ~730 hours)
The per-hour difference is small ($0.0012/hr), but t3.micro is cheaper and more powerful. There's no pricing reason to choose t2.micro — the cost advantage goes to t3 in nearly every region.
CPU Burst Model: The Critical Difference
Both t2 and t3 are burstable instances, meaning they earn CPU credits when idle and spend them when they need more than their baseline. But the burst models work very differently:
| Behavior | t2.micro (Standard) | t3.micro (Unlimited by Default) |
|---|---|---|
| Baseline | 10% of 1 vCPU | 10% of each vCPU (20% total) |
| Credits earned/hr | 6 | 12 |
| Max credit balance | 144 (24 hrs of earning) | 288 (24 hrs of earning) |
| When credits run out | Hard throttle to baseline | Keeps bursting, charges $0.05/vCPU-hr |
| Default mode | Standard (no unlimited) | Unlimited |
t2.micro throttles hard when credits run out — your instance drops to 10% CPU and stays there until credits accumulate. Your application slows to a crawl. t3.micro keeps running at full speed but charges you for the extra CPU. Neither is “free” when you burst beyond your credits — t2 costs you performance, t3 costs you money.
Benchmark Comparison
The t3 family runs on newer Intel Xeon Platinum processors (Skylake or Cascade Lake) with the Nitro hypervisor, while t2 uses older Xen-based infrastructure. In practice, this means:
t3.micro Performance Advantages
For CPU-bound tasks like compiling code, running tests, or processing data, t3.micro's newer silicon and extra vCPU make a meaningful difference. For I/O-bound workloads like serving a low-traffic website, the gap is smaller but t3 still wins on network and storage throughput.
The Credit Exhaustion Trap on t2
This is the single biggest reason to avoid t2.micro for anything remotely important. Here's the scenario:
- You launch a t2.micro and it starts with 30 CPU credits (a launch bonus).
- Your app runs fine for a while because those initial credits let you burst freely.
- Credits slowly drain as your app occasionally uses more than the 10% baseline.
- Credits hit zero — your instance is now hard-capped at 10% of one vCPU.
- Everything becomes painfully slow — SSH takes 10 seconds to respond, web pages take 30+ seconds, background processes stall.
With t3.micro in unlimited mode, step 4 never happens. Your instance keeps running at full speed and you pay a small overage charge (~$0.05 per vCPU-hour of burst). For a t3.micro, that's typically pennies per day even under sustained load.
We've seen developers spend hours debugging “slow application” issues that turned out to be nothing more than t2 credit exhaustion. The app wasn't broken — the CPU was just throttled to 10%. Switching to t3 fixed it instantly.
When t2.micro Still Makes Sense
Despite t3's advantages, there are a few legitimate reasons to choose t2.micro:
- Legacy AMIs that require Xen — Some very old AMIs (pre-2018 custom images, certain legacy marketplace AMIs) don't support the Nitro hypervisor that t3 requires. If your workload depends on a Xen-only AMI, t2 is your only option in the micro tier.
- Region availability — In rare cases, t3 instances may not be available in a specific Availability Zone where you already have infrastructure. t2 has broader legacy availability in older AZs.
- You want hard throttling, not surprise charges — If you're running a non-critical workload and would rather have it slow down than generate unexpected costs, t2's standard mode (with no unlimited option) gives you a hard ceiling on spending. Though you can also disable unlimited mode on t3.
Unless you have a specific technical constraint (legacy AMI, specific AZ), choose t3.micro. It's cheaper, faster, has 2x the CPU baseline, better networking, and won't hard-throttle your application when credits run out. Both are free tier eligible, so there's no cost difference for new accounts.
Frequently Asked Questions
Is t2.micro cheaper than t3.micro?
No. In most regions, t3.micro is about 10% cheaper than t2.micro. In us-east-1, t3.micro costs $0.0104/hr vs $0.0116/hr for t2.micro. The only potential extra cost with t3 is if you sustain high CPU usage in unlimited mode, which generates small overage charges ($0.05/vCPU-hour).
Which is better for a small website?
t3.micro. A small website benefits from t3's 2 vCPUs (web servers like Nginx can use both cores), better networking (up to 5 Gbps vs “Low to Moderate”), and unlimited burst mode (so traffic spikes don't throttle your site). The t3.micro handles concurrent requests noticeably better than t2.micro.
Can I switch from t2.micro to t3.micro?
Yes. Stop your instance, go to Actions → Instance Settings → Change Instance Type, select t3.micro, and start it again. The process takes about a minute. Your data, IP address (if using Elastic IP), and security groups all stay the same. The only requirement is that your AMI must support the Nitro hypervisor — any modern Amazon Linux, Ubuntu, or Debian AMI does.
Do both t2.micro and t3.micro qualify for the free tier?
Yes. As of 2024, both instance types are free tier eligible. New AWS accounts get 750 hours per month of either t2.micro or t3.micro for 12 months. See our AWS Free Tier EC2 guide for the full breakdown of what's included and what costs extra.
Compare All EC2 Instances
Use CloudBench to compare specs, pricing, and benchmarks across every EC2 instance type. Find the best instance for your workload and budget.
Explore Instance Pricing →