Lightning AI

What it is for

Lightning AI is a GPU Studio you open in the browser: a persistent VS Code / Jupyter workspace, a shared Drive, then Jobs or a serving process when you leave the laptop. Use it when the job is code + GPU hours on a credit meter, not when you need a $20 chat seat.

What makes it unique: the cover is the live lightning.ai Studio workspace — lightning_studio file tree, train.py, GPU pane, terminal — not a pricing card.

Key features

• Studio — a persistent cloud workspace with in-browser VS Code / Jupyter, SSH into Cursor / PyCharm, and a machine picker (AI Studio). One workspace. CPU vs T4 vs H200 is SKU of this machine, not five products. The first 4-CPU instance is free.
• Lightning Drive — shared Teamspace storage (browser upload, SDK, or an S3/EFS connect) that survives sleep and is visible to every Studio in the space (same docs, “Native persistent storage”). A disk, not the editor.
• Jobs / multi-node — duplicate the Studio environment onto N machines or reserved job hardware when interactive iteration is done (docs “Scale to multi-node” / “Start 100 jobs”). A batch, not a live editor.
• Serve / inference — leave a Studio running a server, or use the platform model APIs (docs deploy example; pricing req/min and tok/min caps). A process that answers HTTP, not a training loop.
• BYO cloud / VPC — attach AWS or GCP (Enterprise) so the same Studio runs on your account and credits (AI Studio “Connect cloud account”). A tenancy, not a GPU SKU.

pip install lightning-sdk and the in-Studio copilot are install / assist. Credits ($1 = 1 credit) and $/GPU-hour are the meter. Templates on the hub are starting points, not a second product.

Who it's for

Best fit: ML engineers who will burn the 5 free credits (and the free 4-CPU Studio), then buy hours — T4 $0.55/hr, H200 $4.50/hr — instead of renting a bare VM.

Pricing

Verified 2026-10-22 (Asia/Shanghai) from lightning.ai/pricing and AI Studio. Usage is credits (1 credit = $1). Purchased credits expire in 12 months.

• Free start — no card. 5 credits on signup; add a card for 25 more. One free 4-CPU Studio (restarts every 4 hours on Free). Phone verify unlocks up to 80 free GPU hours (SKU-dependent: T4 75, L4 31, L40S 5, A100 40GB 10, A100 80GB 5).
• On-demand GPU (Lightning cloud) — T4 $0.55/GPU-hr; L4 $0.79; L40S $2.14; RTX PRO 6000 $2.89; A100 40GB $2.19; A100 80GB $2.71; H100 $4.68; H200 $4.50 (interruptible H200 $3.82–$4.58). Sleeping Studios cost $0.
• Paid tiers — raise concurrent GPUs, drop the 4-hour Free restart, and unlock multi-node / reserved machines. Confirm the live Free / Pro / Teams / Enterprise matrix on lightning.ai/pricing at checkout — do not invent a seat.
• Enterprise — BYO AWS/GCP, VPC, SOC2 / HIPAA, 24/7. Custom.

The AWS Marketplace “Teams Tier 1 Seat $140/month / 50 credits” card is a marketplace SKU, not the self-serve lightning.ai table. Confirm which door you are buying.

Limitations / costs

• The free CPU Studio restarts every 4 hours. Pro+ drops that restart.
• GPU hours burn credits. The editor looks the same on CPU and H200; the bill does not.
• Credits expire in 12 months. Unused signup credits do too.
• A Studio, not a GPU marketplace with only a price list. For published community-cloud $/hr see Runpod or Salad.

Alternatives

• Runpod — a GPU marketplace when you want to rent a pod, not a persistent Studio.
• Modal — serverless functions when you do not want a long-lived workspace.
• SkyPilot — an OSS control plane when the GPUs are already yours.

Official tutorial

Official Lightning AI channel video: Lightning AI Demo: Build, Train & Deploy AI Models Faster (End-to-End Platform Walkthrough) (oEmbed author: Lightning AI). First-party Studio → Drive → GPU switch → deploy.

Written path: AI Studio plus Pricing.

Website

https://lightning.ai

Repository

https://github.com/Lightning-AI/lightning

Source

Catalog row curated for AI Tools Nav. Cover is the live lightning.ai Studio workspace (train.py + GPU pane). Dollars checked on lightning.ai/pricing and the Studio docs (5 free credits; T4 $0.55/hr; H200 $4.50/hr; 1 credit = $1). Tutorial oEmbed author: Lightning AI.

Last verified

2026-10-22 (Asia/Shanghai) editorial check.