Cloud updates
Cloud servers, app platforms and GPU clouds. Price changes, launches, policy changes and shutdowns, each dated and taken from the company's own pages. Prices in an update are as of its date; the comparison pages always show today's.
October 2026
Nebius switches preemptible GPUs to dynamic spot pricing
Since October 8, 2026, Nebius preemptible GPU prices move with demand, up to every 15 minutes. An H100 starts from $0.79 per GPU-hour, capped one cent below on demand.
AWS raises EC2 Capacity Block GPU prices about 15% from October 7
Reserving B300 GPUs through EC2 Capacity Blocks for ML now costs $16.146 per GPU-hour, up from $14.04. On-Demand and Savings Plans prices did not change.
Nebius raises on-demand GPU prices by up to 21%
From October 1, 2026, a Nebius H100 costs $4.50 per GPU-hour on demand instead of $3.85, and a B300 $9.50 instead of $7.85. The RTX PRO 6000 stays at $1.80.
Render cuts Key Value prices by 20% to 45% and adds 3 GB and 50 GB plans
From October 1, 2026, Render's 1 GB Key Value plan costs $20/mo instead of $32, and the 40 GB plan $600 instead of $1,100. Existing paid instances get the lower prices automatically.
September 2026
Railway offers free 2 vCPU VMs you can start without an account
Since September 25, 2026, running ssh railway.new starts a free Linux VM with 2 vCPU, 2 GB of RAM and coding agents preinstalled. You get 60 minutes to build and 24 hours to claim it.
Supabase moves logs to usage-based pricing, with billing from early 2027
Supabase said on September 25, 2026 that log ingest will be metered. Paid plans include 20 GB a month, then $0.50 per GB; Free includes 1 GB. Billing starts after a grace period.
Google Cloud will shut down T4 and P4 GPUs on August 1, 2027
Google Cloud said on September 22, 2026 that NVIDIA T4 and P4 GPUs reach end of support on August 1, 2027. P100 GPUs were already shut down on September 15, 2026.
Runpod launches global volumes that mount in any data center
Runpod's global volumes, in beta since September 15, 2026, attach to GPU Pods and Serverless workers in any region. Requests cost $0.005 or $0.0005 per 1,000.
Runpod starts charging sales tax on credit purchases
Since September 10, 2026, Runpod adds sales tax on top of credit purchases in some jurisdictions. A $100 top-up at an 18% rate costs $118. A business tax ID may exempt you.
Vercel prices Password Protection on Pro at $20 per project, not $150 a team
Since September 9, 2026, Vercel Pro teams can password-protect a single project for $20 a month. Before, it took a $150/mo team add-on that covered every project.
Vercel launches Flat Rate CDN for Pro teams, with tiers from $20 a month
Since September 8, 2026, Vercel Pro teams can pay a fixed monthly price for CDN traffic. Pro includes 1M requests and 1 TB; paid tiers cost $20, $100 or $300 a month.
August 2026
DigitalOcean launches v5 Droplets, priced per resource with no monthly cap
Since August 25, 2026, DigitalOcean offers v5 Droplets in Atlanta, Richmond, Kansas City and Memphis. You pay by the hour for vCPU, memory and disk, with no monthly cap.
Google Cloud us-west1 outage: fiber work broke Compute Engine, GKE and more
On August 20, 2026, fiber maintenance cut network capacity between data centers in Google Cloud's us-west1 region in Oregon, causing errors and timeouts across more than 20 services.
Hyperstack adds NVIDIA B300 on demand at $7.40 per GPU-hour
Hyperstack lists 8-GPU NVIDIA B300 virtual machines on demand at $7.40 per GPU-hour in its CANADA-1 region. Reserved private B300 clusters are planned for Q4 2026.
Google Cloud's flexible commitments now discount G2 and G4 GPU VMs
Since August 11, 2026, spend-based flexible CUDs cover G4 (RTX PRO 6000) VMs at 16% off for 1 year or 42% for 3 years, and G2 (L4) at 21% or 43%, with no GPU reservation needed.
DigitalOcean launches Spot GPU Droplets for B300, MI355X and MI350X
Spot GPU Droplets entered public preview on August 10, 2026. On October 18 the pricing page listed a B300 at $6.17 per GPU-hour; the price is locked when you create the Droplet.
DigitalOcean raises H100, H200 and MI300X GPU Droplet prices about 30%
From August 1, 2026, an H100 GPU Droplet costs $4.41 per GPU-hour on demand instead of $3.39. The L40S and RTX Ada Droplets kept their prices.
July 2026
Azure West US outage: a widened network repair cut traffic for five hours
On July 23, 2026, traffic into and out of Azure's West US region failed or slowed from 14:44 to 19:41 UTC. Microsoft's review blames a defect that widened an optical repair to a whole datacenter.
Cloudflare Workflows starts charging for steps and storage on paid plans
Cloudflare said on July 7, 2026 that Workflows would bill steps and stored state from August 10. Workers Paid includes 500,000 steps a month, then $0.80 per extra 100,000.
Akamai drops the monthly price cap for G8, GPU and extra IPv4 on Linode
From July 1, 2026, Akamai bills G8 Dedicated Linodes, GPU Linodes and additional IPv4 addresses for every hour used, with no monthly cap. Other compute keeps the cap.
June 2026
Hetzner raises cloud prices for new orders; CPX and CCX more than double
Since June 15, 2026, new Hetzner Cloud servers cost more. In Germany and Finland, CX23 went from $4.99 to $6.49/mo and CPX22 from $9.49 to $22.99/mo, before IPv4 and VAT.
Oracle Cloud halves its Always Free Arm compute to 2 OCPUs and 12 GB
Oracle's Always Free tier now covers 1,500 OCPU hours and 9,000 GB hours of Ampere A1 a month, half the old 3,000 and 18,000: 2 OCPUs and 12 GB of memory instead of 4 and 24.