Home·Comparison·RunPod

Kaldera vs RunPod: the structural difference

This page does not compare price lists. Prices and feature lists change weekly; a page that compares them is lying by the following week. What we compare is the business model: who pays for the broken second, how many line items the invoice has, and which legal system the data sits under.

Kaldera is in Phase 0: a broker MVP. Below, every item says whether it works today or is planned.

Method

What we compare, and what we do not.

We write nothing about a competitor that we cannot verify. If a cell in the RunPod column of the table below is empty, that does not mean "they do not have it" — it means "we did not verify it, check their own sources".

What we do compare

The structural things: business model, billing unit, the wording of the guarantee, whether the commission is stated openly, data sovereignty and the geography of the legal entity. These do not change overnight and can be verified from either side's own pages.

What we do not compare

The volatile things: hourly prices, GPU stock, uptime percentages, customer counts, support speed. These move day to day, and there is no honest way to measure most of them from the outside.

The table

Eight structural rows.

Everything in the Kaldera column either works today or states which phase it arrives in. The "—" in the RunPod column means unverified.

Structural itemKalderaRunPod
You don't pay for seconds that don't work Product commitment — a second that fails the health check enters no line item, and the job moves to another host
Hard budget cap Available — at the cap, jobs pause safely and the checkpoint is kept
Per-second billing Per second — measured at micro-dollar precision, visible on the live meter
Line-item transparency One screen — every line, including the storage charge of a stopped pod, on its own row
Region pinning In Phase 0 you can pick supply from EU-incorporated providers; the full region pinning package is Phase 3
EU + TR legal entity The incorporation plan: a TR legal entity first, then an EU one. Not yet incorporated; once it is, this row is updated with the date
Commission transparency 15% — the rate is written on the page and shown line by line in the host panel (host payouts are Phase 1)
Host scorecard Public — 4 metrics, the formula published, directly setting placement order

The "—" does not mean "the competitor lacks this". Kaldera did not verify that row for RunPod; before you decide, check RunPod's own current pages. The statements in the Kaldera column are our product decisions today, not a contract.

Three promises

We can fit the difference into three sentences.

You don't pay for seconds that don't work

This is not a marketing sentence but a rule in the billing code: a second that fails the health check is written to no line item. If a correction arrives after a period has closed, a refund line is opened.

Hard budget cap

Not a warning email but a real brake. At the cap, jobs are paused, the checkpoint is kept and you get a notice. A forgotten pod does not produce a surprise bill at the end of the month.

Per-second billing

No rounding up to the minute. Usage is held in micro-dollars; the invoice total and the sum of the usage events match exactly. The number you see on screen is the calculation itself.

Honesty

What works today, and what does not.

Kaldera is in Phase 0: a broker layer built on top of existing providers' APIs. The list below exists to show that what we sell and what we do are the same thing.

One account, one API, one invoicePhase 0 — working
Per-second metering and a live cost meterPhase 0 — working
Hard budget capPhase 0 — working
Provider scorecard (benchmark bot)Phase 0 — working
Our own host network and community supplyPhase 1 — in progress
Serverless inference endpointPhase 2 — planned
Sovereignty package and Verified host tierPhase 3 — planned
SOC 2 Type IProcess to be started — no certificate

In fairness

Where RunPod is better than us.

A comparison page is only worth reading if it also writes down the competitor's strengths. These are items we accept.

A mature product: RunPod has been in the field for years and is used in production. Kaldera is in Phase 0; maturity is earned with time, not with claims.
Serverless works today: their serverless inference product is ready and usable. Ours is in Phase 2 — if you need it this week, you cannot wait for us.
A wide template catalogue: they are ahead of us on the variety of ready images and templates. Our own model catalogue is still tagged "soon".
Global footprint: regions and capacity spread around the world. Our focus is EU + TR; we are not competing with them on global coverage.

These items change over time. Verify the current state on the RunPod side from their own pages, and ours from our status page.

Deciding

Which one, when?

RunPod may make more sense today

If you need a serverless endpoint this week, if you lean on a ready-made template catalogue, if you are looking for capacity in a region outside EU + TR, or if the maturity of a platform that has been in the field for years is your first criterion.

Kaldera may make more sense today

If you are tired of paying for broken nodes, if you want a hard cap against surprise bills, if you need to see every line of the invoice on one screen, or if keeping the data inside the EU + TR border is a condition set by your compliance team.

Get started

There is no cheaper way to verify a claim than testing it.

Open the console, start a pod, stop it while the meter runs. Count the seconds on the billing screen — if what we describe and what you pay do not match, write to us.

Kaldera is still at an early stage; this site is a sample marketing page and the numbers are illustrative.

This comparison is structural; prices and features change, so check both sides' current pages before deciding. If you find a row that is wrong, write to us and we will correct it and change the date. Last updated: 29 August 2026