The Akash GPU utilisation number everyone cites is from October 2024
If you have read anything about Akash GPU utilisation in the last year, you have probably seen a figure somewhere near 79%. It gets repeated in research notes, threads and investor decks.
We track this number daily. It is not 79%, and it has not been 79% for a long time.
Where the 79% comes from
The figure traces back to a single blog post published in October 2024 — roughly twenty-two months ago as of writing.
That post has three problems:
Meanwhile, Akash's own Q1 2026 report publishes no utilisation figure at all. There is nothing official to check it against.
What Akash's API actually returns
One command, no key, no signup:
curl -s https://console-api.akash.network/v1/network-capacity | jq '.resources.gpu'
At the time of writing that returns:
{"active": 240, "pending": 2, "available": 180, "total": 424}
That is Akash's endpoint, not ours. We are not asking you to trust our number — we are pointing at theirs.
The daily series
We have measured this every day since 9 August. Numerator and denominator both, because a ratio you cannot decompose is a ratio you cannot audit:
| Date | Active | Total | Utilisation | |---|---|---|---| | 2026-08-09 | 127 | 423 | 30.0% | | 2026-08-11 | 179 | 429 | 41.7% | | 2026-08-13 | 197 | 425 | 46.4% | | 2026-08-15 | 249 | 417 | 59.7% | | 2026-08-18 | 268 | 420 | 63.8% | | 2026-08-21 | 257 | 422 | 60.9% | | 2026-08-22 | 200 | 422 | 47.4% | | 2026-08-25 | 226 | 431 | 52.4% | | 2026-08-27 | 268 | 435 | 61.6% | | 2026-08-28 | 248 | 433 | 57.3% |
Two things worth noticing, and the second is the interesting one.
First: demand roughly doubled while supply barely moved. Active GPUs went 127 → 248 over the window. Total capacity went 423 → 433 — about 2%. That is a genuine demand ramp, not a supply artefact. Second: it is volatile. It touched 63.8% on the 18th, fell to 47.4% on the 22nd, and recovered to 61.6% on the 27th. A sixteen-point swing in five days.That volatility is why a single number quoted without a date is close to meaningless. Any figure — including ours — is a point on a moving series. The honest unit is the series, with the date attached.
Why the denominator is the whole story
Capacity is not demand. Most DePIN networks publish supply — node counts, total capacity, provider tallies. Very few publish the ratio, because the ratio is the number that can embarrass you.
This matters right now for Akash specifically. They passed Provider Incentive Pilot 02 with roughly a $10M budget aimed at substantially increasing GPU supply. If supply grows and the numerator does not follow, utilisation falls — and utilisation is the metric that says whether the incentive bought demand or just capacity.
Nobody is publishing that ratio daily. So we do.
How to check anything here
Every number above decomposes into a numerator and a denominator from a public endpoint, with a timestamp. That is the standard we hold ourselves to: no estimates, no zero-fill, no back-fill. If a metric is not flowing, it is absent, not zero.
- Akash's own API: the curl above
- Our free tier, no signup:
/try - The method and current coverage: utilisation diligence
The general lesson
We are not picking on one blog post. The point is structural: a number with no date, no source and no denominator will outlive the reality it described, because it is easier to quote than to check.
The Akash figure is a clean example — twenty-two months stale, internally inconsistent, twenty points off, and still circulating. There are others. If you are underwriting a DePIN position on a utilisation figure, the first question is not "what is the number" but "when was it measured, from what, over what denominator."
If the answer is not immediately available, the number is decoration.