Helium Moves Mobile Network Data to Apache Iceberg Tables Anyone Can Query With SQL, Switch-Over Due November 1

Helium Network HNT$0.458-2.7% is moving its Mobile network's data out of gzipped protobuf files in Amazon S3 and into Apache Iceberg tables that anyone can query with SQL, according to a blog post Helium published on 13 October 2026. Helium says the switch-over "is due to complete on Nov 1, 2026". The same day, a notice from Helium's X account told anyone who pulls Mobile network data from the old S3 files to switch over before that date.

Who has to move off the old S3 files before November 1

The notice applies to anyone whose software reads Helium Mobile data from the old files: Helium's X post says those readers "need to switch over before November 1, 2026". Helium's blog describes the migration as "underway rather than finished", with some tables fully moved and others "still written both ways" while Helium checks the new path against the old one.

Helium has not said which tables are in which state, or what happens to the old files after 1 November. The change covers the Mobile network only. Helium's post says the IoT network "runs its own oracles and its own tables" and is not part of the migration. As of 13 October 2026, Helium's oracle data documentation still describes only the old S3 bucket.

How the gzipped protobuf pipeline worked, and what it cost to read

Helium's oracles are the services that process the network's data-transfer and reward records, according to Helium's documentation. Under the old design, Helium's blog says, each oracle appended protobuf messages (a compact binary format) to a stream that was gzipped, cut into a new file on a timer or at 50 MB, and uploaded to S3. The millisecond timestamp in each filename "was the only index that existed".

Reading that data meant listing files, downloading each one, decompressing it and decoding it with the matching protobuf definitions. Helium's documentation says the bucket is "Requester Pays", so the reader covers Amazon's transfer and request charges. Helium's blog puts it bluntly: "It cost real money to read", and an analyst with a SQL client "had no way in at all".

Querying the Apache Iceberg tables with Trino, PyIceberg or Spark

Apache Iceberg is an open table format. As Helium's post describes it, the data still sits as unchangeable files in object storage, now in the Parquet format, but metadata alongside the files describes the columns and how the data is partitioned, so a query engine can work out which files to read and which to skip.

Helium's post gives working connection settings for three clients:

• Trino, the SQL engine Helium runs itself and the one configuration Helium says it "can vouch for"
• PyIceberg, a Python library that needs no cluster
• Apache Spark

All three point at the same Iceberg REST catalog, https://catalog.mobile.helium.io/api/catalog, with the warehouse name mobile. Helium says the catalog needs no login and the storage credentials are the literal string anonymous.

The tables sit in Cloudflare R2, which Helium says does not bill for data leaving the service, so "the only real cost of asking is running a Trino instance". Helium adds two cautions. Requests to the catalog are rate-limited, and a query that does not filter on a table's partition column reads the whole table.

What the ten tables show about Helium rewards and HNT emissions

Helium's post lists ten tables in three groups: data_transfer, hotspots and rewards. They cover offload sessions as the oracles ingested them, each hotspot's Data Credit burn, hotspot heartbeats, speed tests and bans, and the payouts to hotspots and service providers for each reward period. One table logs every multiplier ticket, accepted or refused, issued under HIP-150, Helium's location-multiplier scheme for Mobile deployers.

These records connect to Solana through HNT, Helium's token. Helium's documentation says the oracles handle all network data except the transactions that run through Helium's programs on Solana, and the blog calls the oracle files "the audit trail behind on-chain token emissions". The sessions table holds offload sessions, the kind of carrier traffic Helium announced at the first Wakefern grocery stores on 8 October 2026.

Helium's case for the move is that payouts on the network are meant to be verifiable, and that publishing the tables lets an outsider check what traffic a hotspot carried, what it was paid and which records were rejected. The post gives no data volumes, no cost figures and no list of outside parties that read the old files, so the number of teams affected by the 1 November date is not known.