All articles
solananetworkhistoryvalidatorsinfrastructure

Epoch 1000: six years of Solana, by the numbers

Solana mainnet crossed its 1,000th epoch on July 10, 2026: 432 million slots, 530 billion transactions, 8 outages totalling 64.5 hours, and 885 days of unbroken uptime. The honest history, epoch by epoch.

Share
devrels.xyz/a/152short link

At 04:11 UTC on July 10, 2026, Solana mainnet crossed into its 1,000th epoch. There was no ceremony on chain; the leader schedule rotated, stake activations kicked in, rewards paid out, and slot 432,000,000-and-change got produced like any other. That is the point. One thousand epochs of exactly 432,000 slots each, one continuous ledger since March 16, 2020, no reset, no rollback, no restart-from-snapshot-zero. The Solana Foundation marked it with solana.com/epoch1000, a wallet age checker that tells you which epoch your wallet first appeared in, no connect, no signature.

Milestone posts usually come in two flavors: the victory lap that forgets September 2021 ever happened, and the dunk thread that pretends February 2024 was yesterday. Builders deserve the actual ledger. Here it is, both columns.

What an epoch actually is

An epoch is 432,000 slots, fixed, with a 400 ms target per slot. At the theoretical minimum that is 48 hours; real slot times put an epoch at roughly two to two and a half days (older epochs ran longer when slots were slower; across all 1,000 the average works out to about 2.31 days). Three things happen at every boundary:

  • Stake changes land. Pending activations and deactivations take effect, refreshing the active stake set. This is why your stake "warms up" instead of counting instantly.
  • Rewards pay out. Staking rewards are recalculated and distributed; since SIMD-0118, the payout is partitioned across multiple blocks at the start of the epoch instead of jamming one block.
  • The leader schedule rotates. Every validator computes the next epoch's slot-by-slot leader assignment deterministically from stake weights, fixed in advance for the whole epoch.

If you want the current state of the boundary math yourself, it is one RPC call away:

bash
curl -s https://api.mainnet-beta.solana.com -X POST \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"getEpochInfo"}'

# at publication this returned:
# epoch 1003, slotsInEpoch 432000,
# transactionCount 530,025,255,192

That last number is worth pausing on: 530 billion cumulative transactions (consensus votes included, which is most of them, since every validator votes every slot). The user-initiated slice is tracked separately: June 2026 alone saw 3.77 billion non-vote transactions, an all-time monthly record.

The honest outage ledger

Eight incidents took the network fully or effectively down between genesis and today, totalling roughly 64.5 hours. The pattern in the list is more interesting than the list itself: nearly every early outage was some form of "the ingest path trusted the outside world too much," and each one bought a structural fix.

  • Dec 2020, ~6 h: block propagation bug partitioned the network.
  • Sept 14, 2021, 17 h: the Grape Protocol IDO. Bots flooded 300k+ TPS at raw UDP ingest and consensus stalled. The famous one.
  • Jan 2022: days of degraded performance from duplicate transaction spam (no full halt).
  • Apr 30, 2022, 8 h: Candy Machine mint bots at six million requests per second.
  • June 2022, 4.5 h: durable nonce bug.
  • Sept 2022, 8.5 h: fork-choice bug on duplicate blocks.
  • Feb 25, 2023, ~19 h: an oversized block (150,000 shreds) overwhelmed deduplication. The longest outage.
  • Feb 6, 2024, ~5 h: an infinite JIT recompile loop triggered by legacy loader programs. The last one.

The 2021-2022 spam era produced the trio every Solana developer now takes for granted: QUIC replaced raw UDP for transaction ingestion, stake-weighted QoS made connection capacity proportional to stake, and local fee markets gave contended state a price (see our priority fees guide). August 2024's coordinated patch of an ELF-alignment vulnerability is the counterpoint worth noting: a critical bug handled across the validator set with zero downtime.

Since February 6, 2024: 885 days of continuous uptime at the epoch 1000 mark, by far the longest streak in network history (the previous best was under a year, and genesis-to-first-outage was 547 days).

The milestone in numbers

text
Genesis                    March 16, 2020
Epoch 1000                 July 10, 2026, 04:11 UTC
Days elapsed               2,307
Slots per epoch            432,000 (fixed)
Cumulative transactions    530B+ (incl. votes, live RPC)
June 2026 non-vote txs     3.77B (all-time monthly record)
Outages                    8 (~64.5 h total)
Uptime streak              885 days and counting
Active validators          ~700 (live RPC, July 17, 2026)
Active stake               ~425M SOL
Longest-serving validator  Laine: 784 consecutive epochs

Two caveats on validator counts, because published numbers disagree: aggregators were quoting anywhere from 791 to 906 active validators this summer depending on methodology. The figure above is what mainnet RPC's getVoteAccounts returned at publication: 699 current, non-delinquent voting accounts. Stake concentration is the better decentralization lens anyway: a Nakamoto coefficient around 19, with no single validator holding much over 3% of stake.

What changed under the chain along the way

The chain never reset, but almost everything producing it has been replaced. Token Extensions shipped as the flagship of early 2024 (our field guide). Frankendancer, the hybrid of Firedancer's networking stack with the Agave runtime, went live on mainnet at Breakpoint in September 2024, and full Firedancer followed on December 12, 2025, running on a fifth or more of the validator set by mid-2026. Client diversity was Anatoly Yakovenko's long-stated bar for finally dropping the "beta" from mainnet-beta; the tag is technically still there (the RPC endpoint is literally api.mainnet-beta.solana.com), which at this point says more about naming inertia than network maturity.

The next thousand epochs start with the biggest consensus change in the network's history: Alpenglow, which passed its governance vote in September 2025 with 98.27% of participating stake in favor, went live on a community test cluster in May, and targets mainnet in Q3 2026 with ~150 ms finality. Epoch boundaries themselves will keep doing exactly what they have done 1,003 times now. That reliability took six years and eight hard lessons to earn.

Resources

Keep reading

Get new articles in your inbox

Technical deep-dives on Solana tooling, infrastructure, and ecosystem. No noise.