NEAR is an L1 with two fast-growing businesses behind its token. Revenue growth, combined with buybacks, burns and staking, makes it a solid long-term investment. The two businesses are NEAR Intents, a cross-chain swap network, and NEAR AI, a private AI cloud. Intents brings the revenue: it now earns 95% of NEAR's protocol revenue, against 34% a year earlier, and in Oct 2026 that revenue was 5.2× its level a year before. NEAR AI is earlier and publishes no revenue yet, but its usage is growing fast: it served 225.2B tokens on OpenRouter in the week to Nov 8, 2026, 16× its first full week. Burns and buybacks tie that revenue to the token: L1 fees are burned and Intents fees buy NEAR back, and together they offset 19.5% of new NEAR in 2026 Q3, against 1.8% a year earlier. If Intents revenue keeps growing, even at a fading rate, they would offset all of it by Dec 2028. Staking locks NEAR up: on Nov 14, 2026, 45.0% of all NEAR was staked to secure the L1, and 1.5M NEAR was staked with NEAR AI, which opened staking for AI credits on Jul 30, 2026. New supply is also falling: the cap on new issuance has been cut from 5% to 2.5% a year. We are long-term bullish. This report sets out the evidence for that view, the evidence against it, and what would change our mind.
The problem NEAR solves
NEAR was built as an L1: a base-layer blockchain designed to be fast and cheap, and to scale by splitting its work across shards. Fees still average about 1 cent a transaction. It found early traction. Apps such as Sweat, HOT and KaiChing took off on it, and in Jan 2025 the chain carried 7.2M transactions a day.
But blockspace has become a commodity. Many chains now offer fast, cheap transactions, and cheap transactions earn little: in Oct 2026, all of NEAR's L1 gas fees came to $223.6K. The traffic did not last either. One app, KaiChing, sent 65% of all L1 transactions in Mar 2026, and when it stopped the next month, L1 traffic fell with it. By Oct 2026 the L1 carried 729.9K transactions a day.
So NEAR has taken on two more problems: fragmented blockchains and centralised AI.
Fragmented blockchains. Crypto is spread across hundreds of blockchains, each with its own assets and apps. Moving money between them usually takes several steps, a bridge, and a fee at each one. NEAR Intents turns that into one step: a user says what they have and what they want, competing market makers fill the order, and NEAR settles the result. In Oct 2026, Intents handled $3.99B of swaps, 2.3× its level a year earlier. On 1Click, the Intents channel with public swap-by-swap data, 269K users swapped through 352 apps, starting from 37 chains.
Centralised AI. A few companies run the most capable AI models, and they see everything people send them. NEAR AI runs open models inside secure hardware, so no one, not even NEAR AI, can read the prompts. Customers pay per use, or stake NEAR and receive AI credits. It is the newest business. On OpenRouter, a marketplace for AI models, it served 225.2B tokens (units of text) in the week to Nov 8, 2026, 16× its first full week, and Brave's AI assistant has offered a model running on it since Oct 15, 2026.
That makes NEAR three businesses on one token: the L1, NEAR Intents and NEAR AI. The rest of this paper breaks down each business, shows how it contributes to the NEAR token, and explains why we think that makes NEAR a good investment.
NEAR's three layers
NEAR now works as three businesses, each built on the one below it. People, apps and AI agents come in at the top, and every layer connects to the NEAR token.
The layers are built to feed each other. AI agents that need to pay or trade across chains can do it through Intents, and every Intents swap settles on the L1.
Why the NEAR token has value
The token gains when NEAR is taken out of circulation, or locked up, faster than new NEAR is issued. Four mechanisms do that, and each one started or changed in the last 12 months.
Fees and burns. Every L1 transaction pays a fee in NEAR. Until Nov 5, 2026, 30% of each L1 fee went to the contract being called and the rest was burned. Since then, all of it is burned.
Buybacks. NEAR keeps a share of the fees on Intents swaps. Since Feb 23, 2026, that share buys NEAR on the market.
Staking. NEAR is staked with validators to secure the chain: 589.8M NEAR, 45.0% of all NEAR, on Nov 14, 2026. Since Jul 30, 2026, holders can also stake with NEAR AI and receive AI credits. Staked NEAR is locked, not removed, and can be withdrawn.
Less new supply. On Oct 30, 2025, NEAR cut its cap on new issuance from 5% to 2.5% a year. A second cut, HSP-029, passed on Nov 2, 2026 and would lower it to 1.6% by Jan 2029.
How much it adds up to. Burns and buybacks offset 19.5% of the new NEAR issued in 2026 Q3, against 1.8% a year earlier. Counting NEAR newly staked with NEAR AI, the figure reached 43% in Aug 2026 and 45% in Oct 2026. Most new NEAR is still not offset, so the case depends on revenue continuing to grow.
Where it could go. Measured in NEAR, Intents revenue grew at a yearly rate of 116% between 2026 Q1 and 2026 Q3. If that growth halves every 2 years, burns and buybacks offset all new NEAR by Dec 2028. If it holds, they get there by Aug 2028. If revenue stops growing, they do not get there within the 6 years we project. Without HSP-029, the first of those dates moves to Mar 2030. These are projections from a short history, not forecasts. The assumptions are in METHODS.md.
The last year, business by business
This section shows how each of NEAR's three businesses, and the token, changed over the twelve months to Oct 2026. NEAR Intents grew the most. Swap volume rose 2.3×, and its share of native cross-chain swap volume (against THORChain, Chainflip and Maya) rose from 45% to 58% while total DEX volume fell 56%. NEAR AI is new. Its first full week on OpenRouter ended Sep 20, 2026, and staking for AI credits opened on Jul 30, 2026. In Oct 2026 it served 887.2B tokens on OpenRouter and held 1.4M NEAR in stake. The L1 moved the other way. Its own traffic fell after its largest app left, and Intents now makes up 8.9% of its transactions, against 1.6% a year earlier. For the token, burns and buybacks offset 19.7% of new NEAR in Oct 2026, against 4.0% a year earlier.
Business
Measure
Oct 2026
Oct 2025
Change
NEAR Intents
Swap volume
$3.99B
$1.72B
2.3×
NEAR's share of swap fees
$2.5M
$295.8K
8.5×
Distinct 1Click users
269K
132.6K
2.0×
Apps sending swaps through 1Click
352
48
7.3×
Share of native cross-chain swap volume
58.3%
44.5%
+13.8 pts
NEAR L1
Transactions a day
729.9K
4.5M
−84%
Gas fees paid
$223.6K
$304.5K
−27%
Share of transactions that are Intents
8.9%
1.6%
+7.3 pts
NEAR AI
Tokens served on OpenRouter
887.2B
–
new
NEAR staked for AI credits, month end
1.4M
–
new
NEAR token
New NEAR issued
2.8M
5.2M
−47%
Share of new NEAR offset by burns and buybacks
19.7%
4.0%
+15.7 pts
NEAR staked with validators, month end
592M
619.8M
−4%
Share of all NEAR staked, month end
45.2%
48.5%
−3.3 pts
Where NEAR stands today. It is worth being direct about this. Intents is a working, growing business, and it already pays for buybacks. NEAR AI is early: it publishes no revenue, and we estimate its OpenRouter sales at about $975.8K a year. The token is earliest of all, because the mechanisms that send value to it have only just been switched on. But revenue is growing fast, and burns, buybacks and staking connect that growth to the token, which is why our view is long-term bullish.
What would change our mind
This is the test of the thesis. The table takes the thesis line apart: each row is one thing the line claims, with the evidence for it, the evidence against it, and the result that would change our mind. Revenue growth comes first because the other rows lean on it: buybacks only grow if revenue does. Every figure comes from the same data and queries as the charts.
NEAR's protocol revenue (L1 fees burned plus its share of Intents fees) was $2.7M in Oct 2026, 5.2× a year earlier. Intents earned 95% of protocol revenue over Aug 2026–Oct 2026, against 34% a year earlier. None of NEAR AI's revenue is in this figure, because NEAR AI doesn't publish it, so anything it earns comes on top.
The growth is recent: 81% of the twelve-month rise in monthly revenue came in Sep 2026 and Oct 2026. Nearly all of the revenue comes from one business, and swap fees follow trading activity, which can fall as fast as it rose.
Protocol revenue no longer growing over a 12-month span, or NEAR's share of Intents fees being cut.
Burns and buybacks tie that revenue to the token.Chart. Detail: C6.
L1 fees are burned, and since Feb 23, 2026 NEAR's share of Intents fees buys NEAR on the market. Together they offset 19.5% of new NEAR in 2026 Q3, against 1.8% a year earlier. If Intents revenue keeps growing, even at a fading rate, they offset all new NEAR by Dec 2028.
Most new NEAR is still not offset: 80.5% in 2026 Q3. About 53% of the year's rise came from cutting issuance, not from more revenue. The monthly share fell from 26.9% in Sep 2026 to 19.7% in Oct 2026, although Intents revenue rose in dollars: at a higher NEAR price, the same fees buy back less NEAR. The NEAR bought back is held, not burned, and buybacks, the fee split and HSP-029 are governance decisions that can be changed.
The share of new NEAR offset by burns and buybacks no longer rising over a 12-month span, or buybacks being paused, or the NEAR bought back being sold.
NEAR Intents is a fast-growing business.Chart. Detail: C3, C4.
Intents handled $3.99B of swaps in Oct 2026, 2.3× a year earlier, while total DEX volume fell 56%. Its share of native cross-chain swap volume rose from 45% to 58%. On 1Click, the channel with public swap-by-swap data (about 27% of Intents volume), 269K users swapped that month, against 132.6K a year earlier, and the top 10 made up 15% of volume.
The activity is led by trading and concentrated. On 1Click in Oct 2026, 74% of volume involved a stablecoin and one app (swapkit) routed 40% of it. Distinct users fell from 415.7K in Aug 2026 to 269K. Flow like this can leave as quickly as it came.
Intents swap volume or 1Click users no longer growing over a 12-month span, or the top 10 users' share of 1Click volume rising above half.
NEAR AI is a fast-growing business.Chart. Detail: C5.
NEAR AI served 225.2B tokens on OpenRouter in the week to Nov 8, 2026, 16× its first full week. Brave's AI assistant has offered a model running on it since Oct 15, 2026.
NEAR AI is early and unproven. Its first full week on OpenRouter ended Sep 20, 2026, and the latest week was below the peak week (320.7B). It publishes neither revenue nor total usage. Our estimate from OpenRouter, one of its sales channels, is about $975.8K a year, less than NEAR's protocol revenue in Oct 2026 alone. No published mechanism sends that money to the token.
Weekly tokens on OpenRouter falling for a sustained period, or NEAR AI still publishing no revenue a year from now.
45.2% of all NEAR (592M NEAR) was staked with validators at the end of Oct 2026. At every month end since Jan 2025 the share has been between 40.7% and 48.5%. 1.5M NEAR was staked with NEAR AI, up from 10K when staking for AI credits opened on Jul 30, 2026.
Staked NEAR is locked, not removed: it can be withdrawn. The staked share is not rising: it fell from 48.5% to 45.2% between the end of Oct 2025 and the end of Oct 2026, and the amount staked fell from 619.8M to 592M NEAR. Staking with NEAR AI is small, 0.2% of all staked NEAR, and concentrated: the largest staker holds 73% of it and the top ten 95%.
The staked share falling for a second year running, or NEAR staked with NEAR AI shrinking.
The C-numbers in the first column link to the full evidence and caveats behind each row in METHODS.md, which also lists what we couldn't measure.
Sources and methods
Data as of Nov 14, 2026. This is research, not investment advice.
DASHBOARD.md: every chart, the query behind it, and one plain sentence on what it shows. The same page with hover readouts and data tables is site/index.html.
METHODS.md: how each number is computed, each claim's evidence and caveats in detail, the scenario assumptions, and every source with its date.
README.md: how to rebuild everything with make charts and check it against the live sources with make verify.
Dashboard · NEAR HACKATHON, Investment Thesis track
NEAR thesis dashboard
NEAR is an L1 with two fast-growing businesses behind its token. Revenue growth, combined with buybacks, burns and staking, makes it a solid long-term investment.
Every chart behind the thesis, with one plain sentence on what it shows and the query behind it. The sections follow the thesis table: one per claim, with the result that would change our mind. Data as of Nov 14, 2026. The argument is in THESIS.md and the methods in METHODS.md. Hover a chart for values, or open its table.
The thesis in two pictures
Who uses NEAR, what each layer does and how each one reaches the token; then what each layer earned in the latest complete month and how much of it reached the token.
NEAR works as three layers on one token: NEAR AI served 225.2B tokens on OpenRouter in its latest week, NEAR Intents handled $3.99B of swaps in Oct 2026, and the NEAR L1 carried 729.9K transactions a day.
The diagram shows how the layers stack, not how money moves; that is the next chart. Validator stake is one snapshot, on Nov 14, 2026. User, app and chain counts cover the 1Click channel only.
Query behind this chart: NEAR BigQuery SQL · OpenRouter · revenue.near.org · DefiLlama · NEAR Intents Explorer API · NEAR RPC
queries/l1_activity.sql (NEAR BigQuery)
-- L1 activity since 2025-01-01, from one scan of the transactions table (two result shapes):
-- level = 'daily': one row per day with total transactions, distinct signers (proxy for
-- daily active accounts), transactions sent directly to intents.near,
-- and NEAR burnt converting transactions to receipts.
-- level = 'receiver': one row per month and receiving account (top 25 per month), to see
-- which apps drive L1 activity and which large app left in early 2026.
-- GROUPING SETS lets both come from a single pass over four columns, which roughly halves the
-- bytes scanned compared with two separate queries. Longer L1 history (2020+) comes from NearBlocks.
-- Used for: C2 and the L1 layer in Part 2.
WITH grouped AS (
SELECT
IF(GROUPING(block_date) = 0, 'daily', 'receiver') AS level,
block_date,
month,
receiver_account_id,
COUNT(*) AS transactions,
COUNT(DISTINCT signer_account_id) AS signers,
COUNTIF(receiver_account_id = 'intents.near') AS tx_to_intents,
COUNT(DISTINCT IF(receiver_account_id = 'intents.near', signer_account_id, NULL)) AS signers_to_intents,
SUM(receipt_conversion_tokens_burnt) / 1e24 AS tx_conversion_burnt_near
FROM (
SELECT block_date, DATE_TRUNC(block_date, MONTH) AS month, receiver_account_id,
signer_account_id, receipt_conversion_tokens_burnt
FROM `bigquery-public-data.crypto_near_mainnet_us.transactions`
WHERE block_date >= '2025-01-01'
)
GROUP BY GROUPING SETS ((block_date), (month, receiver_account_id))
)
SELECT *
FROM grouped
WHERE TRUE
QUALIFY level = 'daily'
OR ROW_NUMBER() OVER (PARTITION BY level, month ORDER BY transactions DESC) <= 25
ORDER BY level, block_date, month, transactions DESC
OpenRouter
GET https://openrouter.ai/provider/near-ai (tokens per model per day)
GET https://openrouter.ai/api/v1/models/<model>/endpoints (NEAR AI's price per token)
revenue.near.org
GET https://revenue.near.org/ (the page embeds its data; fetch/revenue_tracker.py extracts the monthly revenue series and the daily fees-and-volume series)
DefiLlama
GET https://api.llama.fi/overview/dexs (daily volume per protocol: near-intents, thorchain-dex,
chainflip-amm, maya-protocol, and the total)
GET https://api.llama.fi/overview/bridge-aggregators
NEAR Intents Explorer API
GET https://explorer.near-intents.org/api/v0/transactions
?statuses=SUCCESS&numberOfTransactions=1000&direction=next
&lastDepositAddress=…&lastDepositMemo=… (cursor; Authorization: Bearer <JWT>)
&startTimestampUnix=…&endTimestampUnix=… (time-window refill of skipped ranges)
NEAR RPC
POST https://rpc.mainnet.near.org
{"method": "validators", "params": [null]} (current validators and their stake)
{"method": "block", "params": {"finality": "final"}} (total supply, from the block header)
Table: the numbers in this chart
Layer
What the diagram says
NEAR AI
Runs open AI models in secure hardware, so prompts stay private
NEAR AI
225.2B tokens served on OpenRouter in the week to Nov 8, 2026
NEAR AI
16× the tokens of its first full week on OpenRouter
NEAR AI
Brave's AI assistant Leo has offered a model on it since Oct 15, 2026
NEAR AI
1.5M NEAR staked for AI credits on Nov 14, 2026
NEAR Intents
Swaps between any two chains in one step
NEAR Intents
$3.99B of swaps in Oct 2026, 2.3× the level a year earlier
NEAR Intents
58% of native cross-chain swap volume, against 45% a year earlier
NEAR Intents
On 1Click: 269K users, 352 apps and 37 starting chains in Oct 2026
NEAR Intents
95% of NEAR's protocol revenue over Aug 2026–Oct 2026
NEAR L1
The blockchain where swaps settle and NEAR is staked
NEAR L1
729.9K transactions a day in Oct 2026, at about 1 cent each
NEAR L1
Intents is 8.9% of transactions in Oct 2026, against 1.6% a year earlier
NEAR L1
589.8M NEAR staked with 355 validators, 45.0% of supply (Nov 14, 2026)
NEAR L1
All gas fees are burned since Nov 5, 2026
Follow the money: what each layer earned in Oct 2026, and what reached the token
In Oct 2026, burns and buybacks offset 544.3K NEAR (19.7% of the NEAR issued that month) and staking for NEAR AI locked another 687.3K NEAR, while 45.0% of all NEAR was staked with validators on Nov 14, 2026.
Latest month with complete data for all three layers. NEAR AI sales are an estimate: OpenRouter tokens at NEAR AI's list prices. Staked NEAR is locked, not removed, and can be withdrawn. Total stake is one snapshot of the validator set on Nov 14, 2026; its month-end history is the chart "Share of all NEAR staked with validators".
Query behind this chart: NEAR BigQuery SQL · revenue.near.org · OpenRouter · NEAR RPC · NearBlocks
queries/supply_flows_daily.sql (NEAR BigQuery)
-- Daily NEAR burned and NEAR minted since 2025-01-01, from block-to-block changes in total supply.
-- Within an epoch, total supply only falls: by the NEAR burned in each block (70% of L1 gas fees;
-- the other 30% is paid to contract developers and stays in supply). At each epoch boundary it
-- jumps up by the newly minted NEAR (validator rewards, i.e. inflation). So:
-- burned = sum of block-to-block decreases, minted = sum of increases.
-- Reads two numeric columns from the 32 GB blocks table, far cheaper than summing fees from the
-- 9.8 TB execution_outcomes table (173 GB). Differences are taken within each calendar month, so one
-- block-to-block change per month (the first block of the month) is not counted.
-- Used for: C1 (L1 side of the revenue mix: burned NEAR) and C6 (burn and issuance in the capture ratio).
WITH b AS (
SELECT
block_date,
total_supply - LAG(total_supply) OVER (PARTITION BY DATE_TRUNC(block_date, MONTH) ORDER BY block_height) AS d
FROM `bigquery-public-data.crypto_near_mainnet_us.blocks`
WHERE block_date >= '2025-01-01'
)
SELECT
block_date,
SUM(IF(d < 0, -d, 0)) / 1e24 AS burned_near,
SUM(IF(d > 0, d, 0)) / 1e24 AS minted_near,
COUNTIF(d > 0) AS mint_blocks
FROM b
GROUP BY block_date
ORDER BY block_date
queries/l1_activity.sql (NEAR BigQuery)
-- L1 activity since 2025-01-01, from one scan of the transactions table (two result shapes):
-- level = 'daily': one row per day with total transactions, distinct signers (proxy for
-- daily active accounts), transactions sent directly to intents.near,
-- and NEAR burnt converting transactions to receipts.
-- level = 'receiver': one row per month and receiving account (top 25 per month), to see
-- which apps drive L1 activity and which large app left in early 2026.
-- GROUPING SETS lets both come from a single pass over four columns, which roughly halves the
-- bytes scanned compared with two separate queries. Longer L1 history (2020+) comes from NearBlocks.
-- Used for: C2 and the L1 layer in Part 2.
WITH grouped AS (
SELECT
IF(GROUPING(block_date) = 0, 'daily', 'receiver') AS level,
block_date,
month,
receiver_account_id,
COUNT(*) AS transactions,
COUNT(DISTINCT signer_account_id) AS signers,
COUNTIF(receiver_account_id = 'intents.near') AS tx_to_intents,
COUNT(DISTINCT IF(receiver_account_id = 'intents.near', signer_account_id, NULL)) AS signers_to_intents,
SUM(receipt_conversion_tokens_burnt) / 1e24 AS tx_conversion_burnt_near
FROM (
SELECT block_date, DATE_TRUNC(block_date, MONTH) AS month, receiver_account_id,
signer_account_id, receipt_conversion_tokens_burnt
FROM `bigquery-public-data.crypto_near_mainnet_us.transactions`
WHERE block_date >= '2025-01-01'
)
GROUP BY GROUPING SETS ((block_date), (month, receiver_account_id))
)
SELECT *
FROM grouped
WHERE TRUE
QUALIFY level = 'daily'
OR ROW_NUMBER() OVER (PARTITION BY level, month ORDER BY transactions DESC) <= 25
ORDER BY level, block_date, month, transactions DESC
-- Daily (per-epoch) NEAR balances of the wallets that receive NEAR Intents revenue,
-- as listed on revenue.near.org. The table is clustered by account_id, so this is cheap.
-- Used for: C6 (buybacks held vs burned) and cross-checking the revenue tracker.
SELECT
epoch_date,
account_id,
liquid,
staked,
unstaked_not_liquid,
reward
FROM `bigquery-public-data.crypto_near_mainnet_us.ft_balances_daily`
WHERE account_id IN (
'buybacks.multisignature.near',
'fefundsadmin.sputnik-dao.near',
'1csfundsadmin.sputnik-dao.near'
)
AND epoch_date >= '2025-01-01' -- partition filter (this table has epoch_date, not block_date)
ORDER BY account_id, epoch_date
revenue.near.org
GET https://revenue.near.org/ (the page embeds its data; fetch/revenue_tracker.py extracts the monthly revenue series and the daily fees-and-volume series)
OpenRouter
GET https://openrouter.ai/provider/near-ai (tokens per model per day)
GET https://openrouter.ai/api/v1/models/<model>/endpoints (NEAR AI's price per token)
POST https://rpc.mainnet.near.org
{"method": "validators", "params": [null]} (current validators and their stake)
{"method": "block", "params": {"finality": "final"}} (total supply, from the block header)
NearBlocks
GET https://api.nearblocks.io/v1/charts (daily NEAR price)
Table: the numbers in this chart
Step
Oct 2026
Intents swap volume
$3.99B
Intents fees paid by users
$6.39M
Fee rate
0.160%
NEAR's share of fees (USD)
$2.5M
NEAR's share of fees (% of fees)
39.2%
NEAR's share of fees (NEAR)
512.86K NEAR
Buyback wallet, change over the month
297.36K NEAR
L1 transactions per day (average)
729.93K
L1 gas fees (USD)
$223.61K
NEAR burned by L1 gas fees
31.46K NEAR
Total NEAR supply, Nov 14, 2026
1.31B NEAR
NEAR staked with validators, Nov 14, 2026
589.77M NEAR
Share of all NEAR staked, Nov 14, 2026
45.0%
Validators, Nov 14, 2026
355
NEAR AI tokens served on OpenRouter
887.24B
NEAR AI inference sales on OpenRouter (estimate, list prices)
$79K
NEAR staked with NEAR AI's validator, month end
1.4M NEAR
NEAR newly staked for NEAR AI (locked)
687.33K NEAR
NEAR offset by burns and buybacks (L1 burn + Intents revenue)
544.32K NEAR
NEAR issued
2.76M NEAR
Share of new NEAR offset by burns and buybacks
19.7%
The same share, counting new AI stake
44.6%
1. Revenue is growing
Where NEAR's protocol revenue comes from, month by month.
What would change our mind: Protocol revenue no longer growing over a 12-month span, or NEAR's share of Intents fees being cut. The full table is in THESIS.md.
NEAR's protocol revenue was $2.7M in Oct 2026, 5.2× a year earlier: $2.5M from its share of Intents fees and $157K from L1 fees burned.
L1 side: NEAR burned by gas fees, valued at the daily price. Gross L1 fees = burn ÷ 0.7 before Nov 5, 2026 (when 30% of gas went to the called contract) and = burn after, when all gas is burned. Intents side: NEAR's net revenue after partner payouts. Complete months only.
Query behind this chart: NEAR BigQuery SQL · revenue.near.org · NearBlocks
queries/supply_flows_daily.sql (NEAR BigQuery)
-- Daily NEAR burned and NEAR minted since 2025-01-01, from block-to-block changes in total supply.
-- Within an epoch, total supply only falls: by the NEAR burned in each block (70% of L1 gas fees;
-- the other 30% is paid to contract developers and stays in supply). At each epoch boundary it
-- jumps up by the newly minted NEAR (validator rewards, i.e. inflation). So:
-- burned = sum of block-to-block decreases, minted = sum of increases.
-- Reads two numeric columns from the 32 GB blocks table, far cheaper than summing fees from the
-- 9.8 TB execution_outcomes table (173 GB). Differences are taken within each calendar month, so one
-- block-to-block change per month (the first block of the month) is not counted.
-- Used for: C1 (L1 side of the revenue mix: burned NEAR) and C6 (burn and issuance in the capture ratio).
WITH b AS (
SELECT
block_date,
total_supply - LAG(total_supply) OVER (PARTITION BY DATE_TRUNC(block_date, MONTH) ORDER BY block_height) AS d
FROM `bigquery-public-data.crypto_near_mainnet_us.blocks`
WHERE block_date >= '2025-01-01'
)
SELECT
block_date,
SUM(IF(d < 0, -d, 0)) / 1e24 AS burned_near,
SUM(IF(d > 0, d, 0)) / 1e24 AS minted_near,
COUNTIF(d > 0) AS mint_blocks
FROM b
GROUP BY block_date
ORDER BY block_date
revenue.near.org
GET https://revenue.near.org/ (the page embeds its data; fetch/revenue_tracker.py extracts the monthly revenue series and the daily fees-and-volume series)
NearBlocks
GET https://api.nearblocks.io/v1/charts (daily NEAR price)
Query behind this chart: NEAR BigQuery SQL · revenue.near.org · NearBlocks
queries/supply_flows_daily.sql (NEAR BigQuery)
-- Daily NEAR burned and NEAR minted since 2025-01-01, from block-to-block changes in total supply.
-- Within an epoch, total supply only falls: by the NEAR burned in each block (70% of L1 gas fees;
-- the other 30% is paid to contract developers and stays in supply). At each epoch boundary it
-- jumps up by the newly minted NEAR (validator rewards, i.e. inflation). So:
-- burned = sum of block-to-block decreases, minted = sum of increases.
-- Reads two numeric columns from the 32 GB blocks table, far cheaper than summing fees from the
-- 9.8 TB execution_outcomes table (173 GB). Differences are taken within each calendar month, so one
-- block-to-block change per month (the first block of the month) is not counted.
-- Used for: C1 (L1 side of the revenue mix: burned NEAR) and C6 (burn and issuance in the capture ratio).
WITH b AS (
SELECT
block_date,
total_supply - LAG(total_supply) OVER (PARTITION BY DATE_TRUNC(block_date, MONTH) ORDER BY block_height) AS d
FROM `bigquery-public-data.crypto_near_mainnet_us.blocks`
WHERE block_date >= '2025-01-01'
)
SELECT
block_date,
SUM(IF(d < 0, -d, 0)) / 1e24 AS burned_near,
SUM(IF(d > 0, d, 0)) / 1e24 AS minted_near,
COUNTIF(d > 0) AS mint_blocks
FROM b
GROUP BY block_date
ORDER BY block_date
revenue.near.org
GET https://revenue.near.org/ (the page embeds its data; fetch/revenue_tracker.py extracts the monthly revenue series and the daily fees-and-volume series)
NearBlocks
GET https://api.nearblocks.io/v1/charts (daily NEAR price)
Table: the numbers in this chart
Month
Intents share (net)
Intents share (gross fees)
Jan 2025
0.0%
0.0%
Feb 2025
0.0%
0.0%
Mar 2025
0.0%
0.0%
Apr 2025
0.0%
0.0%
May 2025
0.0%
0.0%
Jun 2025
0.0%
25.6%
Jul 2025
0.0%
40.6%
Aug 2025
10.5%
55.5%
Sep 2025
11.2%
68.8%
Oct 2025
58.1%
89.1%
Nov 2025
78.7%
96.8%
Dec 2025
70.8%
96.1%
Jan 2026
82.9%
96.6%
Feb 2026
87.0%
97.8%
Mar 2026
82.3%
96.7%
Apr 2026
94.1%
98.9%
May 2026
95.6%
98.3%
Jun 2026
95.3%
98.0%
Jul 2026
94.6%
98.4%
Aug 2026
96.0%
98.7%
Sep 2026
96.1%
98.3%
Oct 2026
94.1%
96.6%
2. Burns and buybacks tie that revenue to the token
How much of each month's new NEAR is offset by L1 fees burned and by NEAR bought with Intents revenue, when each mechanism was switched on, where the NEAR from Intents fees sits, and where the trend could go.
What would change our mind: The share of new NEAR offset by burns and buybacks no longer rising over a 12-month span, or buybacks being paused, or the NEAR bought back being sold. The full table is in THESIS.md.
Share of new NEAR offset by burns and buybacks, monthly
The share of new NEAR offset by burns and buybacks rose from about 2% a quarter in 2025 Q1–Q3 to 19.5% in 2026 Q3.
Share of new NEAR offset = (L1 fees burned + NEAR's share of Intents fees, in NEAR) ÷ NEAR issued. METHODS.md calls this the capture ratio. The table shows complete quarters.
Query behind this chart: NEAR BigQuery SQL · revenue.near.org · NEAR RPC
queries/supply_flows_daily.sql (NEAR BigQuery)
-- Daily NEAR burned and NEAR minted since 2025-01-01, from block-to-block changes in total supply.
-- Within an epoch, total supply only falls: by the NEAR burned in each block (70% of L1 gas fees;
-- the other 30% is paid to contract developers and stays in supply). At each epoch boundary it
-- jumps up by the newly minted NEAR (validator rewards, i.e. inflation). So:
-- burned = sum of block-to-block decreases, minted = sum of increases.
-- Reads two numeric columns from the 32 GB blocks table, far cheaper than summing fees from the
-- 9.8 TB execution_outcomes table (173 GB). Differences are taken within each calendar month, so one
-- block-to-block change per month (the first block of the month) is not counted.
-- Used for: C1 (L1 side of the revenue mix: burned NEAR) and C6 (burn and issuance in the capture ratio).
WITH b AS (
SELECT
block_date,
total_supply - LAG(total_supply) OVER (PARTITION BY DATE_TRUNC(block_date, MONTH) ORDER BY block_height) AS d
FROM `bigquery-public-data.crypto_near_mainnet_us.blocks`
WHERE block_date >= '2025-01-01'
)
SELECT
block_date,
SUM(IF(d < 0, -d, 0)) / 1e24 AS burned_near,
SUM(IF(d > 0, d, 0)) / 1e24 AS minted_near,
COUNTIF(d > 0) AS mint_blocks
FROM b
GROUP BY block_date
ORDER BY block_date
revenue.near.org
GET https://revenue.near.org/ (the page embeds its data; fetch/revenue_tracker.py extracts the monthly revenue series and the daily fees-and-volume series)
NEAR offset by burns and buybacks rose to 544.3K in Oct 2026 while NEAR issued halved to about 2.7M a month, but issuance is still the larger number.
Issued and burned come from block-to-block changes in total supply. New AI stake is the month's net change in NEAR staked with nearai.pool.near: locked, not removed, and withdrawable.
Query behind this chart: NEAR BigQuery SQL · revenue.near.org · NEAR RPC
queries/supply_flows_daily.sql (NEAR BigQuery)
-- Daily NEAR burned and NEAR minted since 2025-01-01, from block-to-block changes in total supply.
-- Within an epoch, total supply only falls: by the NEAR burned in each block (70% of L1 gas fees;
-- the other 30% is paid to contract developers and stays in supply). At each epoch boundary it
-- jumps up by the newly minted NEAR (validator rewards, i.e. inflation). So:
-- burned = sum of block-to-block decreases, minted = sum of increases.
-- Reads two numeric columns from the 32 GB blocks table, far cheaper than summing fees from the
-- 9.8 TB execution_outcomes table (173 GB). Differences are taken within each calendar month, so one
-- block-to-block change per month (the first block of the month) is not counted.
-- Used for: C1 (L1 side of the revenue mix: burned NEAR) and C6 (burn and issuance in the capture ratio).
WITH b AS (
SELECT
block_date,
total_supply - LAG(total_supply) OVER (PARTITION BY DATE_TRUNC(block_date, MONTH) ORDER BY block_height) AS d
FROM `bigquery-public-data.crypto_near_mainnet_us.blocks`
WHERE block_date >= '2025-01-01'
)
SELECT
block_date,
SUM(IF(d < 0, -d, 0)) / 1e24 AS burned_near,
SUM(IF(d > 0, d, 0)) / 1e24 AS minted_near,
COUNTIF(d > 0) AS mint_blocks
FROM b
GROUP BY block_date
ORDER BY block_date
revenue.near.org
GET https://revenue.near.org/ (the page embeds its data; fetch/revenue_tracker.py extracts the monthly revenue series and the daily fees-and-volume series)
Issuance was halved in Oct 2025, and buybacks, staking for NEAR AI and burning all L1 gas have all switched on since.
Issuance cap and gas-burn dates come from NEAR's protocol config at past blocks. The HSP-029 start date is an assumption: vote end plus the 90-day grace period.
Query behind this chart: NEAR BigQuery SQL · NEAR archival RPC · config.yaml
queries/supply_flows_daily.sql (NEAR BigQuery)
-- Daily NEAR burned and NEAR minted since 2025-01-01, from block-to-block changes in total supply.
-- Within an epoch, total supply only falls: by the NEAR burned in each block (70% of L1 gas fees;
-- the other 30% is paid to contract developers and stays in supply). At each epoch boundary it
-- jumps up by the newly minted NEAR (validator rewards, i.e. inflation). So:
-- burned = sum of block-to-block decreases, minted = sum of increases.
-- Reads two numeric columns from the 32 GB blocks table, far cheaper than summing fees from the
-- 9.8 TB execution_outcomes table (173 GB). Differences are taken within each calendar month, so one
-- block-to-block change per month (the first block of the month) is not counted.
-- Used for: C1 (L1 side of the revenue mix: burned NEAR) and C6 (burn and issuance in the capture ratio).
WITH b AS (
SELECT
block_date,
total_supply - LAG(total_supply) OVER (PARTITION BY DATE_TRUNC(block_date, MONTH) ORDER BY block_height) AS d
FROM `bigquery-public-data.crypto_near_mainnet_us.blocks`
WHERE block_date >= '2025-01-01'
)
SELECT
block_date,
SUM(IF(d < 0, -d, 0)) / 1e24 AS burned_near,
SUM(IF(d > 0, d, 0)) / 1e24 AS minted_near,
COUNTIF(d > 0) AS mint_blocks
FROM b
GROUP BY block_date
ORDER BY block_date
NEAR archival RPC
POST https://archival-rpc.mainnet.near.org
{"method": "EXPERIMENTAL_protocol_config", "params": {"block_id": <first block of the day>}}
(bisected by day to date each change in max_inflation_rate and burnt_gas_reward)
config.yaml
Scenario inputs: issuance paths, growth window, half-life (no request)
Table: the numbers in this chart
Month
Issuance, actual (annualised)
HSP-029 path
Jan 2025
4.99%
–
Feb 2025
5.01%
–
Mar 2025
4.95%
–
Apr 2025
4.98%
–
May 2025
5.04%
–
Jun 2025
4.98%
–
Jul 2025
4.97%
–
Aug 2025
5.01%
–
Sep 2025
5.00%
–
Oct 2025
4.79%
–
Nov 2025
2.47%
–
Dec 2025
2.48%
–
Jan 2026
2.50%
–
Feb 2026
2.48%
–
Mar 2026
2.50%
–
Apr 2026
2.49%
–
May 2026
2.50%
–
Jun 2026
2.48%
–
Jul 2026
2.48%
–
Aug 2026
2.49%
–
Sep 2026
2.49%
–
Oct 2026
2.49%
2.50%
Nov 2026
–
2.50%
Dec 2026
–
2.50%
Jan 2027
–
2.50%
Feb 2027
–
2.46%
Mar 2027
–
2.43%
Apr 2027
–
2.39%
May 2027
–
2.35%
Jun 2027
–
2.31%
Jul 2027
–
2.27%
Aug 2027
–
2.24%
Sep 2027
–
2.20%
Oct 2027
–
2.16%
Nov 2027
–
2.12%
Dec 2027
–
2.09%
Jan 2028
–
2.05%
Feb 2028
–
2.01%
Mar 2028
–
1.98%
Apr 2028
–
1.94%
May 2028
–
1.90%
Jun 2028
–
1.86%
Jul 2028
–
1.82%
Aug 2028
–
1.79%
Sep 2028
–
1.75%
Oct 2028
–
1.71%
Nov 2028
–
1.68%
Dec 2028
–
1.64%
Jan 2029
–
1.60%
Feb 2029
–
1.60%
Mar 2029
–
1.60%
Apr 2029
–
1.60%
May 2029
–
1.60%
Jun 2029
–
1.60%
Jul 2029
–
1.60%
Where the NEAR from Intents fees sits: revenue-wallet balances, month end
-- Daily (per-epoch) NEAR balances of the wallets that receive NEAR Intents revenue,
-- as listed on revenue.near.org. The table is clustered by account_id, so this is cheap.
-- Used for: C6 (buybacks held vs burned) and cross-checking the revenue tracker.
SELECT
epoch_date,
account_id,
liquid,
staked,
unstaked_not_liquid,
reward
FROM `bigquery-public-data.crypto_near_mainnet_us.ft_balances_daily`
WHERE account_id IN (
'buybacks.multisignature.near',
'fefundsadmin.sputnik-dao.near',
'1csfundsadmin.sputnik-dao.near'
)
AND epoch_date >= '2025-01-01' -- partition filter (this table has epoch_date, not block_date)
ORDER BY account_id, epoch_date
revenue.near.org
GET https://revenue.near.org/ (the page embeds its data; fetch/revenue_tracker.py extracts the monthly revenue series and the daily fees-and-volume series)
Table: the numbers in this chart
Month
1csfundsadmin.sputnik-dao.near
buybacks.multisignature.near
fefundsadmin.sputnik-dao.near
Total held
Cumulative Intents revenue
Nov 2025
0 NEAR
0 NEAR
319.05K NEAR
319.05K NEAR
445.52K NEAR
Dec 2025
0 NEAR
0 NEAR
427.52K NEAR
427.52K NEAR
619.67K NEAR
Jan 2026
0 NEAR
0 NEAR
634.33K NEAR
634.33K NEAR
950.01K NEAR
Feb 2026
182.98K NEAR
0 NEAR
1.04M NEAR
1.22M NEAR
1.35M NEAR
Mar 2026
404.45K NEAR
0 NEAR
1.27M NEAR
1.67M NEAR
1.66M NEAR
Apr 2026
510.07K NEAR
34.65K NEAR
1.28M NEAR
1.82M NEAR
1.96M NEAR
May 2026
510.07K NEAR
262.59K NEAR
1.37M NEAR
2.15M NEAR
2.4M NEAR
Jun 2026
510.07K NEAR
486.48K NEAR
1.44M NEAR
2.44M NEAR
2.84M NEAR
Jul 2026
502.8K NEAR
662.72K NEAR
1.52M NEAR
2.69M NEAR
3.18M NEAR
Aug 2026
502.8K NEAR
940.84K NEAR
1.65M NEAR
3.09M NEAR
3.67M NEAR
Sep 2026
271 NEAR
1.85M NEAR
1.81M NEAR
3.66M NEAR
4.37M NEAR
Oct 2026
271 NEAR
2.15M NEAR
1.96M NEAR
4.11M NEAR
4.88M NEAR
Where it could go: share of new NEAR offset by burns and buybacks, projected with HSP-029
With HSP-029 lowering the issuance cap from 2.5% to 1.6% a year, burns and buybacks offset all new NEAR by Dec 2028 if Intents revenue growth halves every 2 years (base), by Aug 2028 if it holds (bull), and not within 6 years if revenue stops growing (bear).
Projections, not findings. Starting point: 2026 Q3's Intents revenue and L1 burn (19.5% of new NEAR offset in Oct 2026). Growth is in NEAR terms, so a rising NEAR price slows it. L1 burn is held at 2026 Q3's level, scaled to 100% burned from Nov 5, 2026. Lines leave the chart above 200%.
Query behind this chart: NEAR BigQuery SQL · config.yaml · revenue.near.org
queries/supply_flows_daily.sql (NEAR BigQuery)
-- Daily NEAR burned and NEAR minted since 2025-01-01, from block-to-block changes in total supply.
-- Within an epoch, total supply only falls: by the NEAR burned in each block (70% of L1 gas fees;
-- the other 30% is paid to contract developers and stays in supply). At each epoch boundary it
-- jumps up by the newly minted NEAR (validator rewards, i.e. inflation). So:
-- burned = sum of block-to-block decreases, minted = sum of increases.
-- Reads two numeric columns from the 32 GB blocks table, far cheaper than summing fees from the
-- 9.8 TB execution_outcomes table (173 GB). Differences are taken within each calendar month, so one
-- block-to-block change per month (the first block of the month) is not counted.
-- Used for: C1 (L1 side of the revenue mix: burned NEAR) and C6 (burn and issuance in the capture ratio).
WITH b AS (
SELECT
block_date,
total_supply - LAG(total_supply) OVER (PARTITION BY DATE_TRUNC(block_date, MONTH) ORDER BY block_height) AS d
FROM `bigquery-public-data.crypto_near_mainnet_us.blocks`
WHERE block_date >= '2025-01-01'
)
SELECT
block_date,
SUM(IF(d < 0, -d, 0)) / 1e24 AS burned_near,
SUM(IF(d > 0, d, 0)) / 1e24 AS minted_near,
COUNTIF(d > 0) AS mint_blocks
FROM b
GROUP BY block_date
ORDER BY block_date
queries/blocks_daily.sql (NEAR BigQuery)
-- Daily block counts, height range and end-of-day total supply, since genesis.
-- Used for: supply history (C6, scenarios) and checking BigQuery completeness against block heights.
SELECT
block_date,
COUNT(*) AS blocks,
MIN(block_height) AS first_height,
MAX(block_height) AS last_height,
ARRAY_AGG(total_supply ORDER BY block_height DESC LIMIT 1)[OFFSET(0)] / 1e24 AS end_of_day_total_supply_near
FROM `bigquery-public-data.crypto_near_mainnet_us.blocks`
WHERE block_date >= '2020-07-21'
GROUP BY block_date
ORDER BY block_date
config.yaml
Scenario inputs: issuance paths, growth window, half-life (no request)
revenue.near.org
GET https://revenue.near.org/ (the page embeds its data; fetch/revenue_tracker.py extracts the monthly revenue series and the daily fees-and-volume series)
Table: the numbers in this chart
Scenario
Revenue growth a year at start (NEAR terms)
Growth then
After 1 year
After 3 years
Breakeven
Bear
0%
–
22.0%
29.2%
not within 6 years
Base
116%
halves every 2 years
40.7%
138%
Dec 2028
Bull
116%
held
43.4%
264%
Aug 2028
Scenarios with and without HSP-029: share of new NEAR offset, projected monthly
If Intents revenue growth halves every 2 years, burns and buybacks offset all new NEAR by Dec 2028 with HSP-029; if growth holds, by Aug 2028; with no growth, not within 6 years.
Projections, not findings. Starting point: 2026 Q3's Intents revenue and L1 burn (19.5% of new NEAR offset in Oct 2026). Growth is in NEAR terms, so a rising NEAR price slows it. L1 burn is held at 2026 Q3's level, scaled to 100% burned from Nov 5, 2026. Lines leave the chart above 200%.
Query behind this chart: NEAR BigQuery SQL · config.yaml · revenue.near.org
queries/supply_flows_daily.sql (NEAR BigQuery)
-- Daily NEAR burned and NEAR minted since 2025-01-01, from block-to-block changes in total supply.
-- Within an epoch, total supply only falls: by the NEAR burned in each block (70% of L1 gas fees;
-- the other 30% is paid to contract developers and stays in supply). At each epoch boundary it
-- jumps up by the newly minted NEAR (validator rewards, i.e. inflation). So:
-- burned = sum of block-to-block decreases, minted = sum of increases.
-- Reads two numeric columns from the 32 GB blocks table, far cheaper than summing fees from the
-- 9.8 TB execution_outcomes table (173 GB). Differences are taken within each calendar month, so one
-- block-to-block change per month (the first block of the month) is not counted.
-- Used for: C1 (L1 side of the revenue mix: burned NEAR) and C6 (burn and issuance in the capture ratio).
WITH b AS (
SELECT
block_date,
total_supply - LAG(total_supply) OVER (PARTITION BY DATE_TRUNC(block_date, MONTH) ORDER BY block_height) AS d
FROM `bigquery-public-data.crypto_near_mainnet_us.blocks`
WHERE block_date >= '2025-01-01'
)
SELECT
block_date,
SUM(IF(d < 0, -d, 0)) / 1e24 AS burned_near,
SUM(IF(d > 0, d, 0)) / 1e24 AS minted_near,
COUNTIF(d > 0) AS mint_blocks
FROM b
GROUP BY block_date
ORDER BY block_date
queries/blocks_daily.sql (NEAR BigQuery)
-- Daily block counts, height range and end-of-day total supply, since genesis.
-- Used for: supply history (C6, scenarios) and checking BigQuery completeness against block heights.
SELECT
block_date,
COUNT(*) AS blocks,
MIN(block_height) AS first_height,
MAX(block_height) AS last_height,
ARRAY_AGG(total_supply ORDER BY block_height DESC LIMIT 1)[OFFSET(0)] / 1e24 AS end_of_day_total_supply_near
FROM `bigquery-public-data.crypto_near_mainnet_us.blocks`
WHERE block_date >= '2020-07-21'
GROUP BY block_date
ORDER BY block_date
config.yaml
Scenario inputs: issuance paths, growth window, half-life (no request)
revenue.near.org
GET https://revenue.near.org/ (the page embeds its data; fetch/revenue_tracker.py extracts the monthly revenue series and the daily fees-and-volume series)
Table: the numbers in this chart
Issuance path
Scenario
Revenue growth a year at start (NEAR terms)
Growth then
After 1 year
After 3 years
Supply after 6 years
Breakeven
Revenue at breakeven vs today
With HSP-029 (2.5% → 1.6%)
bear
0%
–
22.0%
29.2%
1.458B
not within 6 years
–
With HSP-029 (2.5% → 1.6%)
base
116%
halves every 2 years
40.7%
138%
1.458B
Dec 2028
3.7×
With HSP-029 (2.5% → 1.6%)
bull
116%
held
43.4%
264%
1.458B
Aug 2028
4.1×
If HSP-029 isn't implemented (stays 2.5%)
bear
0%
–
19.4%
18.4%
1.521B
not within 6 years
–
If HSP-029 isn't implemented (stays 2.5%)
base
116%
halves every 2 years
35.7%
87.1%
1.521B
Mar 2030
5.8×
If HSP-029 isn't implemented (stays 2.5%)
bull
116%
held
38.2%
166%
1.521B
Jan 2029
5.7×
3. NEAR Intents is a fast-growing business
Cross-chain swaps settled on NEAR: volume, users, where swaps start, and share of the market.
What would change our mind: Intents swap volume or 1Click users no longer growing over a 12-month span, or the top 10 users' share of 1Click volume rising above half. The full table is in THESIS.md.
GET https://api.llama.fi/overview/dexs (daily volume per protocol: near-intents, thorchain-dex,
chainflip-amm, maya-protocol, and the total)
GET https://api.llama.fi/overview/bridge-aggregators
GET https://api.llama.fi/overview/dexs (daily volume per protocol: near-intents, thorchain-dex,
chainflip-amm, maya-protocol, and the total)
GET https://api.llama.fi/overview/bridge-aggregators
Table: the numbers in this chart
Month
NEAR Intents
Bridge aggregators
All DEX
Intents (USD)
Aggregators (USD)
All DEX (USD)
Oct 2025
100
100
100
$1.72B
$2.43B
$588.09B
Nov 2025
212
80
71
$3.66B
$1.95B
$415.62B
Dec 2025
117
71
56
$2.01B
$1.72B
$329.12B
Jan 2026
158
101
62
$2.72B
$2.46B
$364.4B
Feb 2026
138
100
52
$2.38B
$2.44B
$303.46B
Mar 2026
121
81
41
$2.08B
$1.97B
$240.01B
Apr 2026
110
77
34
$1.9B
$1.88B
$202.71B
May 2026
116
68
35
$2B
$1.66B
$204.08B
Jun 2026
137
75
39
$2.35B
$1.83B
$230.88B
Jul 2026
128
84
36
$2.21B
$2.04B
$208.94B
Aug 2026
161
123
41
$2.78B
$2.99B
$241.34B
Sep 2026
279
238
53
$4.8B
$5.78B
$313.19B
Oct 2026
231
213
44
$3.99B
$5.18B
$258.81B
4. NEAR AI is a fast-growing business
Private AI inference: tokens served on OpenRouter, the one sales channel with public usage data.
What would change our mind: Weekly tokens on OpenRouter falling for a sustained period, or NEAR AI still publishing no revenue a year from now. The full table is in THESIS.md.
GET https://openrouter.ai/provider/near-ai (tokens per model per day)
GET https://openrouter.ai/api/v1/models/<model>/endpoints (NEAR AI's price per token)
Table: the numbers in this chart
Week ending
Total tokens
glm-5.3-flash
qwen3.8-27b
gpt-oss-120b
deepseek-v4.1-flash
Est. revenue
Sep 20, 2026
14.43B
14.43B
0
0
0
$1K
Sep 27, 2026
36.58B
36.58B
0
0
0
$3K
Oct 4, 2026
50.48B
50.48B
0
0
0
$5K
Oct 11, 2026
80.87B
66.61B
14.26B
0
0
$8K
Oct 18, 2026
195.77B
127.39B
49.94B
18.44B
0
$17K
Oct 25, 2026
320.71B
225.54B
52.76B
42.41B
0
$29K
Nov 1, 2026
299.83B
194.52B
59.92B
45.21B
183M
$26K
Nov 8, 2026
225.24B
125.75B
64.98B
34.06B
463.19M
$19K
5. Staking locks NEAR up
NEAR staked with validators, and the part of it staked with NEAR AI's validator for AI credits.
What would change our mind: The staked share falling for a second year running, or NEAR staked with NEAR AI shrinking. The full table is in THESIS.md.
Share of all NEAR staked with validators, month end
The share of all NEAR staked with validators fell from 47.3% to 45.2% between the end of Jan 2025 and the end of Oct 2026: the amount staked rose from 582.6M to 592M NEAR while total supply rose from 1.23B to 1.31B.
Staked NEAR is the summed stake of every validator in the epoch that contains the month's last block; the share divides it by total supply at that block. It is the same definition as the dated snapshot in the money-flow diagram. Staked NEAR is locked, not removed, and can be withdrawn. The lowest month end is Sep 2026: 532.8M NEAR was staked, 90.1M less than a month earlier. Five validators held 74.9M less between them (bisontrails2.poolv1.near, p2p-org.poolv1.near, figment.poolv1.near, galaxydigital.poolv1.near and nearvana.poolv1.near). nearvana.poolv1.near was not in the validator set at that month end, and the node lists it as kicked out for producing too few blocks. Between then and the end of Oct 2026 the amount staked rose from 532.8M to 592M NEAR, and the five largest rises came to 74.8M (bitwise_2.poolv1.near, cnear-1.poolv1.near, cnear-3.poolv1.near, cnear-2.poolv1.near and nearvana.poolv1.near). The two amounts are close, which fits stake moving between validators rather than leaving, but we did not trace the accounts behind it.
Query behind this chart: NEAR Data API, then NEAR archival RPC
NEAR Data API, then NEAR archival RPC
GET https://mainnet.neardata.xyz/v0/block/<last block of the month>/headers
(that block's epoch_id and the total supply)
POST https://archival-rpc.mainnet.near.org
{"method": "validators", "params": {"epoch_id": "<epoch_id>"}}
(that epoch's validators and each one's stake; staked NEAR is their sum)
Table: the numbers in this chart
Month end
NEAR staked
Total supply
Share staked
Validators
Jan 2025
582.63M NEAR
1.23B NEAR
47.25%
235
Feb 2025
590.64M NEAR
1.24B NEAR
47.72%
233
Mar 2025
589.19M NEAR
1.24B NEAR
47.41%
251
Apr 2025
586.64M NEAR
1.25B NEAR
47.02%
262
May 2025
577.91M NEAR
1.25B NEAR
46.12%
253
Jun 2025
573M NEAR
1.26B NEAR
45.55%
267
Jul 2025
591.46M NEAR
1.26B NEAR
46.82%
300
Aug 2025
577.59M NEAR
1.27B NEAR
45.53%
299
Sep 2025
574.91M NEAR
1.27B NEAR
45.14%
298
Oct 2025
619.81M NEAR
1.28B NEAR
48.47%
360
Nov 2025
588.94M NEAR
1.28B NEAR
45.96%
380
Dec 2025
594.6M NEAR
1.28B NEAR
46.31%
382
Jan 2026
606.07M NEAR
1.29B NEAR
47.11%
382
Feb 2026
581.87M NEAR
1.29B NEAR
45.14%
398
Mar 2026
586.42M NEAR
1.29B NEAR
45.40%
439
Apr 2026
600.28M NEAR
1.29B NEAR
46.38%
433
May 2026
594.77M NEAR
1.3B NEAR
45.86%
418
Jun 2026
591.46M NEAR
1.3B NEAR
45.51%
418
Jul 2026
585.99M NEAR
1.3B NEAR
44.99%
418
Aug 2026
622.91M NEAR
1.31B NEAR
47.73%
415
Sep 2026
532.82M NEAR
1.31B NEAR
40.74%
417
Oct 2026
592.03M NEAR
1.31B NEAR
45.18%
381
NEAR staked with NEAR AI's validator (nearai.pool.near)
Query behind this chart: NEAR BigQuery SQL · NEAR RPC · NEAR RPC, stake.dao views
queries/nearai_pool_stake.sql (NEAR BigQuery)
-- Per-epoch balance of the nearai.pool.near staking pool (found among current validators via RPC).
-- Hypothesis to test: this is the pool behind NEAR AI's "Staking for Inference" (launched 2026-07-30,
-- https://near.ai/blog/staking-for-near-ai). NEAR AI's docs don't name the pool, so this is evidence,
-- not confirmation. The table is clustered by account_id, so the scan is small.
-- Used for: C5 (NEAR AI as a working third layer).
SELECT epoch_date, account_id, liquid, staked, unstaked_not_liquid
FROM `bigquery-public-data.crypto_near_mainnet_us.ft_balances_daily`
WHERE account_id = 'nearai.pool.near'
AND epoch_date >= '2025-01-01' -- partition filter (this table has epoch_date, not block_date)
ORDER BY epoch_date
-- Daily activity on stake.dao (House of Stake staking contract for NEAR AI products) since 2026-07-01:
-- function-call receipts sent to it, distinct accounts calling it, and NEAR attached to those calls
-- (deposits for locks, farm stakes and purchases). Gives a day-by-day history of NEAR AI staking
-- activity, which the contract's current-state views can't.
-- Reads four columns for ~4 months. Used for: C5 (NEAR AI staking adoption over time).
SELECT
block_date,
COUNTIF(action_kind = 'FUNCTION_CALL') AS function_calls,
COUNT(DISTINCT IF(action_kind = 'FUNCTION_CALL', receipt_predecessor_account_id, NULL)) AS calling_accounts
FROM `bigquery-public-data.crypto_near_mainnet_us.receipt_actions`
WHERE block_date >= '2026-07-01'
AND receipt_receiver_account_id = 'stake.dao'
GROUP BY block_date
ORDER BY block_date
-- L1 activity since 2025-01-01, from one scan of the transactions table (two result shapes):
-- level = 'daily': one row per day with total transactions, distinct signers (proxy for
-- daily active accounts), transactions sent directly to intents.near,
-- and NEAR burnt converting transactions to receipts.
-- level = 'receiver': one row per month and receiving account (top 25 per month), to see
-- which apps drive L1 activity and which large app left in early 2026.
-- GROUPING SETS lets both come from a single pass over four columns, which roughly halves the
-- bytes scanned compared with two separate queries. Longer L1 history (2020+) comes from NearBlocks.
-- Used for: C2 and the L1 layer in Part 2.
WITH grouped AS (
SELECT
IF(GROUPING(block_date) = 0, 'daily', 'receiver') AS level,
block_date,
month,
receiver_account_id,
COUNT(*) AS transactions,
COUNT(DISTINCT signer_account_id) AS signers,
COUNTIF(receiver_account_id = 'intents.near') AS tx_to_intents,
COUNT(DISTINCT IF(receiver_account_id = 'intents.near', signer_account_id, NULL)) AS signers_to_intents,
SUM(receipt_conversion_tokens_burnt) / 1e24 AS tx_conversion_burnt_near
FROM (
SELECT block_date, DATE_TRUNC(block_date, MONTH) AS month, receiver_account_id,
signer_account_id, receipt_conversion_tokens_burnt
FROM `bigquery-public-data.crypto_near_mainnet_us.transactions`
WHERE block_date >= '2025-01-01'
)
GROUP BY GROUPING SETS ((block_date), (month, receiver_account_id))
)
SELECT *
FROM grouped
WHERE TRUE
QUALIFY level = 'daily'
OR ROW_NUMBER() OVER (PARTITION BY level, month ORDER BY transactions DESC) <= 25
ORDER BY level, block_date, month, transactions DESC
queries/l1_app_split_daily.sql (NEAR BigQuery)
-- Daily L1 transactions since 2025-01-01, split by the two consumer apps that dominated L1 activity:
-- KaiChing: signer is 'kaiching' or any *.kaiching account (65% of March 2026 transactions; left in April 2026)
-- HOT: signer is any *.tg account (HOT's Telegram game and wallet accounts)
-- Signer-based, because these apps sign their users' transactions from their own sub-accounts
-- (queries/l1_drop_mar_apr_2026.sql shows signer and receiver totals for KaiChing are identical).
-- Reads only the signer column. Intents transactions per day come from l1_activity.sql.
-- Used for: C2 (L1 activity with the largest apps removed).
SELECT
block_date,
COUNT(*) AS transactions,
COUNTIF(signer_account_id = 'kaiching' OR ENDS_WITH(signer_account_id, '.kaiching')) AS kaiching_tx,
COUNTIF(ENDS_WITH(signer_account_id, '.tg')) AS hot_tg_tx
FROM `bigquery-public-data.crypto_near_mainnet_us.transactions`
WHERE block_date >= '2025-01-01'
GROUP BY block_date
ORDER BY block_date
Table: the numbers in this chart
Month
All L1
To intents.near
KaiChing
HOT
Everything else
Jan 2025
7.18M
186
2.96M
3.4M
818.52K
Feb 2025
6.87M
284
2.52M
3.48M
866.61K
Mar 2025
6.6M
390
2.27M
3.5M
831.15K
Apr 2025
6.26M
2.62K
2.62M
2.93M
708.54K
May 2025
5.25M
4.07K
2.53M
2.03M
685.54K
Jun 2025
4.89M
19.88K
2.52M
1.72M
629.15K
Jul 2025
4.8M
109.85K
2.4M
1.67M
627.22K
Aug 2025
5.03M
57.03K
2.62M
1.7M
646.6K
Sep 2025
4.68M
38.68K
2.49M
1.53M
613.62K
Oct 2025
4.55M
74.87K
2.48M
1.31M
687.24K
Nov 2025
4.27M
93.1K
2.4M
1.1M
665.82K
Dec 2025
3.84M
40.46K
2.3M
949.34K
557.28K
Jan 2026
3.51M
28.37K
2.27M
637.39K
571.66K
Feb 2026
3.49M
33.05K
2.27M
522.43K
667.79K
Mar 2026
3.48M
32.76K
2.28M
504.22K
666.54K
Apr 2026
946.61K
33.01K
45.77K
335.53K
532.3K
May 2026
803K
32.49K
0.23
214.77K
555.74K
Jun 2026
813.84K
36.21K
0
209.01K
568.61K
Jul 2026
737.9K
42.68K
0
189.51K
505.71K
Aug 2026
681.66K
55.33K
0
148.31K
478.02K
Sep 2026
800.15K
67.73K
1.8K
84.53K
646.09K
Oct 2026
729.93K
64.96K
20.74K
97.38K
546.86K
Accounts signing a transaction on NEAR L1 per day (monthly average)
The number of accounts signing a transaction on NEAR L1 each day fell from 882.5K in Jan 2025 to 72K in Oct 2026 (monthly averages).
Daily active accounts = distinct accounts that signed at least one transaction that day. Accounts signing transactions to intents.near are mostly solvers and relayers, so Intents users are counted from the Explorer API instead (the 1Click users chart). Complete months only.
-- L1 activity since 2025-01-01, from one scan of the transactions table (two result shapes):
-- level = 'daily': one row per day with total transactions, distinct signers (proxy for
-- daily active accounts), transactions sent directly to intents.near,
-- and NEAR burnt converting transactions to receipts.
-- level = 'receiver': one row per month and receiving account (top 25 per month), to see
-- which apps drive L1 activity and which large app left in early 2026.
-- GROUPING SETS lets both come from a single pass over four columns, which roughly halves the
-- bytes scanned compared with two separate queries. Longer L1 history (2020+) comes from NearBlocks.
-- Used for: C2 and the L1 layer in Part 2.
WITH grouped AS (
SELECT
IF(GROUPING(block_date) = 0, 'daily', 'receiver') AS level,
block_date,
month,
receiver_account_id,
COUNT(*) AS transactions,
COUNT(DISTINCT signer_account_id) AS signers,
COUNTIF(receiver_account_id = 'intents.near') AS tx_to_intents,
COUNT(DISTINCT IF(receiver_account_id = 'intents.near', signer_account_id, NULL)) AS signers_to_intents,
SUM(receipt_conversion_tokens_burnt) / 1e24 AS tx_conversion_burnt_near
FROM (
SELECT block_date, DATE_TRUNC(block_date, MONTH) AS month, receiver_account_id,
signer_account_id, receipt_conversion_tokens_burnt
FROM `bigquery-public-data.crypto_near_mainnet_us.transactions`
WHERE block_date >= '2025-01-01'
)
GROUP BY GROUPING SETS ((block_date), (month, receiver_account_id))
)
SELECT *
FROM grouped
WHERE TRUE
QUALIFY level = 'daily'
OR ROW_NUMBER() OVER (PARTITION BY level, month ORDER BY transactions DESC) <= 25
ORDER BY level, block_date, month, transactions DESC
-- Daily L1 transactions since 2025-01-01, split by the two consumer apps that dominated L1 activity:
-- KaiChing: signer is 'kaiching' or any *.kaiching account (65% of March 2026 transactions; left in April 2026)
-- HOT: signer is any *.tg account (HOT's Telegram game and wallet accounts)
-- Signer-based, because these apps sign their users' transactions from their own sub-accounts
-- (queries/l1_drop_mar_apr_2026.sql shows signer and receiver totals for KaiChing are identical).
-- Reads only the signer column. Intents transactions per day come from l1_activity.sql.
-- Used for: C2 (L1 activity with the largest apps removed).
SELECT
block_date,
COUNT(*) AS transactions,
COUNTIF(signer_account_id = 'kaiching' OR ENDS_WITH(signer_account_id, '.kaiching')) AS kaiching_tx,
COUNTIF(ENDS_WITH(signer_account_id, '.tg')) AS hot_tg_tx
FROM `bigquery-public-data.crypto_near_mainnet_us.transactions`
WHERE block_date >= '2025-01-01'
GROUP BY block_date
ORDER BY block_date
stake.dao contract views via NEAR RPC (fetched 2026-11-14)
Build it yourself
Every number, chart and document here is rebuilt from the raw data by one command, with no API keys.
One command
make charts
One command, run in the repository folder, rebuilds everything on this site from the raw data committed in data/raw/. It recomputes every number, redraws all 21 charts and rewrites THESIS.md, METHODS.md, DASHBOARD.md and site/index.html. It needs make and Python 3.11 to 3.14, and no API keys. The first run creates .venv/ and installs the packages pinned in requirements.txt.
The path every chart takes: data/raw/ (as fetched) → analysis/*.py → data/derived/*.csv → viz/build.py → site/charts/*.svg and this site. Each chart on the Dashboard tab names its sources, its data files, its code and its query.
make check deletes every generated file, rebuilds them and confirms each one is identical to the committed copy. make verify compares random slices of the committed data with the live sources (about a minute, no keys). The README lists the commands that download the raw data again, and what each one needs.
Do X, see Y
#
Do
See
0
git clone https://github.com/RehanNek/near-investment-thesis.git, then cd near-investment-thesis
The repo, with README.md, THESIS.md and DASHBOARD.md at the top level
1
make charts
The line Built THESIS.md, METHODS.md, DASHBOARD.md, site/index.html and 21 charts in site/charts/ (data as of 2026-11-14). and then four key points
2
make check
OK: all 53 generated files match the committed copies
3
Open site/index.html in a browser
This site, with every chart under the Dashboard tab
4
In config.yaml, under scenarios: then base:, change half_life_years: 2 to 3. Run make charts
A different break-even date in the third key point, in the thesis and in the scenario charts
5
make verify
A table of committed and live values, then a count of how many match. The run passes when nothing differs
The README sets out each step in full, lists the commands that download the raw data again, and says what this was tested on.
Methods, evidence in detail and sources
The reference pages behind THESIS.md, with data as of Nov 14, 2026. The charts and their queries are on the dashboard. How to rebuild and verify everything is in the README.
Revenue (C1). L1 side: NEAR burned by gas, measured from block-to-block changes in total supply (BigQuery), valued at the daily NEAR price (NearBlocks). Intents side: NEAR's net revenue after partner payouts, from revenue.near.org. The net basis compares what the protocol keeps on each side; the gross basis compares all fees users pay. L1 gross fees = burn ÷ 0.7 before Nov 5, 2026, while 30% of gas went to the called contract, and = burn from that day, when all gas is burned. Before the switch this is an upper bound, because some gas was always burned in full, so C1's Intents share is conservative.
L1 activity (C2). BigQuery transactions table, daily since Jan 2025: transactions, distinct signers (daily active accounts), transactions sent to intents.near, the busiest receiving contracts each month, and the two largest consumer apps split out by signer account (KaiChing: *.kaiching; HOT: *.tg and hot.tg). Every query on a block-level table filters on block_date; the two balance queries filter on epoch_date, the partition column of ft_balances_daily.
Intents breadth (C3). NEAR Intents Explorer API, successful swaps. A user is the first sender address of a swap, or its refund address when no sender is given. The top-ten share is the share of a month's volume sent by its ten largest users. Stablecoin shares use the token on each side of the swap.
Market share (C4). DefiLlama daily volumes. Peer set: NEAR Intents, THORChain, Chainflip and Maya Protocol, the venues that swap native assets across chains. Bridge aggregators and total DEX volume are reference series, kept out of the denominator to avoid double counting.
NEAR AI (C5). Tokens per model per day from OpenRouter's NEAR AI provider page. Revenue estimate: NEAR AI's OpenRouter prices for each model, applied with that model's prompt/completion split. Staking from stake.dao contract views and archival RPC balances of nearai.pool.near; the model catalog from NEAR AI Cloud's public model list.
Capture ratio (C6). The share of new NEAR that burns and buybacks offset. Total supply jumps at each epoch boundary by the newly minted NEAR and otherwise only falls, by the NEAR burned; summing those changes gives NEAR issued and burned per day. NEAR captured as Intents revenue comes from revenue.near.org. New AI stake is the month-end change in nearai.pool.near stake.
Follow the money. The latest month with complete data for all three layers. NEAR AI's sales there are the OpenRouter estimate above, summed over the month. The staking row is a single snapshot, dated in the chart, not a monthly series: total stake is the sum of the current validators' stake from one NEAR RPC validators call, and the staked share divides it by the total supply in the latest block header (fetch/rpc.py).
Staked NEAR, month by month (C6). The same two figures, read at the last block of each complete month from Jan 2025 to Oct 2026. The block's header, from FastNEAR's NEAR Data API, gives its epoch and the total supply; one validators call for that epoch on the archival RPC gives every validator's stake (fetch/staking_history.py). A regular RPC node answers only for the current epoch, which is why the money-flow figure is a single day. The public archival node allows only a few calls at a time and its operator marks it as deprecated, so this fetch is slow and cached. Rebuilding the report reads the saved file and doesn't call it. The lowest month end stands well below the months either side of it, so fetch/staking_validators.py saves every validator's stake at those three month ends, and the note under the chart says which validators' stake fell.
Protocol parameters. Issuance cap and the gas rebate to contracts are read from NEAR's protocol config at past blocks through the archival RPC (fetch/protocol_history.py), which dates each change to the day.
Cross-checks (data/derived/crosschecks_monthly.csv). NEAR price: within 0.5% of CoinGecko in every month. NEAR issued: median gap of 0.5% to the protocol rule (issuance rate × supply). Intents revenue: DefiLlama records NEAR when it is swept into the revenue wallets, in batches, so single months differ; over the last six months the median gap is +21%. L1 gross fees, ours vs DefiLlama +18% and NearBlocks +18% (median monthly gap).
What we couldn't compute
NEAR AI Cloud's revenue and total usage. Its usage and billing APIs need an organisation key, and NEAR AI publishes neither figure. We see OpenRouter traffic and on-chain staking only, and the OpenRouter revenue figure is an estimate at list prices.
Intents users and concentration across all channels. The Explorer API covers the 1Click channel, about 27% of Intents volume in Oct 2026. Swaps through other channels are in the volume figures (revenue.near.org, DefiLlama) but not in the user counts.
Confidential swaps. The Explorer masks them, so they are missing from the user, referrer and chain breakdowns.
The whole cross-chain market. DefiLlama's per-bridge volumes (Relay, deBridge, Across, Mayan and others) are paid-only, so C4's denominator is a peer set, not the whole market.
Gas paid by Intents transactions. We count transactions to intents.near. Attributing gas needs receipt-level tables too large for our query budget.
Where all captured NEAR went. 770.4K NEAR captured as Intents revenue is not in the three revenue wallets we track, and we couldn't trace it.
Whether AI credits get used. Credit consumption happens off-chain.
A price target. We don't model one. The thesis is about where value is created and how much of it reaches the token, and there is no reliable multiple for a token whose revenue mostly funds buybacks.
The evidence in detail, claim by claim
THESIS.md tests the thesis line in one table, with a row for each thing the line claims. Behind those rows are six narrower claims, C1 to C6, each tied to its own charts. This section sets out each one in full: the evidence for it, its caveats and the result that would change our mind about it.
C1. Most of NEAR's revenue now comes from beyond the L1
Evidence for
NEAR's share of Intents fees overtook the L1 fee burn in Oct 2025. Over the last three months (Aug 2026–Oct 2026), Intents was 95.2% of the value the protocol kept (97.8% of all fees paid).
Since Jan 2026, Intents has been above 82% of protocol revenue every month.
Caveats
Part of the shift is the L1 shrinking rather than Intents growing: L1 fees burned fell from $567K in Jan 2025 to $25K–$47K a month in Apr 2026–Aug 2026, as the largest L1 app left (C2) and the NEAR price fell from about $5.12 to $1.80.
L1 gas is cheap by design, so L1 revenue was never large, and a high Intents share is easy to reach.
What would change our mind: Intents' share of protocol revenue falling back below half. Today: 95% of protocol revenue over Aug 2026–Oct 2026, against 34% a year earlier.
L1 transactions fell from 7.2M a day in Jan 2025 to 729.9K in Oct 2026, mostly as KaiChing and HOT wound down. Transactions to intents.near rose from 28.4K to 65K a day over Jan 2026–Oct 2026.
Intents' share of L1 transactions rose from 0.8% in Jan 2026 to 8.9% in Oct 2026. Leaving out KaiChing and HOT, it went from 4.7% to 10.6%.
Caveats
Most of the rising share is the rest of the L1 shrinking: KaiChing accounts sent 65% of all L1 transactions in March 2026, and their stop caused 87% of April's fall.
Transactions to intents.near were higher in Jul 2025 (109.8K a day) than in Oct 2026 (65K), though Intents volume is 17× larger now, so transaction counts don't track Intents' growth.
Leaving out KaiChing and HOT, Intents' share was 14.9% in Jul 2025, above today's 10.6%.
What would change our mind: Intents' share of L1 transactions no longer rising over a 12-month span. Today: 8.9% of L1 transactions in Oct 2026, against 1.6% a year earlier.
Covers the 1Click channel only, about 27% of Intents volume in Oct 2026. The first 1Click swap in the data is on Mar 27, 2025.
Evidence for
In Oct 2026, 1Click handled 505.9K swaps worth $1.1B from 269K distinct users, through 352 referrer apps and 785 chain pairs. A year earlier (Oct 2025): 132.6K users, 48 apps and 338 chain pairs.
91% of Oct 2026 volume moved between two chains other than NEAR. Swaps started on 37 chains; the largest was Ethereum at 31%.
Caveats
Much of the flow is trading: in Oct 2026, 34% of 1Click volume was stablecoin-to-stablecoin (up from 9% a year earlier) and 74% involved a stablecoin.
A few routes carry a lot of it: one referrer app (swapkit) routed 40% of volume, the top 10 users 15% (up from 11% a year earlier), and Zcash was 14% of origin volume.
Distinct users have eased since their peak: 415.7K in Aug 2026, 269K in Oct 2026.
What would change our mind: 1Click volume or users falling over a 12-month span, or the top 10 users' share of volume rising above half. Today: $1.1B of 1Click volume and 269K users in Oct 2026, against $561.8M and 132.6K a year earlier; top 10 users at 15% of volume.
C4. Intents is gaining share, not just moving with the market
Evidence for
Among native cross-chain swap venues tracked by DefiLlama, Intents' share of volume rose from 45% in Oct 2025 to 58% in Oct 2026, after peaking at 79% in Jun 2026.
Over the 12 months to Oct 2026, Intents volume rose 131% while total DEX volume fell 56% and bridge-aggregator volume rose 113%.
Caveats
The share has eased since its Jun 2026 peak (79% → 58%).
The peer set leaves out large bridges whose data is paid (Relay, deBridge, Across, Mayan), so the level is overstated.
Bridge aggregators grew 113% over the same 12 months, close to Intents' 131%.
Recent swings move with the market: in Sep 2026 Intents volume rose 73% while bridge aggregators rose 93% and all DEX volume rose 30%; in Oct 2026 the three moved -17%, -10% and -17%.
What would change our mind: Intents' share of native cross-chain swap volume falling over a 12-month span. Today: 58% of native cross-chain swap volume in Oct 2026, against 45% a year earlier.
Usage. NEAR AI served 225.2B tokens through OpenRouter in the week to Nov 8, 2026, up from 14.4B in its first full week (to Sep 20, 2026). At list prices that week was worth about $19K, or $975.8K a year.
Staking. NEAR staked with NEAR AI's validator rose from 10K on Jul 10, 2026 to 1.5M on Nov 14, 2026. 137 accounts stake for NEAR AI Cloud credits, and 84 have subscribed to NEAR AI Agents.
Product. 64 models in the NEAR AI Cloud catalog, 17 of them running in TEEs for verifiable private inference.
Distribution. Since Oct 15, 2026, Brave's AI assistant Leo runs GLM 5.3 Flash on NEAR AI Cloud in the production browser. Brave has over 128M monthly users, though only Leo users who pick that model use NEAR AI Cloud. NEAR AI also names Venice, the Government of Bermuda and Abound as customers.
Developers. IronClaw, NEAR AI's open-source agent OS, has 12.7K GitHub stars.
Caveats
NEAR AI doesn't publish revenue. Our OpenRouter estimate is about $975.8K a year, and OpenRouter is one channel among several.
Staking is concentrated: the largest staker holds 73% of the credits farm and the top 10 hold 95%.
Staking funds about $13.8K of credits a month, and contract activity has been flat at 168–395 calls a week.
What would change our mind: Weekly tokens on OpenRouter falling for a sustained period, or AI staking shrinking. Today: 225.2B tokens on OpenRouter in the week to Nov 8, 2026; 1.5M NEAR staked on Nov 14, 2026.
C6. The share of value reaching the token is rising
Evidence for
NEAR issued per month fell from about 5.3M to 2.7M after the cut, while NEAR captured by the L1 burn and Intents revenue rose to 544.3K in Oct 2026. New stake for NEAR AI added another 664.8K in Aug 2026 and 687.3K in Oct 2026.
The capture ratio rose from about 2% a quarter in 2025 Q1–Q3 to 19.5% in 2026 Q3. Counting new AI stake, Aug 2026 reaches 43% and Oct 2026 reaches 45%.
Caveats
The ratio is still small: in 2026 Q3, 80.5% of new NEAR was not offset.
About 53% of the rise comes from the inflation cut rather than more revenue: at 2025 Q3's issuance, 2026 Q3's captured NEAR would have offset only 10.0%.
The L1 burn is only 0.8% of issuance.
Bought-back NEAR is held, not burned: 4.1M of the 4.9M NEAR captured as Intents revenue is still in NEAR's revenue wallets.
These mechanisms are governance decisions that can change, and NEAR staked for NEAR AI is locked, not removed, and can be withdrawn.
The share of all NEAR staked with validators fell from 48.5% to 45.2% between the end of Oct 2025 and the end of Oct 2026, and the amount staked fell from 619.8M to 592M NEAR.
What would change our mind: The share of new NEAR offset by burns and buybacks no longer rising, now that the fee switch is on, or the share of all NEAR that is staked falling for a second year running. Today: 19.5% of new NEAR offset in 2026 Q3, against 1.8% a year earlier.
The capture ratio is projected monthly from the last complete quarter. NEAR issued is supply × issuance rate ÷ twelve, under HSP-029's phase-in (the main path) and under today's rate. Intents revenue in NEAR starts at its trailing growth rate, measured from the data over the window set in config.yaml; the base case lets that growth halve every few years, the bull case holds it, the bear case has none. L1 burn is held at the last quarter's level and scaled up from the day all gas is burned. Inputs are in config.yaml.
Intents revenue in NEAR grew 116% a year over 2026 Q1 → 2026 Q3. If that growth halves every 2 years (base case), burns and revenue fully offset new supply by Dec 2028 with HSP-029, or Mar 2030 if issuance stays at 2.5%. If it holds (bull), by Aug 2028; with no growth (bear), not within 6 years.
Issuance path
Scenario
Revenue growth a year at start (NEAR terms)
Growth then
After 1 year
After 3 years
Supply after 6 years
Breakeven
Revenue at breakeven vs today
With HSP-029 (2.5% → 1.6%)
bear
0%
–
22.0%
29.2%
1.458B
not within 6 years
–
With HSP-029 (2.5% → 1.6%)
base
116%
halves every 2 years
40.7%
138%
1.458B
Dec 2028
3.7×
With HSP-029 (2.5% → 1.6%)
bull
116%
held
43.4%
264%
1.458B
Aug 2028
4.1×
If HSP-029 isn't implemented (stays 2.5%)
bear
0%
–
19.4%
18.4%
1.521B
not within 6 years
–
If HSP-029 isn't implemented (stays 2.5%)
base
116%
halves every 2 years
35.7%
87.1%
1.521B
Mar 2030
5.8×
If HSP-029 isn't implemented (stays 2.5%)
bull
116%
held
38.2%
166%
1.521B
Jan 2029
5.7×
Projections, not findings. Starting point: 2026 Q3's Intents revenue and L1 burn (19.5% of new NEAR offset in Oct 2026). Growth is in NEAR terms, so a rising NEAR price slows it. L1 burn is held at 2026 Q3's level, scaled to 100% burned from Nov 5, 2026. Lines leave the chart above 200%.
How much depends on the growth window. Measured over different recent windows, Intents' revenue growth ranges from 116% to 250% a year. With HSP-029, that puts breakeven between Jan 2028 and Dec 2028, depending on the window and on whether growth holds or fades. The scenarios use 2026 Q1 → 2026 Q3 (116% a year).
Trailing window
Growth a year (NEAR terms)
With HSP-029, growth fades
With HSP-029, growth holds
At 2.5%, growth fades
At 2.5%, growth holds
2026 Q2 → 2026 Q3
186%
Jun 2028
Mar 2028
Nov 2028
Jun 2028
2026 Q1 → 2026 Q3
116%
Dec 2028
Aug 2028
Mar 2030
Jan 2029
2025 Q4 → 2026 Q3
250%
Mar 2028
Jan 2028
Jun 2028
Mar 2028
When each mechanism switched on
NEAR's issuance fell from 5.0% to 2.5% a year at the Oct 2025 cut. Every mechanism that sends value to the token has switched on since: buybacks on Feb 23, 2026, staking for NEAR AI on Jul 30, 2026 and burning all L1 gas on Nov 5, 2026. HSP-029 (passed Nov 2, 2026) would lower issuance to 1.6% by Jan 2029.
Mechanism
Since
What it does
Evidence
Issuance cap cut from 5% to 2.5%
Oct 30, 2025
Halves the NEAR issued each year
Protocol config via archival RPC; visible in BigQuery supply flows
All L1 gas burned (HSP-027; the 30% rebate to called contracts ends)
Sourced facts no API exposes: data/manual/facts.yaml. Fetch log with dates: data/raw/_manifest.json.
Tools used
Python with requests, numpy, pandas, pyarrow and PyYAML (pinned in requirements.txt). Running the BigQuery queries again also needs google-cloud-bigquery (pinned in requirements-bigquery.txt); rebuilding does not. Charts are plain SVG drawn by viz/svg.py, with no plotting library; the dashboard's hover layer is site/app.js. Code and drafts were written with Claude Code (AI) and reviewed by the author. NEAR agent skills from near/agent-skills.