On 15 September 2026 Virtuals launched a new product called Occupy, and two days later the platform went live on the Arc blockchain. By 5 October the VIRTUAL token had risen 39 percent, bitcoin 10 percent over the same period. Since the start of the year VIRTUAL is up 33 percent and bitcoin down 2 percent.
The reason the token is supposed to benefit lies in how the platform is built. Every AI agent that launches on Virtuals gets its own token, and that token trades against VIRTUAL. More agents, more trading, more demand for VIRTUAL. That is how simple it sounds.
If that is why you buy VIRTUAL, one number matters: how much VIRTUAL do the agents actually hold?
What Virtuals does: a launchpad for agent tokens
Virtuals is not its own network but a platform on Base, Coinbase's blockchain, and now also on Solana, Ethereum, Robinhood Chain and Arc. It grew out of PathDAO, a gaming guild from 2021. The launchpad on Base has been running since 16 October 2024.
Anyone launching an AI agent, say a bot that posts market commentary on X, issues a token for it. The token starts on a so-called bonding curve: buyers pay in VIRTUAL, and every purchase lifts the price a little. Once 42,000 VIRTUAL have been paid in, about $36,000 today, the agent graduates. It then moves into a regular trading pool where its token and VIRTUAL sit against each other. That pool is locked for ten years.
Every trade in an agent token carries a one percent fee. 70 percent goes to the agent's creator, 30 percent to the Virtuals treasury.
Virtuals does not sell AI. It sells the IPO for AI agents.
What turns that into a story for the VIRTUAL token?
What makes it different: VIRTUAL sits on the other side of every agent
On most launchpads the platform token is an add-on. On Virtuals it is the counter-currency. Buying an agent token puts VIRTUAL into its pool. Selling it takes VIRTUAL back out. The more money sits in agents, the more VIRTUAL is locked in pools and missing from the market.
At the end of 2024 that is exactly what happened. In November 6,776 agents launched, and bots like AIXBT and Luna reached market values in the hundreds of millions of dollars. In December the Virtuals treasury took in $14.49 million in fees according to DefiLlama, in a single month. From the end of May to the end of December 2024 VIRTUAL rose 4,491 percent, bitcoin 39 percent.
If the story had ended there, the case for the token would be proven.
At the end of 2024 the mechanism did exactly what it promises.
Has it kept doing so?
Who uses Virtuals today: many launches, few pools
According to the Virtuals API, 85,254 agents had launched by 6 October 2026. 1,092 of them graduated, plus 195 from the earlier Genesis launch model. 83,967 agents stopped on their bonding curve. That is 98.5 percent.
| Period | Agents launched | of which graduated |
|---|---|---|
| October 2024 to January 2025 | 14,471 | 537 |
| February to December 2025 | 1,668 | 280 |
| 2026 to 6 October | 68,998 | 158 |
2026 brought more new agents than ever, 21,418 in July alone after Virtuals launched on Robinhood Chain. 158 graduated in the whole year.
And most graduated agents barely trade anymore. 1,170 of the 1,287 agents with a pool traded less than $100 in the 24 hours to 6 October. Together they came to $2.74 million, three quarters of it in the ten most traded. The VIRTUAL token alone traded $9.1 million on Binance the same day.
Treasury revenue follows suit. $30.3 million of the $41.7 million to date came in the four months from October 2024 to January 2025. From January to September 2026 it was $2.39 million, against $15.71 million in the same months of 2025.
More agents launch than ever. The trading happens somewhere else: in VIRTUAL itself.
What does that mean for the amount of VIRTUAL tied up in agents?
How much VIRTUAL the agents hold
The answer is on the blockchain. For each of the 956 agent pools on Base, where 1,151 of the 1,287 graduated agents live, the VIRTUAL balance can be queried directly. As of 6 October 2026:
| VIRTUAL | Share of supply | |
|---|---|---|
| in all 956 agent pools on Base | 22.75 m | 2.3 % |
| of which in the ten largest pools | 8.77 m | 0.9 % |
| on bonding curves on Base (estimate from samples) | 1 to 3 m | 0.1 to 0.3 % |
| total supply | 1,000 m | 100 % |
Together that is about 25 million VIRTUAL, roughly $21 million. The token as a whole has a market value of about $565 million, excluding the treasury.
The typical pool is small. The median pool holds 8,444 VIRTUAL, a fifth of what an agent needs to graduate today. Only 69 of the 956 pools are above that threshold. In the rest, holders have sold more agent tokens than they bought since graduation, and the VIRTUAL has left the pool again.
In October 2024, after the first month, Virtuals itself reported 15 million VIRTUAL in agent pools. Two years and 85,000 launches later the figure is 22.75 million.
The agents tie up a fortieth of the token. The other 39 parts trade as if they did not exist.
That is where it stands today. Does it change once the agents start earning money?
The obvious objection
"Pool trading is the old model. The new demand comes from the agent economy: agents paying each other for work."
Since February 2025 Virtuals has been building the Agent Commerce Protocol, ACP for short, and it reports an "agentic GDP" of $482 million. Two things stand against the objection. First, on 12 August 2025 Virtuals switched payment in ACP from VIRTUAL to USDC, the dollar stablecoin. That is what its own changelog says, and since then prices and fees there are set in USDC. Second, Virtuals counts capital raised at token launches into this agent GDP. The revenue ACP has brought the platform since launch was about $452,000, according to a Fundstrat report of 24 June 2026. Virtuals paid for that report.
An agent economy that settles in dollars does not need VIRTUAL.
So who gets the money earned on the platform?
Tokenomics: one billion, one treasury, nothing flowing to holders
VIRTUAL has a fixed supply of one billion tokens. 60 percent went to the public, 5 percent to liquidity and 35 percent to an ecosystem treasury that holders have voted on since July 2025. According to CoinGecko, 658 million are in circulation. The treasury may release up to 10 percent a year if a vote approves it.
Platform fees go into that treasury. According to DefiLlama none of it has flowed to VIRTUAL holders in any month since launch. There is no ongoing buyback.
In February 2025 Virtuals wrote that it had bought back and burned 12.9 million VIRTUAL, permanently reducing supply. According to Unchained, the money bought and burned agent tokens, not VIRTUAL. The blockchain confirms this: on Ethereum, where the token is issued, exactly 1,000,000,000 VIRTUAL exist on 6 October 2026, and the usual burn addresses hold 1,498.
One treasury movement stands out. On 13 February 2026, according to the analytics service Lookonchain, 13.05 million VIRTUAL went from project addresses to Binance, with no explanation from Virtuals. That is more than half of what all agents together hold today.
What Virtuals earns stays in the treasury. What the holder gets is the price.
And what does the price follow?
The price against bitcoin
Since the end of May 2024 VIRTUAL has shown two completely different faces.
| VIRTUAL | BTC | |
|---|---|---|
| $1 at the end of May 2024 became | $10.04 | $1.27 |
| $1 at the peak, 1 January 2025, became | $0.19 | $0.91 |
| Up months | 15 of 28 | 15 of 28 |
| Median month | +2.2 % | +2.2 % |
| Months in which BTC fell | 13 | |
| … VIRTUAL did better | 5 | |
| … VIRTUAL itself was negative | 8 | |
| Largest fall from the peak | −90.5 % | −53.0 % |
| Annual volatility | 157 % | 45 % |
| Year | VIRTUAL | BTC | Treasury revenue |
|---|---|---|---|
| 2024 (from 31 May) | +4,491 % | +38.5 % | $21.0 m (from 16 October) |
| 2025 | −83.6 % | −6.3 % | $18.4 m |
| 2026 (to 5 October) | +33.0 % | −2.1 % | $2.4 m (to September) |
In the median month VIRTUAL and bitcoin are level. A few months make the difference: October and November 2024 brought +551 and +425 percent, the crash after the peak −90 percent in just over three months.
In 2024 and 2025 price and revenue moved together, first up, then down. In 2026 they move apart. The price is rising while revenue is 85 percent below the previous year.
In 2026 VIRTUAL rises with the market for AI tokens, not with its own agents.
Announcements against evidence
| Claim | Who, when | What the data show | Verdict |
|---|---|---|---|
| $60 m revenue by end of 2024, $300 m annualised | Virtuals monthly update, 08 Jan 2025 | DefiLlama: about $21 m in fees over the same period; basis not disclosed | contradicted in magnitude |
| 12.9 m VIRTUAL bought back and burned | Virtuals monthly update, 05 Feb 2025 | supply on Ethereum unchanged at 1 bn; agent tokens were burned (Unchained, 15 Jan 2025) | contradicted |
| Agent GDP $470 m, later $482 m | Virtuals, 12 Feb 2026 and autumn 2026 | definition includes capital from token launches; ACP platform revenue about $0.45 m | not verifiable as economic output |
| Listings on Binance, Upbit, Coinbase | 11 Apr 2025, 31 Jan 2025, 29 Apr 2026 | first prices in exchange data on those days | verified |
| Launch on Robinhood Chain and Arc | Virtuals, 02 Jul 2026 and 17 Sep 2026 | token contracts exist, 92 and 1 graduated agents | verified |
| Launch on BNB Chain and XLayer in Q2 2026 | DEXTools report | no token contract found | did not happen |
| BasisOS as an AI-run yield agent | promoted on Virtuals, 2025 | according to 99Bitcoins run by hand by an insider, about $500,000 lost, refund promised (28 Nov 2025) | contradicted |
| "Stripe for AI agents" | Fundstrat, October 2025 and 24 Jun 2026 | report commissioned and paid for by Virtuals | source not independent |
The exchange listings and the new chains can be checked. The figures on the agent economy cannot.
What this means for you
If you buy VIRTUAL because you believe in AI agents: the link between the two is thinner than the story suggests. The agents hold about 2.5 percent of the token, they have paid for jobs in USDC since August 2025, and fees go to the treasury, not to you. What you buy is a token that swings with the market for AI narratives, more than three times as hard as bitcoin.
If you buy agent tokens on Virtuals: 98.5 percent of all launches never leave the bonding curve. AIXBT, Luna and Game, the flagships of 2024, are 97 to 99 percent below their highs according to CoinGecko.
The agents are the story. What people buy is the token next to it.
Two numbers show whether that changes: the amount of VIRTUAL in agent pools, 22.75 million today, and monthly treasury revenue, $0.19 million in September. The open question is whether Occupy and Arc get more agents over the bonding curve than Robinhood Chain did in July, when 21,418 launches produced 78 graduations.
FAQ
What is a bonding curve? A contract that sets the price of a new token from the amount paid in so far. Every purchase lifts the price, every sale lowers it. On Virtuals the curve ends at 42,000 VIRTUAL paid in, and the token then moves into a regular trading pool.
How were the 22.75 million VIRTUAL counted? For each of the 956 agent pools on Base listed by the Virtuals API, the VIRTUAL balance was queried directly on the blockchain on 6 October 2026 and added up. Pools on Solana, Robinhood Chain, Ethereum and Arc, home to 136 of the 1,287 graduated agents, are not included. The bonding curves on Base are extrapolated from samples of about 2,300 curves.
Why does the 42,000 VIRTUAL threshold not apply to every pool? It applies to launches today. In October 2024 an agent graduated at a $420,000 market value. The comparison shows how small the median pool is today, not that every pool once held exactly 42,000 VIRTUAL.
How are the monthly returns calculated? Month-end close against the previous month-end close, VIRTUAL against USDT on Gate.io, bitcoin against USDT on Binance, without fees. Binance has listed VIRTUAL only since 11 April 2025; from then on the two exchanges differ by 0.04 percent on average.
All VIRTUAL monthly returns in US dollars
| Year | Jan | Feb | Mar | Apr | May | Jun | Jul | Aug | Sep | Oct | Nov | Dec |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2024 | −65.4 | 76.7 | −34.0 | 49.7 | 551.4 | 424.9 | 122.6 | |||||
| 2025 | −49.7 | −44.0 | −47.8 | 166.2 | 31.3 | −25.4 | −16.7 | −11.4 | −8.8 | 34.3 | −33.4 | −28.8 |
| 2026 | 0.8 | 8.8 | −7.2 | 4.5 | 5.2 | −27.2 | 3.7 | 27.8 | 15.6 |
Figures in percent. June 2024 from the close on 31 May.
Not investment advice, not a recommendation, not a forecast — historical patterns are no guarantee.
Study the Past — Improve your Future 🥋