AI-linked cryptocurrencies just nuked the entire AI narrative trade from 2024. Data from CryptoPresales shows AI and big data tokens lost about 75% of their combined value in 2025, erasing more than $53 billion in market cap. Forklog and Cryptonews both ran the numbers on December 25, and they line up.
From $69.9B peak to a $16.8B sector
CryptoPresales reconstructs a two-year boom then a straight drawdown. The AI token basket posted a 1,873% gain in 2023, then a 157% jump in 2024, lifting combined market cap from about $23 billion in August 2024 back to $55.5 billion by year end. The same dataset shows the category pushing to a higher peak of $69.9 billion in December 2024 as neural-network hype peaked and new AI narratives flooded exchanges.
That move reversed hard in 2025. By April, the sector had already surrendered 63% of its value, or roughly $44 billion, according to the CryptoPresales report. CoinMarketCap’s AI and big data screen now sits near $16.8 billion in aggregate value, about a quarter of the December 2024 peak, while CoinGecko’s broader categorization puts the figure closer to $25.4 billion. Forklog highlighted that spread and tied it to index construction rather than any fresh leg lower.
Q4 selloff finished the job
The bleed started in January but the real damage landed into year end. CryptoPresales and Cryptonews both show the basket drifting around $30 billion in mid 2025, then cracking again in the fourth quarter. November shaved off about $4 billion. December added a further $10 billion wipeout, taking the sector to its current $16.8 billion footprint just days before the close of the year.
That timeline matches on-chain and venue-level flows. Cointelegraph flagged 75% to 90% drawdowns across agentic AI names like ARC, ElizaOS and Virtuals as early as February 6, citing CoinGecko data and calling out a rotation of liquidity into other narratives once the first AI mania cooled. By December, CryptoPresales’ chart of the top 10 AI and big data tokens showed almost uniform double-digit billion value destruction across the board.
Leaders take 80%+ drawdowns
The headline 75% sector loss hides even bigger pain at the token level. In the CryptoPresales breakdown, Artificial Superintelligence Alliance (ASI) ranks as the worst hit, down 84% over the past year. Render (RNDR) and The Graph (GRT) each gave up around 82%. Former darling Virtuals Protocol, which exploded more than 3,500% in 2024, sits 73% below its December 2024 level, with Injective (INJ), Filecoin (FIL), Internet Computer (ICP) and NEAR Protocol (NEAR) all showing similar trajectories.
The live tape reflects that reset. Artificial Superintelligence Alliance’s FET token trades around $0.21 today, according to CoinMarketCap, versus an all-time high of $3.47 recorded on March 28, 2024. That is a 94% slide from peak to current levels, even deeper than the 84% one-year drawdown CryptoPresales calculates from December 2024. RNDR now hovers near $1.29, while GRT trades just above $0.036, based on spot market quotes across major exchanges.
The sector lag is broad, not just limited to the pure AI agent names. NEAR sits close to $1.50. Smaller AI infrastructure and data plays show similar charts, with long, flat bases from the 2022 bear market, a vertical spike across late 2023 and early 2024, then a steep grind lower across most of 2025 as liquidity thinned and issuance plus unlocks met dwindling inflows.
“onchain economy = agentic economy”
Stephen (@meta_hess) on X, reacting to the AI token washout.
Crypto AI trade diverges from real AI boom
The timing clashes hard with what happened in listed AI equities and private markets this year. Moneybox’s year-end review of Nvidia’s 2025 performance shows NVDA up roughly 34% to 39% year to date and becoming the first public company to clear a $5 trillion valuation in late October. Nvidia’s data center revenue printed record quarters and AI infrastructure spending kept ramping, even as the tokenized AI trade went in the opposite direction.
Macro explains part of the split. CryptoPresales points to the US clampdown on AI chip exports to China, geopolitical risk and profit-taking in mid 2024 as the first trigger, followed by tighter liquidity and a rotation away from smaller, higher-beta AI tokens once the Federal Reserve turned more hawkish in early 2025. Cointelegraph’s February reporting echoed that view, with traders describing AI agents as collateral damage when capital chased fresh themes and the Trump memecoin sucked oxygen out of smaller narratives.
The other part is structural. Nvidia and the other AI mega-caps now sit on visible revenue from GPUs and cloud contracts. Most AI-branded tokens still rely on promises, roadmaps and thin protocol fees. CryptoPresales’ dataset makes that gap obvious. The market punished anything that looked like narrative-only exposure, kept selling into every bounce and rewarded only a small handful of projects with durable cash flows or defensible infrastructure.
What this reset actually signals
This is not just another altcoin winter chart. It is the first full stress test of the AI token story against a real AI capex supercycle. Investors just saw that a roaring AI trade in equities and private markets does not automatically transfer to every token with an “AI” tag. The 75% sector collapse in 2025 forces market makers, allocators and builders to separate agent memes and presales from networks that can actually clear usage through a token.
For now, the receipts are clear. AI and big data tokens tracked by CryptoPresales have dropped from a $69.9 billion peak to a $16.8 billion sector in twelve months, with eight of the top ten names down more than 70%. That is the correction the narrative needed. Whether any of these tokens graduate from trade to infrastructure will show up in the next cycle’s revenue lines, not in another wave of AI-branded tickers.
