Christie's closed its dedicated digital art department in September 2025. That same month, the London auction house held its first auction entirely devoted to artificial intelligence, titled "Augmented Intelligence." Total sales: $728,784. Sell-through rate: 82%. The contradiction is only apparent. It reveals a market moving from the margins to the mainstream—a market growing by becoming invisible.
The numbers tell a story of a much vaster territory than a niche phenomenon. The AI-generated art market was valued at $3.2 billion in 2024, according to Grand View Research, with a projection to $40.4 billion by 2033—a compound annual growth rate of 28.9%. Compared to the global art market ($57.5 billion in 2024, down 12% year-on-year according to Art Basel/UBS), AI art already constitutes a substantial segment. Crucially, it is the only one whose curve points upwards while the rest of the market declines.
From Foundational Misunderstanding to Programmed Records
It all began with an anomaly. On October 25, 2018, Edmond de Belamy, a portrait produced by the French collective Obvious using a Generative Adversarial Network (GAN), sold for $432,500 at Christie's. The high estimate was $10,000. A multiplier factor of 43. Observers spoke of a bubble, a media stunt, a curiosity. They were right on all counts—and wrong to believe the story would end there.
In March 2021, Beeple sold Everydays: the First 5000 Days for $69.3 million at Christie's, the first purely digital non-fungible token (NFT) to cross the threshold of a major auction house. The shockwave masked a deeper movement: the rise of artists working with AI as a creative tool, not as a provocation. Refik Anadol reached $2,353,864 at Sotheby's Hong Kong the same year. Mario Klingemann, a pioneer of neural art, had already secured $51,000 at Sotheby's in 2019 for Memories of Passersby I.
Prices now follow a clear stratification. At the top, a handful of institutionalized names. Anadol, whose Unsupervised joined MoMA's permanent collection in October 2023, reached $900,000 at Sotheby's "Origins" sale in Riyadh in February 2025. Holly Herndon, a Stanford PhD and TIME100 AI figure, sold for $94,500 at Christie's with Embedding Study (co-created with Mat Dryhurst). Claire Silver, the first AI artist signed by the WME agency, was auctioned for $44,100.
In the middle of the spectrum, an unprecedented case. Botto, an autonomous program created by Mario Klingemann and ElevenYellow, driven by a decentralized autonomous organization (DAO), has generated over six million dollars in sales since 2021, spread across approximately 150 works. The artist is neither human nor robot. It's a protocol. Collectors still buy.
Ai-Da: When the Creator Becomes the Selling Point
On November 7, 2024, an unexpected lot crossed the million-dollar threshold at Sotheby's. Estimate: $120,000 to $180,000. Final price: $1,084,800. The artist: Ai-Da, a humanoid robot equipped with cameras for eyes and an articulated arm wielding a paintbrush. A nine-fold factor between the low estimate and the hammer price—a direct echo of the Belamy effect.
The question of authorship, far from hindering sales, has become an accelerator. The very ambiguity of Ai-Da's status—machine or creator?—fuels the narrative that drives up bids. Collectors are not buying despite the absence of a human author. They are buying because of it. The legal vacuum surrounding copyright for AI-generated works, which we analyze in a separate part of this investigation, does not slow down transactions. It gives them a speculative flavor that some buyers actively seek.
Who Buys?
Data from Christie's "Augmented Intelligence" sale reveals a sharp shift in client demographics. Forty-eight percent of registered bidders belonged to the millennial and Gen Z generations. Thirty-seven percent had never bid at Christie's before—in any category.
This 37% figure is worth dwelling on. Auction houses operate on a base of recurring clients, loyal over decades, trained in the codes of an opaque market. AI art breaks this lock. It attracts an audience shaped by tech culture, familiar with NFTs and cryptocurrencies, indifferent to the genre hierarchy that structures the classical market.
The 2025 Art Basel/UBS report confirms the trend. Fifty-one percent of high-net-worth collectors report having purchased digital art. The share of digital art in collections has increased from 3% to 13%. Art Basel now ranks digital art as the third-largest spending sector.
However, the market is not naive. Eighty-two percent of surveyed collectors demand transparent labeling identifying AI use. Sixty percent express concerns about authenticity. Only forty percent believe in continued growth. Digital collecting is a fact. Trust still needs to be built.
Three Poles, Three Logics
The AI art market has no single center.
London remains the institutional pivot. Christie's held "Augmented Intelligence" there; Sotheby's sold Ai-Da there. The closure of Christie's dedicated digital department does not signal a retreat. It's a strategic choice: to integrate AI art into mainstream sales rather than confining it to a silo. The message is clear—AI art is no longer a curiosity; it's a revenue stream.
The Gulf is accelerating. Sotheby's inaugurated "Origins" in Riyadh in February 2025, with Anadol headlining at $900,000. The Saudi kingdom, which is investing heavily in AI through its Vision 2030 strategy, sees digital art as a vector of soft power. Gulf money offers AI artists what the Western market still denies them: legitimacy backed by high prices, without the moral reservations that accompany each auction in London or New York.
Asia is forging its own path. The Mori Art Museum in Tokyo, with the "Machine Love" exhibition, recorded record attendance. The Asian market embraces AI art without the hesitations of the Western debate—Japanese regulation, where Article 30-4 of the Copyright Act offers the world's most permissive framework for training AI models, contributes to this.
The Second Lung: NFTs
AI art does not circulate solely in the hushed halls of auction houses. The NFT market, valued at $3.6 billion in 2024, constitutes its parallel circuit. Thirty percent of new NFT projects incorporate an artificial intelligence component, according to the Blockchain Council. SuperRare, Foundation, Art Blocks: these platforms host a creation and sales ecosystem that bypasses traditional gatekeepers.
The boundary between the two circuits is more porous than it appears. Anadol built his reputation in the crypto ecosystem before integrating MoMA's collections. Beeple went from raw NFT to consecration at Christie's. Claire Silver took the opposite path, leaving decentralized platforms for WME. Artists move between worlds. So do collectors. Some buyers at the "Augmented Intelligence" sale came from the crypto universe—a demographic the traditional art market had not known how to reach.
Dissenting Voices
The market is growing. But it doesn't have unanimous support.
Jerry Saltz, New York Magazine critic and Pulitzer Prize winner, described Anadol's work as a "glorified lava lamp." The phrase circulated. Bids did not falter.
Structured opposition came from artists. More than 6,000 creators signed an open letter against Christie's AI sale in February 2025. Karla Ortiz, artist and plaintiff in the Andersen v. Stability AI lawsuit, summarized the sentiment in one sentence: "Christie's doesn't care about artists." The accusation targets the auction house in its legitimizing role: by selling AI art, Christie's normalizes a creative process built, according to its detractors, on the non-consensual exploitation of human works used to train the models. The link between this market and the social divide it fuels among creators is the subject of another part of this investigation.
Nicole Sales Giles, Head of Sale at Christie's, offers a different interpretation: "AI learns everything... to create something new. It's influence. Not theft." The formula has the merit of clarity. Not that of consensus.
The market moves faster than the law. US courts ruled in March 2025, in the case Thaler v. Perlmutter, that copyright requires a human author. The US Copyright Office specified in January 2025 that a simple prompt does not constitute a protectable act of creation. AI art sells, sometimes for a very high price. It benefits from no stable legal protection.
Institutions: Validation and Caution
The entry of AI art into museum collections has redrawn the map of its legitimacy. MoMA acquired Anadol's Unsupervised in October 2023. The Getty Museum integrated its first AI-generated photograph in January 2025—the same name as Getty Images, the image bank suing Stability AI for copyright infringement. The Centre Pompidou acquired NFTs. Each acquisition functions as a seal that the market brandishes in the face of skeptics.
The phenomenon goes further. In spring 2026, DATALAND, the first museum entirely dedicated to AI art, is set to open in Los Angeles. Its founder: Refik Anadol. The artist becomes an institution—a configuration reminiscent of Renaissance workshops, where the creator was also an entrepreneur and tastemaker. The artist who sells for the highest price builds the setting that will legitimize the next generation.
Thirty-five percent of fine art auctions now feature works incorporating an AI component, according to some estimates based on market data compiled for 2025. The figure signals a silent shift. AI art is no longer an event. It's a recurring line item in catalogs.
What Investors Need to Know
Behind the records and growth rates, three structural risks.
Legal risk. Nearly 60 AI-related copyright lawsuits were ongoing as of late October 2025. The Andersen v. Stability AI trial is scheduled for September 2026. Disney and Universal are suing Midjourney for $150,000 per work. An unfavorable decision could call into question the creative models that fuel part of the market. The European AI Act, fully applicable from August 2026, will impose transparency obligations on training data.
Technological risk. The "model collapse" theory, documented by Ilia Shumailov and his co-authors (Nature, 2024), postulates that AI models trained on their own output degrade. If this hypothesis proves true by 2028-2030, purely AI-created works could lose their appeal, while "augmented" art—involving a hybrid creative process—would gain relative value.
Liquidity risk. The market is young, concentrated on a handful of names, and its resale mechanisms remain embryonic. Will an Anadol purchased for $900,000 in Riyadh find a buyer at that price in five years? The NFT market, a natural liquidity circuit, has experienced a massive contraction since its 2021 peaks. The $3.6 billion in 2024 represents a fraction of what it was three years earlier.
A Market That Makes Its Own Rules
AI art has no established canon, no dominant critique, and no pricing grid based on decades of comparable transactions. It recruits its buyers where the traditional market did not look—forty-eight percent millennials and Gen Z among registered bidders at Christie's, thirty-seven percent first-time buyers. It is being built simultaneously in London auction rooms, Riyadh palaces, and decentralized platforms. It thrives in a legal vacuum that nearly 60 lawsuits are trying to fill.
Auction houses have made their choice. Christie's, by absorbing AI art into its general sales, is betting on normalization. Sotheby's, by exporting the format to the Gulf and Asia, is betting on globalization. Both strategies converge: AI art generates transactions, attracts new clients. The market will not wait for it.
For collectors, two categories of assets are emerging. Institutionalized artists—Anadol at MoMA, Herndon in TIME100, Silver at WME—offer a risk profile comparable to emerging artists in the traditional market, backed by multiple validations. Autonomous creations—Botto, Ai-Da, purely generative projects—represent a speculative bet on the very redefinition of the concept of authorship.
Spring 2026 will provide a first full-scale test. DATALAND opens in Los Angeles. Can the AI art market sustain a permanent institution, or does it remain dependent on events and records? The answer will determine whether the $40.4 billion projected for 2033 by Grand View Research is a forecast—or a prophecy.
This article is part of the "Artificial Intelligence and Creative Industries" dossier. See also: the analysis of the legal vacuum of generative AI (article 3) and the investigation into the social divide caused by creative automation (article 4).
