The medium survived

The global art market turned over about $59.6bn in 2025. It is a large number and it is below the 2022 peak of $67.8bn, after two years of decline and a small recovery. Underneath it, the middle of the market has been thinning for a decade.

The concentration is the part worth reading twice. At public auction, lots above $1m were under one per cent of everything sold and about fifty-four per cent of the value. So it is not really a $60bn market. It is a market for everybody, with a small and very expensive room attached, and most of the value is in the room.

The buyers in that room are old, and something over $80tn is expected to pass between generations in the next two decades. That is the structural fact under all of this, and it cuts one way: the people about to inherit did not grow up with auction catalogues.

Secondary trading in NFTs runs at roughly $2bn a year, down about ninety per cent from the 2021 peak. That is traded volume, not market capitalisation. The two get used interchangeably in this argument and they differ by an order of magnitude.

The three figures in the first two paragraphs come from the Art Basel and UBS report and are recorded in common/claims.ts as not yet checked against the published report by a human. Everything else on this page has a primary source that resolves in one click.

Centre Pompidou took in works by thirteen artists in 2023, including a CryptoPunk and an Autoglyph, framed by its curator as a continuation of the museum’s existing holdings in conceptual and video art rather than a break from them. LACMA received twenty-two generative and blockchain works from the collector Cozomo de’ Medici in February of the same year. ICA Miami holds CryptoPunk #5293, given by a trustee, which was the first NFT in any major museum collection.

The larger part of all three arrived as gifts rather than purchases, and a gift is not a price. Every version of this argument lists the acquisitions and leaves out how they were paid for. What a donation records is a curatorial judgement, which is a real thing and a different thing, and an argument that needs it to be a market signal is an argument that has not been checked.

The judgement is still the part that matters, because it is how photography got in too. The Buffalo AKG showed a survey of blockchain art in 2022. The same institution hosted what is generally counted the first major museum exhibition of photography in America, in 1910, roughly seventy years after the medium was invented and while people were still arguing about whether it was art.

The popular telling of this argument runs Impressionism from its rejection in 1863 to a record-setting auction in 1987 and calls the lag a hundred and twenty-four years. That endpoint is chosen to flatter the conclusion. Paul Durand-Ruel had built a working market for those painters by the 1890s, Renoir entered the Louvre in 1907, and Monet died wealthy in 1926.

Read against institutional acceptance, which is what the argument is actually about, the movements land in a band roughly thirty years wide.

The shorter number is the stronger claim. A hundred-year lag asks a reader to believe something they cannot check and would not live to see settled. A thirty-year one puts on-chain art around year eight, which is a position, not a prophecy, and it is falsifiable: if the institutional line goes quiet for five years, the reading was wrong.

The lineage this sits in, at the level of ideas rather than sales, is at the lineage room. LeWitt is the load-bearing figure in both. He published the sentence about the idea becoming a machine that makes the art in 1967, and his wall drawings are instructions executed by other people, which is a smart contract with a different runtime.

Pace opened a dedicated digital arm in November 2021 and gave Tyler Hobbs a solo show at its New York space in March 2023, in the room that has shown Rothko and Calder. Mint passes for his QQL project had sold for about $17m the previous September, in the middle of a bad market.

Pace represents the LeWitt estate. That is either a coincidence or the single most legible signal in this entire argument, and it is not a coincidence.

Sotheby’s and Christie’s both built dedicated venues for this work and kept them running through the downturn. The auction houses did not need the business. They are, along with the blue-chip galleries, the most reputationally cautious actors in the trade, and their entire product is being right about what will still matter in fifty years.

Everydays: The First 5000 Days sold at Christie’s in March 2021 for $69.3m. HUMAN ONE followed that November at about $29m. Pak’s The Merge took about $91.8m in December 2021, though that figure is not comparable to the others: it was sold in units to roughly twenty-eight thousand buyers rather than as one lot, and anyone quoting it beside a single-lot record is comparing two different things.

The one that carries the argument is Ringers #879, which sold at Sotheby’s in June 2023 for $6.2m. June 2023 was the floor of a bear market. That sale is not hype money, and it is the closest thing this category has to a clean reading of conviction with the speculation stripped out.

The votes keep arriving. When the Tel Aviv Museum of Art went looking for motion work for its screens, what it picked in 2026 was the same system, staged as a live software work rather than as stills; the display argument behind that show has its own page.

None of the three sinks the reading. All three are things a reader who knows the field would raise, and an argument that will not state its own strongest objection has not finished being made. The same page sits in front of the ask in every document we have written about the company, for the same reason.

The decade of attempts to make pictures liquid, and what broke each one, is at the art of code. What this gallery has actually chosen to hang, with an argument attached to each entry, is at the canon. What we hold, and where it might colour what we write, is at the disclosures.