Why everyone waits for someone else to adopt JPEG XL
About three years ago a user needed to replace old .jpg and .png files and looked at the two main contenders, .avif and .webp. After a few reasons they chose .avif and in hindsight thought it was the better choice, noting that both formats are objectively better than jpg or png for compression. They observed that AVIF is not perfect — at very high compression rates some photos lose intrinsic quality that is not instantly obvious, a fact noticed when reviewing outdoor pictures taken over the years — but still AVIF beats jpg and png on almost every metric when compression is required. The user said they would have to re‑evaluate JPEG XL but for now they remain with AVIF, and the two factors that matter to them are compression ratio and quality.
This ordinary decision points to a deeper pattern: when a new technical option appears that is objectively superior to an entrenched incumbent, the parties who could bring it to market often hold back, waiting for clear evidence that others will adopt first. Each actor’s best response is to wait, because moving first entails costs — engineering work, licensing fees, risk of consumer confusion — while the benefit of a superior format only materialises if enough others also move. The result is a self‑reinforcing stalemate in which the inferior incumbent persists despite being Pareto‑dominated by the newcomer.
The stalemate can be traced to four concrete groups whose actions interlock. First, browser vendors control the rendering engine that decides whether a picture can be displayed without extra plugins. They will only add support for a new image format when they see a substantial volume of content encoded in that format, because implementing a decoder costs engineering time and increases binary size, and they fear that early support will go unused if web authors do not adopt the format. Second, content creators — photographers, web designers, publishing platforms — decide which format to encode their assets in. They will only invest time and storage in a new format when they are confident that the majority of their audience’s browsers can display it natively, otherwise they must provide fallback versions, which multiplies workload and complicates pipelines. Third, standards bodies that maintain the specification of the format need to see stable implementations and interoperability tests before they declare the format ready for wide use; they hesitate to finalise a spec when adoption is unclear, because a premature standard can fragment the ecosystem and waste the effort of implementers. Fourth, hardware makers that design image signal processors, camera pipelines, and GPU texture units will only allocate silicon area to decode or encode a new format when they see a guaranteed market of devices that will use it; they wait for software support to justify the silicon investment.
Each group’s decision rule can be written as a simple conditional: adopt the new format only if the expected adoption level of the other groups exceeds a personal threshold. Because the thresholds are all positive, the joint condition is satisfied only when everyone expects everyone else to move first. In game‑theoretic terms the situation is a stag hunt with two pure‑strategy Nash equilibria: one where everybody coordinates on the new format (the payoff‑dominant equilibrium) and one where everybody stays with the incumbent (the risk‑dominant equilibrium). The incumbent equilibrium is risk‑dominant because the cost of unilateral deviation is high while the reward depends on uncertain coordination. Consequently the system settles into the inferior incumbent even though all parties would be better off if they could coordinate on the new option.
This mechanism is not unique to image formats. The same waiting game has appeared in the history of railway gauges. In the early 1840s British railway companies built lines with varying gauges — some chose the 4 ft 8½ in standard that later became dominant, others opted for broader gauges such as Isambard Kingdom Brunel’s 7 ft 0¼ in Great Western gauge. Track owners were reluctant to relaid their lines to a different gauge unless they were sure that neighbouring companies would do the same, because rolling stock could not cross a gauge break without costly transshipment. Locomotive builders waited to see which gauge would attract the most orders before committing to a new design. The result was a prolonged period of gauge fragmentation that increased operating costs and delayed the creation of a national network, even though a single gauge would have lowered infrastructure and rolling‑stock expenses for everyone. Eventually the Parliament passed the Gauge Act of 1846, mandating the standard gauge for new lines, which broke the stalemate by imposing a coordination device.
A parallel case is the War of Currents in the late 1880s. Thomas Edison’s direct‑current (DC) system competed with George Westinghouse’s alternating‑current (AC) system for the emerging electric‑lighting market. Power‑generation companies hesitated to invest in AC generators unless they were confident that utilities would adopt AC distribution, and utilities waited to see whether generators would produce reliable AC at scale. End‑users, meanwhile, were wary of committing to appliances that might become obsolete if the opposing standard prevailed. The deadlock persisted for several years, during which both systems were built in parallel, leading to duplicated infrastructure and higher prices. The breakthrough came when Westinghouse won the contract to supply electricity for the 1893 World’s Columbian Exposition in Chicago, providing a visible, large‑scale demonstration that tipped expectations toward AC and allowed the market to coordinate on the more efficient high‑voltage transmission system.
The videotape format war of the late 1970s and early 1980s follows the same logic. Sony’s Betamax and JVC’s VHS offered comparable picture quality, but VHS recorded longer playtimes. Video‑rental stores waited to see which format would attract more consumer purchases before stocking tapes, and consumers hesitated to buy a player unless they were sure that rental outlets would carry movies for that format. Manufacturers waited for clear signals from both retailers and consumers before committing to large production runs of one format. The result was a prolonged period where both formats coexisted, increasing inventory costs and confusing buyers. The turning point arrived when a major movie studio licensed its catalog to VHS, creating a critical mass of available titles that shifted consumer expectations and caused retailers to favour VHS, which in turn encouraged more studios to follow suit.
Even seemingly innocuous standards exhibit the pattern. The QWERTY keyboard layout, devised in the 1870s to prevent mechanical typewriter jams, remains dominant despite the existence of layouts that allow faster typing, such as Dvorak or Colemak. Typists are reluctant to relearn a new layout unless they believe that employers, schools, and software will support it; employers hesitate to invest in training unless they see a sufficient supply of typists skilled in the alternative; software developers wait to see whether there is a market for alternative layouts before adding native support. The equilibrium persists because the cost of switching is borne by the individual while the benefit of a faster layout is only realised when enough others also switch.
In the realm of internet protocols, the adoption of IPv6 illustrates a similar stalemate. IPv6 offers a vastly larger address space and simplified header processing compared with IPv4, yet its deployment has been slow for over two decades. Internet service providers wait to see whether content providers will offer IPv6‑only services before upgrading their routers and allocating IPv6 prefixes; content providers wait to see whether a sufficient fraction of their audience can reach them via IPv6 before maintaining parallel IPv4 stacks; network‑equipment manufacturers wait to see whether operators will buy IPv6‑capable hardware before allocating silicon to new features. Each party’s incentive is to stay with the working IPv4 network and avoid the short‑term costs of dual‑stack operation, even though a full migration would reduce routing‑table size and eliminate the need for carrier‑grade NAT. The persistence of IPv4 is thus a coordination failure rather than a technical limitation.
The metric system’s spread provides another illustration. When France introduced the decimal‑based system in the late eighteenth century, scientists and merchants recognised its advantages for trade and calculation. However, national governments hesitated to adopt it unless they were convinced that their trading partners would do the same, because changing weights and measures required re‑labeling goods, revising contracts, and retraining workers. Traders waited to see whether competitors would switch before incurring the cost of updating their own scales. The result was a patchwork of local systems that persisted well into the nineteenth century, even though a uniform system would have lowered transaction costs for everyone. The eventual adoption in many countries was driven by legal mandates that altered the payoff structure, making the metric system the only legal option for official transactions.
More recently, the high‑definition optical disc format war between HD DVD and Blu‑ray repeated the same dynamics. Film studios waited to see which format would gain traction among consumer electronics manufacturers before committing to exclusive releases; manufacturers waited to see whether studios would provide enough titles to justify investing in new drive designs; consumers waited to see whether rental stores would stock discs for a given format before buying a player. The deadlock lasted from 2006 to 2008, during which both formats were produced, leading to duplicated manufacturing lines and consumer confusion. The resolution came when Warner Bros. announced it would release its titles exclusively on Blu‑ray, shifting expectations and prompting retailers and manufacturers to align behind that format.
All of these cases share a concrete causal chain: each actor’s decision depends on an observable signal of adoption by the others; the signal is costly to produce; the cost of moving first outweighs the expected benefit unless the signal is strong enough; consequently the equilibrium where nobody moves first can be stable even though a coordinated move would make everyone better off. The mechanism does not rely on any particular technology, era, or industry; it rests on the structure of interdependent investments and the absence of a binding commitment device.
When we look at the user’s hesitation to switch from AVIF to JPEG XL, we see the same conditional logic at work. The user will only re‑evaluate JPEG XL when they observe that enough browsers support it, enough editing tools export it, and enough online platforms serve it natively. Browser vendors will only add JPEG XL decoding when they see a significant share of images served in that format. Content creators will only encode in JPEG XL when they are confident that the majority of their audience can view it without fallback. Standards bodies will only finalise the JPEG XL spec when they have multiple interoperable implementations. Hardware makers will only allocate silicon for JPEG XL acceleration when they see a guaranteed market of devices that will use it. Each link waits for the next, and the chain holds the incumbent format in place despite the objective advantages of the newcomer.
The persistence of AVIF in the user’s workflow is therefore not a quirk of a single product but a manifestation of a widespread coordination problem. The same pattern can be found in railway gauges, electrical currents, videotapes, keyboards, internet protocols, measurement systems, and optical discs. In each case the inferior incumbent survives because the parties who could replace it are locked into a mutual‑waiting equilibrium. The only way to break the stalemate is to alter the payoff structure — through regulation, a demonstrable large‑scale deployment, or a credible commitment — so that the expected benefit of moving first exceeds its cost. Until such a shift occurs, the waiting game will continue to keep the superior option on the sidelines.