MIDI at 40: how competitors agreed on a standard and accidentally built the future

1983 was a big year.

On 10 March, MTV played "Billie Jean" — and by the end of that month Michael Jackson was the first artist in the channel's heavy rotation, which is the moment the most powerful outlet in music stopped pretending its format had nothing to do with race.

In December, Jonathan Demme took his cameras into the Pantages Theatre in Hollywood and filmed Talking Heads across several nights. That became Stop Making Sense — still, four decades on, the concert film every other concert film gets measured against.

(It's also the year John Mayer's "83" is set in, though he wrote it eighteen years later, looking back at his old neighbourhood through the eyes of his six-year-old self. Which is its own comment on 1983: a year big enough that people were still writing songs about being small in it.)

It also happens to be the year I was born.

But for all the music that came out of 1983, the most consequential thing to happen that year was a specification document — agreed between companies who were actively trying to put each other out of business.

MIDI — Musical Instrument Digital Interface — was worked out between 1981 and 1983 by companies that were competing with each other for the same customers. Sequential Circuits in California, led by Dave Smith, and Roland in Japan, founded by Ikutaro Kakehashi, brought Yamaha, Korg and Kawai into the discussion in October 1981 and agreed on a shared interface that would let their rival synthesisers talk to one another — with no patent, no licence fee and no owner. It was first demonstrated publicly at the Winter NAMM show in January 1983, when a Sequential Prophet-600 was connected to a Roland Jupiter-6, and the MIDI 1.0 Specification is dated 5 August 1983. It has survived more than four decades essentially unchanged, and the reason is that nobody ever owned it. MIDI 2.0, ratified in January 2020, adds far higher resolution and two-way communication while remaining backward-compatible.

That version is accurate. It is also the version that makes the whole thing sound inevitable, which it emphatically was not.

Rivals with a shared problem

By 1981 the polyphonic synthesiser had stopped being a laboratory curiosity and become a product category, and the category had a problem it was actively making worse. Every manufacturer used its own control voltage scheme, its own trigger logic, its own arrangement of sockets. If you owned two synths from two companies, you owned two instruments. If you wanted to play them together, you played them together with your hands.

In June 1981, at a NAMM show, Roland's founder Ikutaro Kakehashi put the idea of an international standard for synthesiser communication to Tom Oberheim of Oberheim Electronics. Oberheim took it to Dave Smith at Sequential Circuits. By October the three of them were talking it through with representatives of Yamaha, Korg and Kawai — which is to say, with essentially the entire competitive field.

Smith already had a proposal in hand. That same month he and his Sequential colleague Chet Wood presented a paper at the 70th Audio Engineering Society Convention in New York — AES Convention 70 ran from 30 October to 2 November 1981 — titled "The USI, or Universal Synthesizer Interface". It proposed a serial link running at 19.2 kBaud over standard quarter-inch phone jacks, connecting synthesisers, sequencers and home computers.

It is worth being precise about what happened next, because the popular telling flattens it. The USI was not MIDI. It was one company's proposal, and the Japanese manufacturers argued it into something else — a different data rate, a different connector, a broader message set. What arrived in 1983 was the residue of a negotiation between rivals, not a single inventor's design handed down. That is unusual, and it matters, because negotiated standards tend to be the durable ones. Everyone at the table had already lost the argument they most wanted to win, which meant nobody had a proprietary reason to defect later.

Anaheim, January 1983: the cable nobody had tested

Winter NAMM, the American music trade show, is a floor full of noise and competing demonstrations. At the Sequential Circuits booth in January 1983, Dave Smith and John Bowen had a Prophet-600 — Sequential's new polysynth, and the first instrument to ship with the interface built in. Kakehashi brought over a Jupiter-6, Roland's answer to it, not yet released.

Here is the part that gets skipped. Sequential had implemented the interface, tested it, and confirmed it worked — between a Prophet-600 and another Prophet-600. Smith has since said plainly that he did not know whether it would work with Roland's machine, because nobody had ever tried it. Two engineering teams on opposite sides of the Pacific had each read the same document and each built their own hardware against it, without ever connecting the two.

So they plugged the cable in, in front of gathered press, on a trade show floor, with no rehearsal and no fallback.

It worked. Play a key on the Prophet and the Jupiter sounded. That is the entire demonstration, and it is why we are still talking about it — not because it was spectacular, but because for the first time two instruments from two competing companies did the obvious thing that nothing had done before. If it had failed that morning, in public, it is genuinely hard to know what would have happened to the momentum. The specification was still months from publication.

Why give it away

Now the part I find most interesting, and the part most retrospectives glide over on their way to admiring the compatibility.

Five companies had, between them, produced a piece of infrastructure that every one of their customers wanted. The standard commercial move — then, and conspicuously now — is to patent it, brand it, license it, and use compatibility as a moat. Make your competitors either pay you or lock themselves out. The tech industry has run this play so many times since that it barely registers as a choice anymore.

They didn't. In October 1983 the Japan MIDI Standards Committee affirmed that MIDI "should remain open to the public", with manufacturers working collaboratively on future changes. Stewardship went to bodies with no product line to protect: the International MIDI Association from August 1983, and then the MIDI Manufacturers Association, established in 1985 by the original developers of the 1.0 specification. Not a company. A committee. The organisation that maintains the standard today, the MIDI Association, is a non-profit trade body representing over 100 member companies.

The romantic reading is that these were generous people. Some of them may well have been. But the commercial logic is stronger than the sentiment, and Smith articulated it clearly enough over the years: a standard is worth nothing without critical mass, and no single company can compel its rivals to adopt a protocol it owns and charges rent on. Owning MIDI would have meant owning a thing nobody used. Giving it away meant everybody used it — including, crucially, everybody's customers, who could now buy your synth without abandoning their existing rig.

Restraint, in other words, was also self-interest correctly calculated over a long enough horizon. That is a rarer skill than generosity, and a considerably rarer one in 2026 than it sounds.

Slow on purpose, and right by accident

The technical specification looks, from here, almost comically modest. MIDI 1.0 runs serially at 31.25 kbit/s in 8-N-1 asynchronous format — one start bit, eight data bits, no parity, one stop bit — for a ceiling of roughly 3,125 bytes per second. Most values in the protocol are 7-bit: note number, velocity, controller position, all of them landing somewhere in 128 steps from 0 to 127. Your entire dynamic range, on a 1983 synth, is 128 rungs on a ladder.

Modern ears hear that as a limitation. It was a specification. In 1983 the microprocessors inside these instruments were 8-bit parts running at a few megahertz and already busy generating sound. Feeding them a faster or wider data stream would not have produced better music; it would have produced instruments that could not keep up with their own inputs. Asked in 2013 what he would change about the original specification, Smith's answer was "Nothing! 30 years of version 1.0 pretty much says we got it right", and he attributed the perennial complaints about MIDI timing to the sluggish processors of the era rather than to the protocol.

This is the accidental future in the title. Because MIDI was designed exactly to the limit of what cheap 1983 hardware could handle, it was also cheap enough to put in everything, and simple enough that nobody could plausibly claim their implementation was the correct one. A working DX7 from 1983 can still be driven from a 2026 DAW through an interface that costs less than a set of strings — because the protocol never grew a version its old hardware couldn't parse. The constraint that looks primitive is the same constraint that made forty-three years of compatibility possible.

There is a version of this that goes wrong, and we have all lived through it: the format that arrives feature-rich, expensive to implement, controlled by one vendor, and dead within a decade. MIDI's poverty was its immune system.

You notice it mostly by not noticing it. Pick up a controller built this decade and it will very likely carry a USB port and a five-pin DIN socket on the same panel — a 1983 connector and a modern one side by side, both live, and nobody standing there wondering which is doing the work.

That invisibility is the point. Infrastructure that works is infrastructure you stop seeing, which is also why insuring the gear itself gets more thought than the cable joining it together.

Everything else from 1983 in that opening list needed the year to be remembered. MIDI is the one that stopped needing to be.

Two rivals, one award

On 9 February 2013, at the Special Merit Awards ceremony ahead of the 55th Annual GRAMMY Awards, the Recording Academy presented Dave Smith and Ikutaro Kakehashi with a joint Technical GRAMMY for the development of MIDI. Thirty years on, the two men who had plugged a cable between their competing synthesisers in Anaheim stood on the same stage for it.

Neither had long. Kakehashi died on 1 April 2017, aged 87; Smith died on 31 May 2022, aged 72. What strikes me is not the sentiment of it but the timing: they lived long enough to watch their compromise outlast nearly every product either company shipped, and to be told so out loud. Engineers who build durable infrastructure rarely get that. Most standards work is thanked, if at all, posthumously — ask anyone who has read up on the patent dispute behind the metronome.

What MIDI 2.0 actually changes

MIDI 2.0 was adopted unanimously by MMA members at the association's annual meeting on 19 January 2020, during Winter NAMM in Anaheim — the same show, thirty-seven years on. Five core specifications were ratified together, covering MIDI Capability Inquiry, the Universal MIDI Packet and MIDI 2.0 protocol, and common rules for profiles and property exchange.

Set against the 1983 numbers, the changes are substantial. MIDI 2.0 replaces the old two-and-three-byte messages with Universal MIDI Packets of 32 to 128 bits, lifts resolution from 128 values to as many as 65,536, adds per-note controllers for genuinely independent expression across a chord, and makes the connection bidirectional: with MIDI-CI, two devices negotiate what they can each do instead of assuming. Backward compatibility is handled by translation between Universal MIDI Packets and MIDI 1.0 messages, so old gear stays in the chain.

Adoption is the part worth being honest about. The plumbing is real: Logic Pro added MIDI 2.0 support in version 10.6, Apple shipped OS-level support in October 2021, Android 13 added it in 2022, and Microsoft has been rolling out Windows MIDI Services in Windows 11. Hardware exists. But the MIDI Association's own February 2026 update frames the current moment as a transition "from architectural vision to practical implementation" — with Apple, Avid, Bitwig and Steinberg jointly building open-source tooling to carry MIDI 2.0 into plug-in formats through JUCE, and further profiles and transports still in progress. Six years past ratification, that is a standard shipping steadily rather than a standard that has arrived.

Which is roughly what you would expect, and arguably what the 1983 decision guaranteed. A protocol nobody owns cannot be mandated by anybody. It spreads only when enough independent parties decide, separately, that it is worth implementing. That is slower than a corporate rollout, and it is also the reason the last one lasted forty-three years.

Common questions

Who invented MIDI and in what year?
MIDI was developed jointly rather than by one person. Dave Smith of Sequential Circuits and Ikutaro Kakehashi of Roland are credited as its principal architects, with Tom Oberheim, Yamaha, Korg and Kawai involved in the 1981 discussions. The first cross-manufacturer demonstration took place in January 1983, and the MIDI 1.0 Specification document is dated 5 August 1983.

Why was MIDI made free and royalty-free?
In October 1983 the Japan MIDI Standards Committee affirmed that MIDI should remain open to the public, with manufacturers collaborating on future changes. The commercial logic reinforced the principle: a standard is worthless without critical mass, and no single company can compel its rivals to adopt a protocol it owns and charges for. Open was the only route to universal.

What instruments were used in the first public MIDI demonstration?
A Sequential Circuits Prophet-600 was connected to a Roland Jupiter-6 at Winter NAMM in Anaheim in January 1983. Sequential had only ever tested a Prophet-600 against another Prophet-600, so Dave Smith has said he did not know beforehand whether two competing manufacturers' machines would actually communicate. They did, in front of press, unrehearsed.

What is the difference between MIDI 1.0 and MIDI 2.0?
MIDI 1.0 is one-way, runs at 31.25 kbit/s, and uses 7-bit values — 128 steps of velocity or controller resolution. MIDI 2.0 uses Universal MIDI Packets of 32 to 128 bits, supports up to 65,536 values, adds per-note controllers, and is bidirectional: devices negotiate capabilities through MIDI Capability Inquiry rather than assuming what the other end supports.

Is MIDI 2.0 backward compatible with old MIDI gear?
Yes. MIDI 2.0 translates between Universal MIDI Packets and MIDI 1.0 messages, and a MIDI 2.0 device that gets no response to a capability inquiry simply behaves as a MIDI 1.0 device. A 1983 synthesiser with five-pin DIN sockets still functions inside a MIDI 2.0 signal chain, which was a deliberate design requirement rather than a courtesy.

Did Dave Smith and Ikutaro Kakehashi win a GRAMMY for MIDI?
Yes. The Recording Academy presented the two of them with a joint Technical GRAMMY Award for the development of MIDI at the Special Merit Awards ceremony on 9 February 2013, ahead of the 55th Annual GRAMMY Awards. Kakehashi died in April 2017 and Smith in May 2022, both having seen the standard's endurance formally recognised.