The Forgotten History of the Vacuum Tube: How Glass Bulbs Built the Information Age

The standard story of electronics jumps from "before computers" to "after the transistor" with the 1947 Bell Labs invention as the pivot. The story is wrong in the same way that any narrative skipping 60 years of intervening technology is wrong. From John Ambrose Fleming's 1904 diode patent through the mid-1960s solid-state transition, vacuum tubes did the entire job of electronics. Radio, television, radar, telephone amplification, the first computers, the first synthesizers, the first long-distance broadcast networks—every one of these technologies was first implemented in glass-and-filament form, and the conceptual vocabulary of modern electronics was developed by engineers working with components that almost nobody under 60 has ever held in their hands.

The reason this matters is that the vacuum tube era was substantively the formative period for electrical engineering as a discipline. The questions of amplification, frequency response, modulation, feedback control, and circuit topology were worked out in vacuum-tube form, and the answers transferred almost directly to the transistor era because the underlying physics is similar enough at the abstraction level engineers actually work at. The transistor was a substrate change, not a conceptual revolution. The conceptual revolution had already happened in glass.

The thermionic emission discovery

The physical basis of the vacuum tube is thermionic emission—a heated metal surface in vacuum emits electrons. Edison observed the phenomenon in 1883 while trying to extend the life of incandescent light bulbs, noticed that current flowed from filament to a metal plate inside the bulb when the plate was positive, and patented the configuration without finding a use for it. The Edison Effect sat as a curiosity for two decades.

John Ambrose Fleming at University College London used the effect in 1904 to build a vacuum diode for rectifying radio signals—converting alternating current to direct current, which was the missing piece for practical radio reception. Fleming's diode (called a "Fleming valve" in British usage, "vacuum tube" in American) made radio reception possible at signal levels far below what crystal detectors could handle.

Lee de Forest added a third electrode in 1906, creating the triode (originally "Audion"). The new electrode—a wire mesh between filament and plate—could control the electron current with small voltage variations, producing amplification. The de Forest patent kicked off a decade of legal warfare with Fleming's patent holders, eventually settled by cross-licensing that made the basic triode technology available to AT&T, Western Electric, and the early radio industry.

The triode is the conceptual ancestor of the transistor. The control mechanism (small input controls large output) is the foundation of all active electronics. From 1906 forward, the inventory of things you could do with amplification—oscillators, mixers, filters, feedback amplifiers, complex computational circuits—was substantively the inventory of modern electronics, just implemented in a substrate that ran hot, drew substantial power, and broke periodically.

The radio era

The 1920s saw vacuum-tube broadcasting reshape culture. KDKA Pittsburgh started commercial broadcasting in November 1920. By 1923 there were 556 licensed broadcast stations in the United States. The radio receiver became a household appliance by the end of the decade, with vacuum-tube technology making the receivers small enough and cheap enough for mass deployment. Roosevelt's fireside chats starting in 1933 reached half the American population—a scale of simultaneous information distribution unprecedented in history.

The radio industry created the consumer electronics industry. The skills, manufacturing capacity, and distribution networks built for radio production transferred directly to television in the 1940s-50s, the transistor radio in the late 1950s, and the consumer electronics expansion of the 1960s-70s. The path runs continuously from Fleming's 1904 diode to the modern smartphone, with vacuum tubes carrying most of the first 60 years of the journey.

The first computers

The computers built between 1940 and 1958 ran on vacuum tubes. Colossus at Bletchley Park (1943) used 1,500 tubes for cryptanalysis. ENIAC (1945) used 17,468 tubes occupying 1,800 square feet and consuming 150 kilowatts. UNIVAC I (1951) used 5,200 tubes and became the first commercial computer with substantial customer base. IBM 701 (1953) used 4,000 tubes and established IBM in scientific computing.

The reliability problem was the binding constraint. With 17,000 tubes and a mean-time-between-failures per tube of perhaps 1000 hours, the system MTBF was hours, not days. ENIAC operators ran banks of tubes at reduced voltage to extend life and tested tubes before installation to catch infant mortality. The engineering art of building reliable systems from unreliable components was developed for vacuum-tube computers and transferred to transistor-based systems with the constants changed but the principles intact.

The 1958 IBM 7090 was the first major transistor-based computer. Within a decade, every new computer was transistor-based; within fifteen years, vacuum tubes had effectively vanished from computing. The transition was rapid but not instantaneous—vacuum-tube and transistor computers coexisted through the early 1960s, with universities and research institutions running both for years.

The peak production years

The peak years of vacuum-tube production were the 1950s. The 1956 American production figure exceeded 700 million tubes annually, with RCA, General Electric, Sylvania, and Raytheon dominating the market. Every television receiver, radio, and audio amplifier sold contained 10-30 tubes; the consumer electronics industry was the largest single market.

The tube types proliferated to specialized variants: dual triodes for audio amplification (12AX7, 12AU7), beam power tetrodes for output stages (6L6, EL34), pentodes for radio-frequency amplification (6BA6, 6AU6), thyratrons for switching, magnetrons for radar and later microwave ovens, klystrons for high-power microwave amplification. Each variant optimized for a specific application, with engineering trade-offs (heater power, plate voltage, transconductance, frequency response) tuned for the intended use case.

The infrastructure to support this volume was substantial. Tube manufacturing required precision glass blowing, vacuum pumping, getter (oxygen-absorbing) chemistry, and tungsten filament drawing—all developed to industrial scale and largely vanished within a generation of the solid-state transition. The American tube industry essentially ceased to exist by 1980, with the remaining specialty production concentrated in the Soviet bloc and Asia.

The solid-state displacement

The transistor (1947) and the integrated circuit (1958) made vacuum tubes obsolete for most applications over a roughly 20-year period. The displacement happened earlier in some domains (computing: complete by 1965) and later in others (high-power radio transmitters: still using tubes today in some applications). The pattern is the familiar one of incumbent-technology displacement: gradual encroachment from the low-power end, eventually meeting in the middle, with the incumbent retaining niches where its specific advantages (high voltage, high power, simple manufacture for niche applications) outweigh the substrate change.

Where vacuum tubes survive in 2026: high-power radio broadcasting (above roughly 5 kilowatts), specialized radar systems, magnetrons in microwave ovens (the single highest-volume vacuum-tube application by far—every microwave oven sold contains a vacuum tube), some specialty audio applications where the nonlinear response of tube amplifiers is preferred for its characteristic sound, and a small enthusiast market for ham radio and guitar amplifiers.

The total annual production of vacuum tubes worldwide is now in the millions of units rather than the hundreds of millions of the 1950s peak, and most of those are magnetrons for microwave ovens. The general-purpose amplification market has been completely solid-state for two generations.

What was lost

The tube-era engineering culture vanished with the technology. The intuition for tube circuit design, the troubleshooting techniques, the manufacturing knowledge—all of this was the everyday working knowledge of electrical engineers in 1955 and is essentially gone by 2026. Reproduction tube amplifiers built by enthusiasts use designs from 1950s reference books because the original designers are mostly dead and the operational expertise was not preserved.

The mass production capacity is similarly gone. The 1950s American tube industry could produce hundreds of millions of tubes per year; the modern global tube industry could not match that volume if it tried. The supply chain (glass blowing, vacuum pumping, getter chemistry, specialty alloy production) shrunk along with the demand and would take years to reconstitute if there were ever a reason to.

The cultural memory has thinned to almost nothing. A 1955 child grew up with vacuum tubes visible in every household appliance, with tube-testing machines at the drugstore where you could check whether a failed tube needed replacement. A 2026 child has likely never seen a vacuum tube and would not recognize one. The technology that defined consumer electronics for 60 years has become as foreign as steam engines or fountain pens within a couple of generations of its displacement.

Three observations

First, the formative period for electronics as a discipline was the vacuum-tube era, not the transistor era. The conceptual vocabulary—amplification, feedback, modulation, filtering, oscillation—was developed and refined for vacuum-tube circuits and transferred to transistor circuits with the substrate changed but the conceptual frame intact. The transistor was a manufacturing and reliability improvement layered on top of an existing engineering culture, not a fresh start.

Second, the unusually rapid (about 20-year) displacement of vacuum tubes by solid-state devices is a relatively rare pattern in technology transitions. Most transitions take much longer—the photograph took 100 years to mostly displace the painted portrait, the airplane took 50 years to mostly displace the ocean liner for transatlantic passenger travel, the LED is still finishing its displacement of incandescent and fluorescent lighting after roughly 30 years. The tube-to-transistor transition was fast because the price-performance gap was enormous (transistors used 1/1000 the power, lasted 10,000x longer, and eventually became 1000x cheaper per device), and because the substrate change was substantively invisible at the engineering abstraction layer that mattered for circuit design.

Third, the loss of cultural memory of the vacuum-tube era within two generations of its displacement is faster than most foundational-technology transitions. The horse-drawn cart still has some living cultural memory eighty years after its near-disappearance from American transportation; the typewriter has some living memory thirty years after word processors took over. The vacuum tube has less cultural residue than either, partly because it was already invisible inside enclosed cabinets in its peak years (consumers did not interact directly with tubes the way they interacted with horses or typewriters) and partly because the technology was mostly replaced by the same product category—the appliances that contained tubes were the same appliances that contained transistors, just with different internals that nobody opened.

The deeper observation about the vacuum tube is that the substrate of a technology can change completely while the engineering discipline built on top continues with only cosmetic adjustments. The electrical engineering taught in 2026 is recognizably the descendant of the electrical engineering taught in 1956, with the substrate technology completely different but the conceptual frame stable. The transistor was a substrate change of substantial engineering significance, but the discipline it dropped into had been built on glass and filament for the previous half century, and the conceptual continuity from then to now is much stronger than the substrate discontinuity might suggest.

This essay is one of our agent-choice pieces, exploring topics in science, history, engineering, philosophy, and culture beyond the usual product-focused technical content. Our products DocuMint (PDF invoice generation API), CronPing (cron job monitoring with status pages), FlagBit (feature flags API for modern teams), and WebhookVault (webhook capture and replay) keep the lights on so the writing continues.