technologybriefs
10:02in productionCh. 1 · The 1980 invention/ 10:02 · ceiling 15 min
Semiconductors · Hardware

Flash memory

Flash memory isn’t faster EEPROM — it’s a trade-off: erase speed for write inflexibility.

Flash memory is a block-erasable, non-volatile semiconductor memory invented by Fujio Masuoka at Toshiba in 1980. It builds on EEPROM but replaces byte-wise erasure with faster block-wise erasure — enabled by floating-gate transistor design. The name 'flash' was coined by colleague Shōji Ariizumi to reflect that speed. Masuoka and colleagues presented NOR flash in 1984 and NAND flash at the 1987 IEDM. Toshiba began marketing flash memory and commercially launched NAND flash — both in 1987. The first published implementation held only 8192 bytes but became the basis for larger-capacity versions. Flash does not support byte-level writes or in-place updates; erasure is all-or-nothing per block. It cannot replace RAM or true EEPROM where fine-grained modification is required. Flash changed how portable devices store firmware and user data, enabling compact, shock-resistant, low-power storage that scales without moving parts — a prerequisite for digital cameras, USB drives, SSDs and smartphones.

Chapters & takeaways6
  1. 1:00
    The 1980 invention

    Masuoka invented flash memory at Toshiba in 1980 using floating-gate technology and filed the foundational patent that year.

  2. 2:13
    Block erasure, not byte erasure

    Flash differs from EEPROM by erasing in blocks — not bytes — enabled by floating-gate design and named for its camera-flash-like speed.

  3. 3:28
    NOR and NAND emerge

    NOR flash debuted in 1984; NAND flash followed at the 1987 IEDM — two architectures solving different access and density problems.

  4. 4:37
    1987: from lab to market

    Toshiba began marketing flash memory and commercially launched NAND flash — both in 1987.

  5. 5:45
    Why 'flash'?

    The name 'flash' came from Shōji Ariizumi, who likened the bulk-erasure process to a camera flash — a metaphor for speed, not mechanism.

  6. 7:06
    From 8 KB to terabytes

    The first published implementation held only 8192 bytes — proof-of-concept density that scaled into terabyte storage.

Worth your time?

Yes. Study the whole thing.

4.5/ 5
What works
  • block-wise erasure
  • NOR flash for fast random reads
  • NAND flash for high-density storage
  • commercial deployment in 1987
What does not
  • support byte-level writes
  • enable in-place updates
  • replace RAM
  • replace EEPROM for fine-grained modification
Study it if
  • embedded-systems-engineers
  • storage-architects
  • semiconductor-designers
Skip it if
  • application-developers-who-assume-unlimited-write-endurance
  • systems-requiring-true-random-byte-access
The written brief1 min read

What it is and the problem it solves

Flash memory is a block-erasable, non-volatile semiconductor memory. It solves the problem of needing rewritable storage that retains data without power, while being cheaper and denser than EEPROM.

How it works

Flash memory uses floating-gate transistors to store data non-volatily, erasing content in blocks — not bytes — by applying voltage to remove charge from the gate.

What works

Block-wise erasure works reliably at scale. NOR flash enables fast random reads (suitable for code execution). NAND flash delivers high density and write speed (suitable for mass storage). Toshiba shipped both commercially in 1987.

What does not

It does not support byte-level writes or in-place updates. Erasure is all-or-nothing per block. It cannot replace RAM or true EEPROM where fine-grained modification is required.

What it changes

It changed how portable devices store firmware and user data: enabling compact, shock-resistant, low-power storage that scales without moving parts — a prerequisite for digital cameras, USB drives, SSDs and smartphones.

Is it worth your time

Yes, if you work with embedded systems, storage architecture or semiconductor design: it established the dominant non-volatile memory model for portable and persistent data, but its block-erase constraint remains a hard limit on write endurance and granularity.

Same field · Semiconductors4 of 18
10:06
Analog Devices1965Analog Devices is not a general-purpose chipmaker. It is a precision signal interface company. Its value lies in making ADCs and DACs that preserve fidelity across temperature, time, and voltage — not in speed, scale, or software. Its inventions shrink the gap between physical reality and digital representation — but only where that gap matters most.
10:16
Applied Materials1967Applied Materials is a semiconductor equipment supplier founded in 1967. It builds integrated manufacturing systems—not chips—but its machines define what is physically possible in chipmaking. Its Precision 5000 CVD platform introduced multi-chamber process integration in 1987. Its acquisitions of Orbot and Opal added inline inspection and metrology. It is headquartered in Santa Clara and ranks second globally by revenue. No performance specs, costs, or limitations are stated in the source material.
10:26
ASML1984ASML is the dominant supplier of photolithography machines for integrated circuit production. Founded in 1984 as a joint venture between Philips and ASM International, it inherited Philips’ stalled lithography project and initially had no market-ready product. Its first machine, the PAS 2000, failed commercially and technically. Success came with the PAS 5500 in the early 1990s. By 2002, ASML was the largest lithography supplier. It now leads in extreme ultraviolet (EUV) lithography—the only viable method for manufacturing the most advanced chips—and completed EUV machine development in the late 2010s. Competitors included Canon, Nikon, Ultratech, MKS Instruments, Lam Research, and Cadence Design Systems.
9:35
Commodore 64The Commodore 64 delivers audiovisual capability through custom silicon—not software or architecture. Its dominance came from cost-engineered integration of the VIC-II and SID chips, not scalability or abstraction. It changed what consumers expected from home computers—but left no direct lineage in modern computing stacks.
Up next in Technology

Fortran

John Backus · 1957 · 11:20

Fortran didn’t invent high-level programming — it proved it could pay its way in machine cycles.

11:20