December 2002

At the end of 2002, Intel Corporation occupied an unusual position.

The company remained the dominant supplier of microprocessors for personal computers. Its x86 architecture had survived Motorola, RISC processors, AMD clones, the rise and fall of numerous PC manufacturers, and several generations of computing technology. Microsoft Windows still overwhelmingly ran on x86. Hundreds of millions of users recognized the phrase “Intel Inside.”

But an Intel shareholder looking backward five years had little to celebrate.

Intel’s stock had closed 1997 at a split-adjusted $17.56. By 1999 it had risen to $41.16, and during 2000 it reached $75.81. By the end of 2002 it was back to $15.57. The investor had watched the stock rise more than fourfold and then lose almost 80% from its high.

The business had not disappeared.

Revenue was actually higher in 2002 than in 1997:

1997: $25.1 billion
2002: $26.8 billion

Yet operating income had fallen from $9.9 billion to $4.4 billion, and net income from $6.9 billion to $3.1 billion. Intel’s operating margin had fallen from roughly 39% to 16%.

The puzzle was larger than a stock-market bubble.

Intel had become one of the strongest technology franchises in modern business history. It had repeatedly reinvented itself, established the dominant PC architecture, built a powerful consumer brand, controlled much of the surrounding PC platform, and maintained roughly 80% of the world’s general-purpose microprocessor market.

How had that happened?

And why had a company with such an extraordinary competitive position produced almost no stock return for an investor buying at the end of 1997 and holding through 2002?

I. 1968–1980: The Memory Company Builds a Second Business

A company founded around memory

When Robert Noyce and Gordon Moore founded Intel in July 1968, the central idea was semiconductor memory.

Computers still relied heavily on magnetic-core memory. Moore believed advances in semiconductor manufacturing could replace core memory with smaller, faster, progressively cheaper integrated circuits.

Intel’s 1103 DRAM, introduced in 1970, validated the idea. By the end of 1971 it had become the world’s best-selling semiconductor device; by 1972, 14 of 18 major mainframe manufacturers in the United States, Europe and Japan were using it.

Intel went public on October 13, 1971, at $23.50 per share, raising $6.8 million. That year it generated $9.4 million of revenue and its first annual profit.

An investor buying the IPO was therefore not buying “the future PC CPU monopoly.”

He was primarily buying an innovative memory semiconductor company.

A second business appears

One month after Intel’s IPO, the company publicly introduced the 4004 microprocessor.

The processor originated from a project for Japanese calculator manufacturer Busicom. Instead of producing many specialized logic chips, Intel engineers designed a programmable general-purpose processor whose behavior could be changed through software.

Strategically, this was revolutionary.

Financially, it was initially small.

Intel spent most of the 1970s earning its money from memory chips and memory systems while gradually developing a second business around microprocessors, peripheral chips, development systems and software tools.

Exhibit 1 — Intel in the memory era

FY	Revenue	YoY	Net income	Net margin	R&D
1971	$9.4m	—	>$1m	~11%	—
1972	$23.4m	+148%	$3.1m	13%	$3.4m
1973	$66.2m	+183%	~$9m	~14%	—
1974	$134.5m	+103%	$19.8m	15%	$10.5m
1975	$136.8m	+2%	$16.3m	12%	$14.5m
1976	$226m	+65%	$25m	11%	~$21m
1977	$283m	+25%	$32m	11%	~$28m
1978	~$400m	+41%	~$44m	11%	~$41m
1979	~$661m	+65%	~$78m	12%	~$67m
1980	$855m	+29%	$96.7m	11%	$96m

Intel’s 1975 annual report makes the corporate identity clear. Management discussed expanding sales of memory systems while describing the need to “educate” customers about microcomputers—the former was an established business; the latter was still an emerging market.

Memory was still the larger business in 1980

The distinction is important.

By 1978–80, microprocessors had become important, but Intel had not yet become primarily a CPU company.

Contemporary industry estimates put Intel’s broad product-group revenue approximately as follows:

$m	1979	1980	1981
Memory Products Group	$376m	$406m	$284m
Microcomputer Group	$232m	$371m	$415m
Systems and other	~$123m	~$169m	~$169m
Eliminations	(~$70m)	(~$91m)	(~$79m)
Intel revenue	$661m	$855m	$789m

Thus, as late as 1980, memory remained Intel’s largest broad operating group.

But the two lines were almost equal:

Memory: ~$406 million
Microcomputer: ~$371 million.

Intel was approaching a crossover without yet knowing what that crossover would imply.

And there was an important complication inside “memory.” Intel had already begun deemphasizing some intensely competitive DRAM products and remained particularly strong in higher-value EPROM and programmable-memory products.

The problem was therefore not simply “all memory is bad.”

It was that standardized DRAM was becoming a fundamentally different economic business from proprietary microprocessors.

II. The Battle for the CPU: Intel Was Not Yet the Obvious Winner

Intel introduced the 8086 in 1978 and the lower-cost 8088 in 1979.

But Intel did not own the future of computing.

Motorola’s 68000 was a formidable competing architecture, while Zilog’s Z8000 and processors from National Semiconductor and Texas Instruments competed for 16-bit design wins. These were largely CISC-era competitors, offering different combinations of instruction-set sophistication, performance, memory addressing and system support.

Intel responded in 1979 with Operation Crush.

Rather than selling purely on chip performance, Intel mobilized:

software tools;
engineering support;
peripheral chips;
documentation;
salespeople;
advertising;
seminars;
a multigeneration product road map.

The campaign produced almost 2,500 design wins.

This was an early strategic lesson that Intel would repeat for the next two decades:

The processor did not have to win as an isolated chip.
The ecosystem around the processor had to win.

One of those design wins was IBM.

III. Turning Point I — 1981: Two Curves Cross

1981 was one of the strangest years in Intel’s history.

Financially, it was terrible.

Strategically, it may have been the most important year the company ever experienced.

Business One: memory economics break

Intel’s 1981 revenue fell 8%.

Net income fell 72%, from $96.7 million to $27.4 million.

Management identified one factor above all others: memory.

Intel reported that strong price erosion caused by industry overcapacity devastated the business. Despite record memory unit shipments, both memory revenue and memory pretax profit fell by more than $100 million from the prior year.

This distinction was fundamental:

Memory units ↑
Memory prices ↓ much faster
Revenue ↓
Profit ↓↓↓

Intel was discovering the danger of a product whose technological progress primarily translated into falling price per bit.

Business Two: microcomputers keep growing

Intel’s other business behaved differently.

The company reported that microcomputer-component revenue still increased during 1981, despite price declines in many products.

The two curves therefore crossed:

1980

Memory: $406m
Microcomputer: $371m

1981

Memory: $284m
Microcomputer: $415m

This is the first real turning point in Intel’s history.

Not because Intel decided in 1981 to abandon memory—it did not.

Rather:

1981 is when the old business ceased to be Intel’s economic growth engine at precisely the moment the new business received its most consequential external validation.

IBM chooses the 8088

On August 12, 1981, IBM introduced the IBM Personal Computer using Intel’s 8088 processor.

The immediate financial contribution was modest.

IBM sold only tens of thousands of PCs during 1981. Intel’s terrible 1981 financial results therefore cannot be described as the beginning of a PC-driven earnings boom.

The value of IBM was architectural, not yet financial.

IBM possessed immense credibility with corporate purchasers. If IBM standardized on Intel architecture, software developers had a stronger incentive to write for it. PC suppliers had an incentive to make compatible peripherals. Other manufacturers had an incentive to build compatible machines.

IBM’s relatively open hardware design would ultimately allow competitors to clone much of the PC.

Ironically, that hurt IBM more than Intel.

What customers increasingly needed to preserve was not an IBM-built computer, but compatibility with:

x86 + DOS, and eventually Windows.

IBM’s own historical account notes that the PC architecture became an industry standard.

IBM creates one of Intel’s future competitors

IBM was unwilling to depend completely on a single supplier.

Intel therefore entered a technology-sharing arrangement with Advanced Micro Devices, formalized in 1982, allowing AMD to second-source Intel processors.

The logic was sensible in 1981:

Intel needed IBM more than IBM needed Intel.

An exclusive CPU architecture with inadequate supply security could have lost the design.

The strategic cost would become visible years later.

Intel had effectively helped create a legally sanctioned x86 competitor in order to establish x86 as the industry standard.

That trade-off—share the architecture to accelerate adoption, then later try to reclaim its economics—became one of the central competitive tensions of Intel’s history.

IV. 1982–1984: From Design Win to Platform

It is tempting to describe 1981–84 simply as “the PC boom.”

That is too neat.

The correct chronology is:

1981: architecture win
1982: recession and ecosystem formation
1983–84: PC/x86 acceleration inside a broader semiconductor boom

1982: IBM invests before the earnings arrive

Intel’s revenue recovered to about $900 million in 1982, but net income remained only around $30 million.

The company was still struggling with weak semiconductor economics.

Yet IBM made a remarkable decision.

Late in 1982, IBM agreed to invest $250 million in Intel, initially obtaining approximately 12% of the company.

For IBM, the objective was strategic: ensure that a critical technology supplier possessed enough financial strength to continue funding new semiconductor technology.

The investment implied something important.

The processor relationship had moved beyond:

“Intel sells chips to IBM.”

It had become:

“IBM needs Intel to remain technologically and financially capable of executing the PC road map.”

V. 1983–1984: Now the PC Boom Shows Up in the Numbers

By 1983 the underlying market had changed dramatically.

Approximate IBM PC-family shipments illustrate the acceleration:

Calendar year	Approx. IBM PC-family units
1981	~20,000
1982	~140,000–180,000
1983	~500,000
1984	~1.2 million
1985	~1.4 million

But IBM was no longer the whole story.

Compaq and other manufacturers were producing IBM-compatible PCs. More than one million PCs using Intel’s 8088 architecture were reportedly shipped during 1983 alone.

A broader measure of North American business-oriented desktop computers shows the scale of the expansion:

Year	Units
1980	402k
1981	622k
1982	1.09m
1983	2.33m
1984	3.96m

The important economic event was therefore not merely that IBM sold more computers.

It was that:

IBM-compatible computing became a category.

Intel’s results reflected the transition

Exhibit 2 — The first PC cycle

FY	Revenue	Growth	Gross margin	Operating income	Net income	R&D	Capex
1980	$855m	+29%	~53%	$183m	$96.7m	$96m	$152m
1981	$789m	-8%	~42%	~$30m	$27m	$116m	$154m
1982	~$900m	+14%	~40%	~$28m	~$30m	~$131m	~$138m
1983	$1.12bn	+25%	~44%	~$139m	$116m	~$142m	~$145m
1984	$1.63bn	+45%	~46%	$250m	$198m	~$180m	$388m
1985	$1.36bn	-16%	31%	$(60)m	~$2m	$195m	~$236m

Intel itself described the strong 1983–84 semiconductor market and identified microprocessors and related components as its greatest growth area.

Was the boom really driven by PCs?

Partly—and increasingly.

It was not exclusively a PC boom. Industrial electronics, telecommunications, automotive products and other semiconductor markets were also strong.

But Intel was unusually leveraged to the fastest-growing segment.

Its 1984 report noted that both the surge in orders during the first half and the cancellations later in the year were heaviest in microprocessors and related components. It specifically identified PC and office-automation manufacturers as customers that had accumulated too much inventory.

This gives a much better interpretation of 1983–84:

A broad semiconductor recovery provided the macro cycle.
The explosion of IBM-compatible PCs gave Intel disproportionate upside.

And the product mix confirms it.

By the mid-1980s, Intel’s microprocessor, peripheral and microcontroller businesses were growing much faster than memory.

The corporate identity had already started changing before management formally acknowledged it.

VI. Turning Point II — 1985: Choose What Company You Are

In 1984 Intel’s financial performance looked exceptional.

Revenue rose 45%.

Net income rose 71%.

The company responded exactly as semiconductor companies often respond to shortages:

it built capacity.

Capex increased to almost $400 million, roughly twice annual net income.

Then the cycle reversed.

The semiconductor shortage becomes excess capacity

By late 1984, PC and office-automation growth slowed.

Customers that had stockpiled scarce chips suddenly had excess inventories.

Semiconductor supply caught up with demand.

Prices fell.

Intel’s 1985 revenue declined from $1.63 billion to $1.36 billion.

Gross margin fell from roughly 46% to 31%.

Operating income moved:

+$250 million → -$60 million

and net income nearly disappeared:

$198 million → ~$2 million.

Intel specifically cited intensified pricing pressure, industry excess capacity and aggressive Japanese competition in memory products.

Intel had seen this movie before.

The disturbing part was that the problem was no longer purely cyclical.

Japanese memory competitors change the strategic equation

DRAM competition increasingly centered on:

manufacturing yield;
production scale;
process discipline;
cost per bit;
capacity timing;
willingness to accept low margins to win share.

Japanese producers such as NEC, Hitachi, Toshiba and Fujitsu had become formidable competitors.

Intel could continue investing billions over time merely to defend its position in a market in which the product itself became progressively less differentiated.

Meanwhile, its microprocessor franchise exhibited almost the opposite economics.

Memory

More industry capacity
→ more supply
→ lower price/bit
→ weak differentiation
→ returns migrate toward the lowest-cost producer.

Microprocessor

Larger installed base
→ more software written for architecture
→ more OEM design wins
→ larger production scale
→ more R&D resources
→ better next-generation processors
→ still larger installed base.

The first was tending toward a commodity cycle.

The second had the potential to become a platform network effect.

Intel exits DRAM

In 1985, Intel decided to withdraw from DRAM.

The decision was psychologically difficult.

Memory was not a minor failed product line. It was the business Intel had been founded to pursue.

Gordon Moore and Andy Grove famously reframed the decision by asking what a new management team would do if it took over the company.

The answer was clear:

it would leave memory.

Intel’s later corporate history described the DRAM withdrawal as its first major strategic inflection point.

And management made another revealing choice.

Despite essentially zero net income in 1985, Intel increased R&D to roughly $195 million.

It was not cutting technology investment proportionately to current earnings.

It was reallocating investment toward the business it believed could create future economics.

The same year, Intel launches the 386

Intel introduced the 80386 in 1985.

The coincidence makes 1985 the second—and arguably decisive—turning point:

Intel exits the technology on which it was founded and launches the processor that will help establish its monopoly-like franchise.

1981 had told Intel where the future might be.

1985 forced Intel to choose that future.

VII. 1986–1992: Converting an Architecture Into a Franchise

Intel’s problems did not disappear immediately.

1986 was another difficult year. Revenue fell again to roughly $1.27 billion, restructuring continued, and the company remained under pressure.

But the competitive structure had changed.

Intel was no longer fighting mainly over commodity semiconductor capacity.

It was fighting over who owned compatibility.

From second sourcing to proprietary architecture

Early x86 processors had been licensed to other manufacturers because large customers demanded multiple suppliers.

Once x86 had become established, this arrangement looked increasingly unattractive to Intel.

Why allow AMD to obtain each new generation of the very architecture whose installed base Intel had spent years building?

With the 386, Intel became much more aggressive about controlling access to its intellectual property.

AMD fought back through arbitration, litigation, reverse engineering and eventually independent compatible designs.

The strategic issue was no longer:

Who can manufacture an Intel-designed chip?

It became:

Who controls the evolution of x86?

Intel increasingly did.

IBM loses control of the standard

Meanwhile, IBM’s own position weakened.

Compaq and other clone manufacturers learned to produce machines compatible with IBM PCs without depending completely on IBM.

This produced a remarkable redistribution of industry power.

IBM controlled less of the PC.

Intel controlled more of the architecture.

Microsoft controlled more of the operating system.

The industry increasingly became known as Wintel, not “IBM computing.”

This is one of the central strategic lessons of the case:

The company that creates a market does not necessarily own its most valuable layer.

IBM owned the original system design.

Intel and Microsoft owned the compatibility points that survived across manufacturers.

VIII. The Economics Begin to Change

Intel’s results recovered sharply.

By 1990:

revenue had reached about $3.9 billion;
gross margin was roughly 51%;
operating income was about $858 million;
net income was about $650 million.

Intel explicitly attributed its improving gross margin partly to a richer mix of higher-margin proprietary products.

By 1992, revenue had reached $5.84 billion, net income $1.07 billion and gross margin approximately 56%. Intel said its 486 family accounted for the majority of revenue and an even more substantial portion of gross margin.

Exhibit 3 — The franchise emerges

FY	Revenue	Net income	Approx. market cap	P/S	P/E
1975	$137m	$16m	$503m	3.7x	31x
1980	$855m	$97m	$1.76bn	2.1x	18x
1985	$1.36bn	~$2m	$3.45bn	2.5x	NM
1990	$3.92bn	~$650m	$7.81bn	2.0x	~12x
1995	$16.20bn	$3.57bn	$50.17bn	3.1x	14x
1998	$26.27bn	$6.07bn	$197.8bn	7.5x	~33x

The 1975–98 market-cap series shows a striking pattern: between roughly 1980 and 1995 Intel usually traded at only two to four times book value despite the franchise being progressively established.

The market did not immediately capitalize x86 as an enduring monopoly.

IX. 1991: Intel Inside Changes Who Has Bargaining Power

Intel still had one structural weakness.

Consumers bought:

IBM PCs, Compaq PCs, Dell PCs, Gateway PCs.

They did not necessarily buy Intel processors.

As long as the CPU remained an invisible component, OEMs could potentially substitute a compatible AMD or Cyrix chip and keep the savings.

Intel therefore did something unusual for a semiconductor supplier.

It advertised directly to consumers.

The Intel Inside cooperative campaign launched in 1991. Intel subsidized OEM advertising if computer manufacturers prominently identified the Intel processor.

By the end of 1992:

more than 500 OEMs had joined;
roughly 70% of eligible OEM advertising displayed the logo;
Intel’s initial commitment was approximately $250 million.

This was not merely branding.

It was an attack on OEM bargaining power.

Intel wanted the consumer to tell Dell or Compaq:

“I specifically want the Intel version.”

Once that happened, the processor was no longer a commodity input to the PC manufacturer.

It had become an ingredient brand with consumer pull-through.

X. 1993–1997: From CPU Supplier to PC Platform Owner

The Pentium arrived in 1993.

But by then Intel’s competitive advantage extended well beyond the CPU itself.

Intel influenced or controlled increasingly important parts of:

instruction-set architecture;
processor road maps;
chipsets;
motherboard reference designs;
bus standards;
development tools;
software optimization;
manufacturing processes;
OEM launch schedules.

The company’s fabs were particularly important.

Because Intel sold very high processor volumes, it could afford leading-edge manufacturing investments. Better manufacturing improved cost and performance. Better products increased volume. Larger volume funded the next generation of fabs.

Scale reinforced technology.

Technology reinforced scale.

The network effect around Windows

The software ecosystem reinforced the hardware economics.

More x86 PCs:

→ more developers write x86-compatible software.

More x86 software:

→ customers prefer x86 PCs.

More customers:

→ OEMs build more x86 machines.

More x86 machines:

→ Intel has more resources to invest in the next generation.

A competing architecture therefore did not simply have to produce a superior CPU.

It had to overcome the accumulated value of software compatibility.

The FTC would later identify exactly these barriers: enormous design and manufacturing costs, economies of scale, existing software investments, developer support and Intel’s reputation.

XI. AMD Changes Strategy

Intel never enjoyed competition-free dominance.

AMD’s strategy evolved with the industry.

Initially it was a licensed second source.

When Intel increasingly restricted access to new generations, AMD:

litigated for contractual rights;
reverse-engineered compatible processors;
eventually designed independently engineered x86-compatible CPUs.

Intel and AMD finally settled major portions of their technology dispute in 1995. AMD retained certain rights involving older generations but could not simply copy Intel’s future Pentium/P6 microcode.

AMD therefore had to become a real CPU designer, not merely Intel’s shadow manufacturer.

That raised AMD’s risk—but also eventually made it a more meaningful competitor.

XII. The Financial Machine, 1992–1997

The economics of Intel’s new business were radically better than those of the old one.

Exhibit 4 — Intel becomes the PC profit pool

FY	Revenue	Operating income	Operating margin	Net income	Diluted EPS*
1992	$5.84bn	$1.49bn	25.5%	$1.07bn	$0.16
1993	$8.78bn	$3.39bn	38.6%	$2.30bn	$0.33
1994	$11.52bn	$3.39bn	29.4%	$2.29bn	$0.33
1995	$16.20bn	$5.25bn	32.4%	$3.57bn	$0.50
1996	$20.85bn	$7.55bn	36.2%	$5.16bn	$0.73
1997	$25.07bn	$9.89bn	39.4%	$6.95bn	$0.97

*Restated for subsequent stock splits through 2000.

Between 1992 and 1997:

Revenue: 4.3×

Net income: 6.5×

Intel was not merely benefiting from PC growth.

Its margins were expanding because it had captured a more valuable layer of the industry.

XIII. 1997: How Expensive Was an Obvious Monopoly?

This question becomes important later.

By the end of 1997, virtually nobody could plausibly describe Intel as an obscure emerging company.

It had:

the dominant PC architecture;
a massive installed software base;
extraordinarily strong OEM relationships;
Intel Inside;
leading-edge fabs;
roughly $7 billion of annual profit.

Yet its end-1997 equity value was about $114 billion.

That implied roughly:

P/E = $114bn / $6.95bn ≈ 16.5x

P/S = $114bn / $25.1bn ≈ 4.6x

The competitive position looked spectacular.

The valuation, by later technology-stock standards, did not.

This would matter enormously to the investor who bought at that point.

XIV. 1998–2000: Competitors Fight Back

Intel’s dominance was strongest precisely when competitive responses intensified.

AMD’s K6 attacked the value segment.

AMD acknowledged how difficult the position was: because of Intel’s dominance and brand strength, AMD said it priced its K6 processors at least 25% below Intel products with comparable performance. Intel’s price moves therefore directly constrained AMD’s own margins.

By Q4 1998, AMD said the K6 family had reached approximately 16% of worldwide Windows-compatible processor units, more than twice its prior-year share. Yet AMD still lost more than $100 million during 1998.

Intel responded with Celeron, attacking the lower-priced PC market rather than allowing AMD to own a protected profit pool.

This demonstrated an important form of monopoly power.

Intel did not need to prevent competitors from selling chips.

It could make the segments they entered less economically attractive.

AMD’s Athlon becomes genuinely competitive

AMD’s next-generation Athlon represented a bigger challenge.

It was not simply a cheaper clone. In certain periods it offered performance competitive with or superior to Intel’s products.

By 2000 AMD estimated that it had roughly 17% of worldwide PC processor units.

But Intel still retained overwhelming scale, brand, OEM distribution, software compatibility and manufacturing resources.

Competition had become real.

Dominance remained intact.

XV. Intel Becomes a Regulatory Problem

By 1998, Intel’s market position was powerful enough to attract U.S. antitrust scrutiny.

The Federal Trade Commission said Intel’s own analyses indicated approximately 80% of worldwide dollar sales of general-purpose microprocessors. Other FTC analyses put Intel near 80% of units and almost 90% of dollar sales depending on the precise processor market definition.

The FTC also focused on something more subtle than processor prices.

OEMs depended on Intel for advance technical information and product samples so that they could design machines around future processors and launch them on schedule.

The FTC alleged that Intel had used access to such information in intellectual-property disputes with customers such as Digital Equipment, Intergraph and Compaq. Intel subsequently settled the case without admitting the substantive allegations.

The case revealed how Intel’s power had evolved.

1971

Intel supplied memory components.

1981

Intel supplied IBM’s CPU.

1990

Intel increasingly controlled x86 architecture.

1995

Intel created consumer demand for its CPU.

1998

Intel could influence when major computer manufacturers were technologically able to launch their next PCs.

Intel had moved from supplier to platform gatekeeper.

XVI. 1998–2000: The Business Improves—and the Valuation Improves Faster

Intel’s profits continued increasing despite AMD’s gains.

Exhibit 5 — The late-1990s boom

FY	Revenue	Gross margin	Operating income	Net income	Diluted EPS
1997	$25.07bn	60.3%	$9.89bn	$6.95bn	$0.97
1998	$26.27bn	54.0%	$8.38bn	$6.07bn	$0.86
1999	$29.39bn	59.7%	$9.77bn	$7.31bn	$1.05
2000	$33.73bn	62.5%	$10.40bn	$10.54bn	$1.51

One qualification is important: Intel’s 2000 net income included unusually large investment-related gains, so operating income gives a cleaner picture of the underlying semiconductor business. Still, the core economics were exceptional.

What changed even more dramatically was what investors were willing to pay.

XVII. Exhibit 6 — Intel’s Valuation Cycle

For consistency, the valuation ratios below use market capitalization / reported company financials, rather than combining raw historical stock prices with EPS from different split bases.

Date	Market cap	Revenue	Net income	P/S	P/E
1975 YE	$0.50bn	$0.137bn	$0.016bn	3.7x	31x
1980 YE	$1.76bn	$0.855bn	$0.097bn	2.1x	18x
1985 YE	$3.45bn	$1.365bn	~$0.002bn	2.5x	NM
1990 YE	$7.81bn	$3.92bn	~$0.65bn	2.0x	~12x
1995 YE	$50.2bn	$16.20bn	$3.57bn	3.1x	14x
1997 YE	$114.4bn	$25.07bn	$6.95bn	4.6x	16.5x
1998 YE	$196.5bn	$26.27bn	$6.07bn	7.5x	32x
1999 YE	$274.4bn	$29.39bn	$7.31bn	9.3x	37.5x
2000 peak	~$495bn	$33.73bn†	$7.31bn‡	~14.7x†	~68x‡
2000 YE	$202.1bn	$33.73bn	$10.54bn	6.0x	19x
2001 YE	$210.4bn	$26.54bn	$1.29bn	7.9x	NM/trough
2002 YE	$102.4bn	$26.76bn	$3.12bn	3.8x	33x

†Using eventual FY2000 revenue for perspective.
‡Using FY1999 earnings as the trailing earnings available entering the 2000 peak.

Historical 1975–98 market values are from the Intel valuation history assembled by Greenwald et al.; 1997 onward year-end market caps are independently consistent with reconstructed market-cap data.

The pattern is striking.

At the end of 1997, Intel was already an overwhelmingly dominant technology platform at only ~16.5x earnings.

By the end of 1999, investors paid ~37.5x.

Around the 2000 stock-market high, Intel’s equity value approached half a trillion dollars.

Thus, Intel’s final stock-market surge was not primarily caused by Intel suddenly becoming a better monopoly.

It was caused by the market assigning a much larger value to the monopoly it already knew existed.

XVIII. Exhibit 7 — The Stock Price Tells the Story

Intel split its stock repeatedly, including 2-for-1 splits in 1997, 1999 and 2000. All prices below are presented on a consistent split-adjusted basis.

Year	Year-end price	Annual change
1980	$0.42	—
1981	$0.23	-44%
1982	$0.40	+72%
1983	$0.88	+117%
1984	$0.58	-33%
1985	$0.61	+5%
1986	$0.44	-28%
1987	$0.83	+89%
1990	$1.20	+12%
1992	$2.72	+78%
1995	$7.09	+78%
1996	$16.37	+131%
1997	$17.56	+7%
1998	$29.64	+69%
1999	$41.16	+39%
2000 high	$75.81	—
2000 YE	$30.06	-27%
2001 YE	$31.45	+5%
2002 YE	$15.57	-50%

The investor experience from the end of 1997 is extraordinary:

$17.56 → $75.81 = +332%

then:

$75.81 → $15.57 = -79%

and ultimately:

$17.56 → $15.57 = -11% over five years.

There was no smooth five-year stagnation.

There was a spectacular late-cycle melt-up followed by a nearly complete round trip.

XIX. 2000–2002: Dominance Does Not Prevent a Downcycle

The Internet boom had encouraged enormous technology investment.

Businesses built networks.

Telecommunications companies ordered equipment.

PC manufacturers and corporations upgraded systems.

Semiconductor producers built capacity.

Then the cycle turned.

Intel’s revenue fell:

$33.7bn → $26.5bn in 2001, a decline of about 21%.

Reported net income fell:

$10.5bn → $1.3bn.

Some of that enormous earnings decline reflected investment gains and acquisition-related accounting, but the underlying operating deterioration was still severe.

Cost of sales rose from roughly 38% of revenue to more than 50%.

Factories designed to produce enormous volumes still carried fixed costs when those volumes disappeared.

By 2002 revenue stabilized at $26.8 billion and net income recovered to $3.1 billion, but Intel remained far below its peak profitability.

XX. The Great Paradox: Intel Did Not Lose Its Moat

This is the most important point in understanding the 2000–02 decline.

Intel did not suddenly lose the PC architecture.

Windows did not abandon x86.

PC manufacturers did not abandon Intel.

AMD did not take 70% of the market.

Intel remained the leading microprocessor supplier.

What disappeared was the assumption that peak demand, peak utilization and peak margins were permanent.

A moat answers:

Who captures the industry’s economics?

It does not answer:

How large will those industry economics be every year?

Intel could own most of the CPU profit pool while the CPU profit pool itself temporarily shrank.

XXI. Revenue Barely Changed; Profitability Did

Compare 1997 with 2002.

	1997	2002	Change
Revenue	$25.07bn	$26.76bn	+7%
Operating income	$9.89bn	$4.38bn	-56%
Operating margin	39.4%	16.4%	-23 pts
Net income	$6.95bn	$3.12bn	-55%
Market cap	$114bn	$102bn	-10%

This is why the five-year stock result cannot be explained simply by “multiple compression.”

In fact, using year-end reported earnings:

1997 P/E ≈ 16.5x

2002 P/E ≈ 33x

The multiple expanded because 2002 earnings were depressed.

What collapsed was the denominator.

XXII. Intel’s Five Strategic Transformations

1. 1968–1980 — Semiconductor memory company

Intel’s competitive advantage originated in semiconductor process innovation.

The business grew spectacularly, but increasingly standardized memory products attracted competitors and price declines.

The microprocessor was still the second business.

2. 1981 — Two businesses cross

Memory revenue and profit collapse despite record unit volumes.

Microcomputer revenue continues growing.

IBM chooses x86.

The old business deteriorates at precisely the moment the future business gains external validation.

This is Turning Point I.

3. 1982–1984 — A design win becomes an ecosystem

IBM’s PC sales accelerate.

Compaq and other clones expand the installed base.

DOS software reinforces compatibility.

Intel’s processor-related revenues grow much faster than memory.

The boom is partly a broad semiconductor cycle, but PCs increasingly explain Intel’s disproportionate growth.

4. 1985 — Intel chooses the new company

The capacity cycle breaks.

Japanese memory competition exposes DRAM’s structural economics.

Intel exits DRAM and launches the 386.

Resources move toward proprietary microprocessors.

This is Turning Point II.

5. 1986–2000 — Architecture becomes platform

Intel restricts second sourcing.

AMD becomes an independent compatible competitor.

Windows/x86 creates network effects.

Intel Inside creates end-user demand.

Chipsets, motherboards and development tools expand Intel’s control.

By the late 1990s, Intel does not merely supply CPUs.

It helps define what a PC is.

XXIII. Exhibit 8 — Competitive Landscape and Competitor Behavior

Era	Main competitive threat	Competitor behavior	Intel response
1968–75	Core memory, U.S. semiconductor firms	Compete on memory density/process	Rapid memory innovation; 1103
1975–80	TI, Mostek, NEC, Hitachi, Toshiba; Motorola in CPUs	Memory price/scale competition; Motorola wins CPU designs	EPROM differentiation; Operation Crush
1981–85	Japanese memory + AMD as x86 second source	Japan drives memory prices lower; AMD supplies compatible Intel processors	Let AMD second-source to satisfy customer needs; increasingly shift resources to microprocessors
1985–92	AMD clones; Motorola/RISC	Reverse engineer x86 or promote alternative architectures	Exit DRAM; protect 386/486; exploit software compatibility
1991–97	AMD, Cyrix, PowerPC, RISC	Lower-cost compatible CPUs or alternative architectures	Intel Inside; fabs; faster product cadence; platform components
1998–2000	AMD K6/Athlon	AMD discounts K6 ≥25%; attacks value then performance segments	Celeron, pricing response, manufacturing scale
2001–02	AMD Athlon/XP; PC downturn	Performance/value competition during weak market	Pentium 4, process investment, maintain R&D through downturn

AMD’s own SEC filings provide perhaps the clearest description of the asymmetry. AMD argued that Intel’s financial strength and market dominance allowed Intel to set x86 standards and alter pricing in ways that directly affected competitors’ margins.

XXIV. What Actually Created Intel’s Moat?

By 2000, Intel’s advantage contained at least six reinforcing elements.

Architecture

The installed x86 software base made switching costly.

Developers

Developers had powerful incentives to optimize for the largest installed platform.

OEM relationships

Manufacturers designed product road maps around Intel’s releases.

Manufacturing scale

Enormous processor volume funded leading-edge fabs and process transitions.

Platform control

Chipsets, motherboards, technical standards and advance product information expanded Intel’s influence beyond the CPU.

Brand

Intel Inside created consumer demand for a component users rarely physically saw.

No single element explains Intel’s dominance.

The moat came from the interaction among them.

XXV. The Investor’s Problem Was Different From the Strategist’s Problem

A corporate strategist examining Intel in 1997 might have correctly concluded:

Intel has an extraordinary competitive advantage.

A stock investor needed to answer two additional questions:

What level of earnings is sustainable?

and:

How much of the future is already reflected in the price?

Those questions became particularly difficult in 1999–2000 because both the business cycle and the valuation cycle pointed upward simultaneously.

At the 2000 high, an investor was implicitly capitalizing exceptionally favorable semiconductor economics at an exceptionally favorable multiple.

That combination—not the disappearance of Intel’s moat—created the eventual stock-market damage.

XXVI. December 2002: The Decision

By the end of 2002, Intel was valued at roughly $102 billion.

The stock had returned almost exactly to where it had traded five years earlier.

Yet much had changed.

Intel still controlled the dominant PC architecture.

AMD had demonstrated that technically credible competition was possible.

PC growth was slowing relative to the 1990s.

Intel continued spending billions of dollars on R&D and manufacturing.

The Internet had not disappeared.

Neither had personal computing.

But investors could no longer assume that the extraordinary earnings trajectory of the 1990s would repeat indefinitely.

The central question was therefore no longer whether Intel was a good company.

It plainly was.

The question was:

What were Intel’s normalized earnings—and what growth rate should an investor pay for after the extraordinary PC buildout of the 1990s?

Intel’s history had demonstrated three very different forms of value creation.

The first came from inventing a superior product: semiconductor memory.

The second came from owning a standard: x86.

The third came from capitalizing that standard at an increasingly high stock-market valuation.

Only the first two were controlled primarily by Intel.

The third belonged to the market.