Which businesses to be in, what the centre adds, and where a chip designer’s boundaries should stop
The preceding sessions examined how a firm creates and sustains advantage. Corporate strategy asks two different questions: which
businesses the firm should be in, and how the corporate office should manage them. Porter studied 33 large US
companies from 1950 to 1986 and found that most had divested many more acquisitions than they kept (1987, p. 1).
Corporate strategy must make the whole worth more than its parts; Porter’s historical sample challenges easy claims of synergy.
NVIDIA designs chips it does not make. It bought Mellanox for almost $7 billion, tried and failed to buy ARM for
$40 billion, and now sells 375-pound AI systems and rents cloud servers, in competition with the customers that buy
its chips (case, pp. 5–7). Today you put those moves through Porter’s tests and through transaction-cost
logic, and decide where its boundaries belong.
01What does corporate strategy decide that business strategy cannot, and why can diversification destroy value?
02When does owning a business add value: Porter’s three tests, his four concepts, synergy and core competencies?
03Where should NVIDIA’s boundaries sit: which businesses belong inside, and what should it make, buy or ally for?
Facts from the Yoffie and von Bargen case (set in late 2024) are marked Case;
where this page reads the case through a framework, the card is marked Analysis.
Porter page numbers refer to the 32-page HBR.org print of his 1987 article; textbook pages to Rothaermel, 6th edition.
Session 9 · your learning targets
Learning outcomes
By the end of this session, you should be able to:
Distinguish vertical integration, diversification and a contractual relationship by locating the activity and ownership boundary. Locate the firm’s boundary →
Apply attractiveness, cost-of-entry and better-off tests separately, identifying evidence that is still missing. Run Porter’s three tests →
Explain a concrete corporate value mechanism through restructuring, skill transfer or shared activities rather than a generic synergy claim. Name the parenting mechanism →
Compare parenting opportunities and harmful misfit with the contribution of an alternative owner, and justify a provisional portfolio judgment. Test parenting advantage →
Defend a make, buy or ally recommendation for a NVIDIA activity using the 2024 case, contracting hazards, capability gaps and a downside condition. Evaluate a boundary decision →
Use the readings, case evidence and practice to explain your reasoning. Browser checks provide practice feedback; formal assessment follows the course outline and instructor guidance.
Recall → connect → carry forward · Sessions 1–8
The strategy story so far: Sessions 1–8
Session 9 changes the question from how one business competes to which activities and businesses it should own. The answer still depends on every earlier lesson. Trace Sessions 1-8 through Nila Foods, a fictional Indian meal-service business: its focused promise leads to a purpose and target, external and internal tests, a value-creation argument, a competitive position and a possible complement ecosystem. None of those steps automatically makes Nila the right owner of another business.
One continuing classroom story. Nila Foods is a fictional Indian meal-service for working adults. Follow its choices from Session 1; later expansion ideas are proposals to test.
Operational effectiveness improves similar activities; strategy requires positioning, trade-offs and fit. State objective, scope and advantage, then compare intended plans with emergent action using evidence.
Nila Foods: Nila proposes preordered weekday meals on clustered routes, declines instant citywide delivery and studies actual customer behaviour.
Purpose and values guide what endures, while a quantified ambition opens a same-date gap from the baseline; closing that gap requires external and internal evidence.
Nila Foods: Nila separated easing meal burdens and honest food-safety disclosure from its growth target, then compared that target with its projected baseline.
Industry boundaries, PESTEL and five forces connect external change to price, cost and investment; complements and possible responses then raise questions about the wider game.
Nila Foods: Nila traced ingredient-price pressure, supplier bargaining, lunch substitutes and easy entry before asking whether a neighbouring seller could also help it.
Value nets and PARTS show how co-opetition changes added value, while uncertainty shapes commitments; credible moves therefore require examining internal strengths.
Nila Foods: Nila considered sharing a collection point with a rival and testing the rules through a limited pilot while demand remained uncertain.
Resources and capabilities support advantage when they improve buyer benefits or relative costs and pass tests against rivals and substitutes; value-chain analysis then checks their economic contribution.
Nila Foods: Nila examined kitchen routines, route know-how and trust, asking which linked activities rivals could imitate and whether it could retain their benefits.
Use the value stick to distinguish value created from value captured. Test shared value against both business and social effects, including burdens shifted elsewhere.
Nila Foods: Nila asks whether demand forecasts and reusable containers reduce waste and cost, including the return and cleaning work.
Porter distinguishes cost leadership, differentiation and focus; Kim and Mauborgne test value innovation; Ghemawat and Pisano examine threats, commitments and capabilities.
Nila Foods: Nila compares dependable delivery and relative cost within its chosen segment. Its value curve and four-actions proposal must demonstrate new demand and viable economics before supporting a blue ocean claim.
Eisenmann, Parker and Van Alstyne test platform participation and pricing; Bower and Christensen distinguish disruptive trajectories from sustaining improvements.
Nila Foods: To extend its offer, Nila asks whether customers and nutrition creators could benefit from one another. An experiment tests complement quality, retained use and cross-side effects; success alone would not establish disruption.
Why this leads to Session 9
A useful complement establishes a reason to coordinate, not necessarily a reason to own. If Nila considers owning its packaging stage, that is a vertical-integration question; a separate nutrition business raises diversification questions. Apply attractiveness, cost-of-entry and better-off tests separately. Specify the activity or skill creating the benefit, identify contracting hazards and capability gaps, then compare Nila's net parenting contribution with a specialist owner and contractual alternatives.
Retrieve it before moving on
Tell Nila's story in order from Session 1 through Session 8, then assess a proposal to own a packaging supplier. Which earlier decisions must the ownership argument preserve, and what would remain unproven even if the supplier served an attractive market?
Compare your reasoning with a model answer
Nila chooses dependable weekday meals for working adults, supported by preorders and clustered routes. It sets purpose, a dated ambition and a same-date value gap. Five forces challenge the economics; co-opetition and uncertainty turn office-partner ideas into experiments with anticipated responses. The internal audit tests forecasting and routing routines against rivals. Activity and shared-value analysis ask whether those routines improve economic value and reduce waste. Business strategy tests the resulting relative value and cost position. Platform analysis then asks whether nutrition creators could reinforce customer participation, without presuming network effects or disruption. Packaging ownership must strengthen this focused system. An attractive supplier market clears only one test: Nila must still justify the entry cost and a concrete better-off mechanism, show that its capabilities fit the unit without harmful interference, and explain why ownership beats contracts or a better specialist parent.
Read the argument · compare the lenses
Value creation is necessary. Being the best owner is harder.
Does the scope move create more value than it costs?
Attractive industry economics, an entry price that leaves room for returns, and a concrete better-off mechanism are separate tests. Passing only one is not enough. Corporate strategy should reinforce the competitive position of its businesses, rather than merely collect businesses under one name.
Distinguish a portfolio manager, a restructuring parent, a transfer of skills, and shared activities. Each has a different reason for ownership and a different coordination burden. A skill transfer may create a one-time gain; a shared activity may need continuing coordination. Ask when ownership adds something a contract would not.
Case question: For NVIDIA’s networking, systems or cloud activities, which specific activity creates the better-off effect? What do the firm’s customers or partners give up as NVIDIA’s scope expands?
Could another parent do more with the same business?
Parenting advantage asks whether this parent creates more net value than alternative owners. Match its capabilities to a business’s improvement opportunities, while testing where its habits could damage that business. Product similarity alone is not enough.
Heartland: useful parenting and little harmful misfit. Ballast: understood, but little scope to help. Value trap: an apparent improvement opportunity with dangerous misfit. Alien territory: weak opportunity and high misfit. Edge of heartland: promising fit with unresolved gaps. Treat these as evidence-led judgments, not permanent labels.
Case question: If NVIDIA has strong chip and software capabilities, what additional capabilities would it need to operate a very different downstream business? Compare with a specialist alternative owner.
Core question · who should own the business?
A shared capability is a starting hypothesis
Make the ownership argument specific. List the unit’s critical success factors, identify an improvement opportunity, and name the parent’s relevant skill. Then look for a harmful mismatch: control systems, time horizons, channels, talent practices or customer relationships. The parent’s contribution must survive both the cost of coordination and comparison with another owner.
Vertical integration: adjacent stages
Owning an upstream input or a downstream route changes the boundary along a value chain. Compare supplier hold-up and coordination problems with the capital, incentive and operating burdens created inside the firm. Long contracts or alliances may secure some benefits without full ownership.
Diversification: additional businesses
Related diversification needs a transferable competence or a valuable shared activity. Unrelated scope needs another defensible parenting mechanism. Reduced volatility alone does not show that the corporate parent improves the businesses.
The NVIDIA case lets you contrast a designer’s dependence on manufacturing partners, the extension from chips into systems, and the use of a common computing platform across markets. These are different scope choices; do not give them one generic “synergy” score.
The first part of the course was business-level strategy: how one business wins in one industry. Rothaermel puts the
switch in one line: business strategy is about how to compete in a single product market, corporate strategy is about
where to compete (2024, p. 298). The tools you already have do not retire. Each becomes one of Porter’s tests.
Sessions 3 and 4 · industry structure
The five forces become the attractiveness test
Porter’s first test for any new business is the analysis
you ran in Session 3
and Session 4: is the
industry’s structure good enough to earn more than the cost of capital, or can the entrant make it so
(1987, p. 8)?
Session 5 · value chain and competencies
Synergy lives in value activities
Porter defines the two kinds of synergy through the value chain
from Session 5:
transferring skills between similar value chains, or sharing an activity such as a sales force (p. 19). The
textbook adds that a firm’s boundaries are set by its knowledge and core competencies (p. 301).
Sessions 6 to 8 · advantage in one business
Only business units compete
Session 8
read NVIDIA as a technology and platform story inside one industry. Porter’s first premise keeps that work central:
diversified companies do not compete, their business units do, so a corporate strategy must grow out of and
reinforce each unit’s competitive strategy (p. 6).
The trap this session sets
Bigger is not the same as more valuable. Porter says executives most often skip the test of whether a new unit is
better off inside the company, partly because they confuse company size with shareholder value (p. 11). And a
hot industry is not an attractive one: “Industries are profitable not because they are sexy or high tech; they
are profitable only if their structures are attractive” (p. 9). Hold both thoughts when you reach NVIDIA.
A diversified company has two levels of strategy. Business-unit strategy is about creating competitive advantage in each
business the company is in. Corporate strategy answers two other questions: what businesses the company should be in,
and how the corporate office should manage the set of units (Porter, 1987, p. 1). The textbook frames the same
choice as the firm’s boundaries, set along three dimensions (Rothaermel, 2024, p. 298).
The three dimensions of corporate strategy, redrawn from Rothaermel (2024, Exhibit 8.2, p. 299)
Each axis is a question the CEO answers. The three question cards are the text version.
Vertical integration · p. 298
In what stages of the industry value chain should we take part?
The industry value chain runs from raw materials to finished goods
and services in distinct stages. Owning more of it is backward or forward integration;
Vertical integration extension takes it apart with Amazon.
Horizontal diversification · p. 298
What range of products and services should we offer?
Single business, dominant business, related or unrelated
diversification, and the discount markets apply to conglomerates, are for
Diversification extension.
Geographic scope · p. 298
Where should we compete: regional, national or global markets?
Global strategy and the multinational are
Session 11.
Today the three dimensions are a map for describing NVIDIA, not yet a theory of each.
Four ideas the textbook uses to decide where to compete
Core competencies
Unique strengths embedded deep in the firm that let it
differentiate, or offer comparable value at lower cost. What the firm knows how to do well belongs in-house;
payroll and facility maintenance can go outside (p. 301).
Economies of scale
Average cost per unit falls as output rises. A reason to grow
within a business until minimum efficient scale is reached (pp. 300–301).
Economies of scope
Producing two or more outputs together costs less than producing
each on its own with the same resources and technology. The economic case for related businesses (p. 302).
Transaction costs
All the costs of an economic exchange. They decide whether it
pays to widen the firm’s boundaries or to use the market; the subject of the
make-or-buy section below (p. 302).
The record: Porter’s sober picture
Porter set aside announcement-day stock reactions and
shareholder returns as measures: a profitable base business can hide a poor diversification record, as at CBS and
General Mills. He counted instead how many acquired units each company kept (1987, pp. 2, 6).
33
large, well-regarded US diversifiers studied over 1950–1986. Each entered on average 80 new industries and 27 new fields (pp. 1–2).
70%+
of new entries were acquisitions; 22% were start-ups and 8% joint ventures (pp. 2–3).
> 60%
of acquisitions in entirely new fields were later divested, and more than half of those in new industries. Fourteen companies left more than 70% of their new-field acquisitions (p. 4).
74%
average divestment rate for unrelated acquisitions. Six of the 33 companies were themselves taken over before the study ended (pp. 4, 6).
“Only the lawyers, investment
bankers, and original sellers have prospered in most of these acquisitions, not the shareholders.”
Porter (1987, p. 6)
Three premises Porter treats as facts of life
01
Competition happens at the business-unit level
Diversified companies do not compete; their units do. A
corporate strategy that does not put each unit’s success first will fail, however elegant it looks (pp. 6–7).
02
Diversification adds costs and constraints
Beyond allocated overhead, units explain decisions upward, comply
with planning systems and corporate policies, and lose the chance to motivate staff with direct equity. These costs
shrink but never vanish (p. 7).
03
Shareholders can diversify themselves
They can buy shares at market prices and avoid acquisition
premiums, so they often diversify more cheaply than a corporation can (p. 7).
The consequence. Corporate strategy succeeds
only if it truly adds value: to units, through tangible benefits that outweigh their lost independence, and to
shareholders, through diversification they could not copy on their own (p. 7). The three tests turn that into a checklist.
Passing the essential tests · Porter (1987), pp. 7–11
Three tests every diversification move must pass
Porter turns “add value” into three conditions under which diversification creates shareholder value. Most
companies make sure their plans pass some of them; his study found that ignoring one or two produced disastrous results
(p. 8). Notice that the first test is about the industry, the second about the price of getting in, and only the
third about the company.
Test 1 · pp. 8–9
Attractiveness
The industry must be structurally attractive, or capable of being made so.
Its structure must support returns above the cost of capital;
otherwise the entrant must restructure the industry or win an advantage well above the average. The industry need not
be attractive already, since early entry can reshape it. Companies drop this test for three reasons: a comforting sense
of “fit” (Shell and other oil majors in chemicals), a low price, or fast growth mistaken for profit
potential, which burned many entrants to personal computers, video games and robotics.
Test 2 · pp. 9–10
Cost of entry
The cost of entry must not capitalise all the future profits.
An acquirer faces an efficient market for companies, where
premiums price in the target’s prospects. Philip Morris paid more than four times book value for Seven-Up, so profits
had to more than quadruple to hold its return; they did not, and it sold. A start-up instead pays the entry barriers in
full, and attractive industries are attractive because those barriers are high. “The more attractive a new
industry, the more expensive it is to get into” (p. 10).
Test 3 · pp. 10–11
Better-off
The new unit must gain advantage from its link with the corporation, or the corporation from it.
Some benefits arrive once, as when a parent overhauls strategy
or installs a strong team; then the parent should sell the unit when the results show. Others last, as when the new
unit sells through sister units’ distribution, a logic behind the Baxter Travenol and American Hospital Supply
merger. Spreading corporate risk does not count: shareholders can do it themselves.
Activity · score the moves
Six NVIDIA scope moves through Porter’s three tests
Done
Five moves NVIDIA made or attempted, and one option the
case leaves open: making its own chips. Rate each move on each test using only what the case reports. When you have rated all three
tests for a move, this page’s reading appears beside yours. “Doubtful” is a legitimate answer when the
case does not give the evidence the test needs; say what evidence you would want.
0 of 6 moves scored.
Concepts of corporate strategy · Porter (1987), pp. 11–27
Four ways a corporate centre can add value
From the companies he studied, Porter identifies four concepts of corporate strategy in use. Each rests on a different
mechanism for creating value and asks the company to organise itself differently. The first two need no connections
between business units; the last two depend on them (p. 12). His conclusion is directional: a company creates more
value as its strategy moves from portfolio management towards sharing activities (p. 26).
Porter's four concepts of corporate strategy, summarised in this page's words
Concept
What the centre does
Where the value comes from
Where it breaks
Porter’s examples
Portfolio managementpp. 13–15 · no links needed
Buys sound, attractive companies whose managers stay; keeps them autonomous; supplies capital and professional management; reviews results; moves cash from cash generators to high-potential units.
Efficient capital markets price good companies fully; capital is easy to raise; managing many unrelated units overwhelms the centre. Markets apply a “conglomerate discount”.
Gulf & Western, Consolidated Foods (Sara Lee) and ITT ended in wholesale divestment. Still workable in developing countries.
Restructuringpp. 16–18 · no links needed
Seeks undeveloped, sick or threatened companies, or industries on the edge of change; replaces management, shifts strategy, adds technology; sells the unit once results are clear.
Buying problems cheaply keeps the premium low; active intervention makes the unit better off. Passes all three tests when done well.
Needs rare insight in unfamiliar businesses; the pool of targets gets bid up; the hardest part is selling a unit once it performs, so the restructurer drifts into portfolio management.
Hanson Trust, Loew’s, BTR, General Cinema.
Transferring skillspp. 18–22 · links needed
Moves know-how between units whose value chains are separate but similar, reassigning key people and backing the transfer from the top.
Proprietary expertise lowers the cost of entry and changes how the receiving unit competes.
Mere similarity; vague hopes of transfer; no incentives for units to take part. Once nothing is left to transfer, the unit should be sold.
3M, PepsiCo; a toiletries unit teaching a cough-syrup unit positioning, promotion and packaging.
Sharing activitiespp. 22–25 · links needed
Runs one activity for several units: a sales force, a distribution system, a service network. Builds horizontal mechanisms: corporate identity, cross-unit task forces, incentives beyond unit results.
Lower cost through scale, utilisation and learning; more or cheaper differentiation. Meets the better-off and cost-of-entry tests.
Coordination costs; compromising an activity to suit two units; sharing only overhead. Close fit never excuses an unattractive industry.
P&G in paper towels and diapers; McKesson; GE in major appliances; Du Pont, IBM.
Summary in this page’s words of Porter (1987), pp. 12–25. It is not a copy of his Exhibit 4.
The outline’s three approaches, mapped onto the readings
The course outline names a portfolio approach, a synergy
approach and core competencies. Porter’s four concepts and the textbook cover all three, from different angles.
Portfolio approach
The centre as banker and reviewer
This is Porter’s portfolio management, with restructuring as
its active cousin; in both, the centre is “a selector, a banker, and an intervenor” (p. 18). The
textbook’s tool is the BCG growth-share matrix, which sorts units by relative market share and market growth into
dogs, cash cows, stars and question marks, each with an investment rule (Rothaermel, pp. 331–332). Internal
capital markets create value only if headquarters allocates capital better than outside markets would (p. 332).
Synergy approach
Links between units, defined precisely
Porter’s transferring skills and sharing activities. He
warns that annual reports make almost anything look related, and that “imagined synergy is much more common than
real synergy” (p. 18). GM bought Hughes Aircraft because cars were going electronic; he calls that paper
synergy. Real synergy is defined at the level of value activities, with the conditions below.
Core competencies
What the firm knows how to do well
The textbook’s idea, not Porter’s word. Core competencies
are unique strengths deep in the firm (p. 301). They set the firm’s boundaries, and the textbook’s
diversification types turn on whether units share them (p. 321). Its core competence–market matrix, adapted
from Hamel and Prahalad (1994), crosses existing or new competencies with existing or new markets (p. 324).
Synergy, on Porter’s conditions
Transferring skills · p. 20
Similarity is not enough. Three conditions must hold.
1. Similar enough to matter. The activities are close enough for shared expertise to mean something. Broad likeness, such as being marketing-intensive or bending metal, is not a basis for diversifying.
2. Important to advantage. The skills concern activities that drive competitive advantage, not peripheral ones such as government relations or real estate.
3. Advanced and proprietary. For the receiving unit, the skills are a significant source of advantage, beyond what competitors can do.
The transfer must be specific, staffed and rewarded; it does not
happen “by osmosis” (pp. 20–21).
Sharing activities · pp. 22–24
A cost–benefit test, activity by activity
Benefits
Lower cost from scale, fuller use of capacity, faster learning; more differentiation,
or differentiation at lower cost. Sometimes the shared activity can be redesigned entirely.
Costs
Coordination, and above all compromise: a salesperson selling for two units works in a
way neither would choose alone. If the activity is not scale-sensitive, coordination costs win.
The activity shared must matter to advantage, as P&G’s
distribution does for bulky, costly-to-ship diapers and paper towels. Sharing only corporate overhead is rarely, if
ever, a reason to diversify (p. 23).
Parenting advantage · the outline’s second reading
Is this parent the best owner of this business?
The parenting lens adds a comparison with alternative owners. Use the reading lenses and parenting studio to test both useful intervention and harmful misfit. Begin with three questions:
1. Could this business raise capital, talent and legitimacy just as easily on its own?
2. Does the parent actively intervene, with capital, talent or coordination, rather than just collect dividends?
3. Would the unit be worth more inside the group than its standalone market value?
An easy standalone option raises the burden of proof for ownership. Set the
three questions beside Porter’s better-off test and his second premise, that belonging to a group always costs a
unit something (pp. 7, 10). The deck is linked under Go deeper.
Analysis
Where core competence and transferring skills meet
Porter’s third condition for transferring skills, that the
skills be advanced and proprietary enough to be beyond competitors (p. 20), describes something close to what the
textbook calls a core competence (p. 301). Read together, they give one hard question for any scope move: which
specific competence does the new business borrow, and would a rival entering the same business lack it?
The answer for NVIDIA is not obvious. Its GPU and software
competence is shared across gaming, data centres, workstations and cars, yet the case reports that its revenue outside
gaming stayed low from 2012 to 2020 (p. 5).
Activity · classify the centre
Ten things a corporate centre does. Which concept is each?
Done
Eight descriptions paraphrase Porter’s own examples;
two come from the NVIDIA case. Pick the concept each one illustrates. Feedback cites the page, and one item has two
defensible answers.
Transaction costs and the scope of the firm · Rothaermel (2024), pp. 302–318
Make, buy or ally: where the firm’s boundary should sit
Porter asks whether a business should be inside the company. Transaction cost economics asks the same of an activity.
Ronald Coase started the line of work with a blunt question: if free markets are so efficient, why do firms exist at
all? The answer is that markets and firms are two ways of organising an exchange, and each has its own costs
(Rothaermel, 2024, p. 302).
Using the market · p. 303
External transaction costs
Searching for someone to contract with, then negotiating,
monitoring and enforcing the contract.
Using the hierarchy · p. 303
Internal transaction costs
Recruiting and paying people, space and equipment, supervising
work, and the administrative cost of coordinating units: transfer prices, capital budgeting, resource allocation.
They rise with size and complexity.
The rule · p. 303
Cin-house < Cmarket → make
When an activity costs less to organise inside the firm than to
buy through the market, the firm should integrate it. When the market is cheaper, buy.
Firms and markets each fail in their own way
Advantages and disadvantages of organising activity in firms and in markets
Advantages
Disadvantages
Firm (make)
Decisions by fiat along clear lines of authority; coordination of complex, specialised work; investments specific
to the transaction that are worth little outside it; a community of knowledge that cannot be bought (p. 304).
Administrative costs of bureaucracy; low-powered incentives such as salaries; the principal–agent problem,
where managers pursue their own interests (pp. 304–305).
Market (buy)
High-powered incentives for independent suppliers; flexibility to compare prices and switch providers (p. 305).
Search costs; opportunism, including hold-up; incomplete contracts that cannot foresee every contingency or measure
performance; costly enforcement; information asymmetry, which can leave only “lemons” on offer
(pp. 305–306).
Paraphrased from Rothaermel (2024), Exhibit 8.4 and pp. 304–306.
Not a switch but a continuum
Between buying on the open market and doing the work
in-house sit hybrid arrangements. Alliances let partners make investments specific to the relationship without the internal
costs of owning each other (pp. 306–307). On this page, “ally” means the middle of the line.
Buy
Arm’s-length market
Buy
Short-term contracts
Under a year, competitive bids; no incentive for suppliers to invest in specific assets
Ally
Long-term contracts
Licensing, franchising
Ally
Equity alliances
A stake signals commitment
Ally
Joint ventures
A new firm, jointly owned
Make
Parent–subsidiary
Owned; still prone to turf battles
Make
In-house
Order and labels follow Rothaermel (2024), Exhibit 8.5, p. 306; notes from pp. 306–310.
Specialised assets · pp. 305, 315–316
Why specific investments invite hold-up
An asset is specialised when it is worth far more in its intended
use than in the next best one. It can be specific to a site, to one customer’s engineering needs, or to the
know-how of one organisation. Once such an investment is made, a partner can withhold cooperation to gain bargaining
power. The textbook’s remedy for hold-up in supply is a second source, “of, say, semiconductors”
(p. 305); backward integration is the other.
Competence sets the limit · pp. 301, 316–317
Integrating what you cannot do well
Integration can raise costs, lower quality, reduce flexibility and
invite legal trouble. The main reason to integrate is vertical market failure: the market is too risky and the
alternatives too costly. The textbook reports that integrating two or more steps away from a firm’s core
competency fails two-thirds of the time (p. 317).
Middle routes · p. 318
Taper integration and strategic outsourcing
A tapered firm makes some of an input and buys the rest, or sells
through its own outlets and others’, which keeps in-house units exposed to market comparison. Strategic outsourcing
moves an activity outside the boundary. The full benefits and risks of vertical integration belong to
Vertical integration extension.
The case · Yoffie & von Bargen (2024)
NVIDIA in 2024: chips, systems, software, and a question of scope
Once a niche supplier of graphics chips for video games, NVIDIA had become a diversified semiconductor, systems and
software company by 2024 (case, p. 6). Huang knew semiconductors were cyclical and wanted to widen the footprint
from GPUs and networking to complete systems and cloud services (p. 6). The case ends on the scope question: how far
up or down the value chain should NVIDIA play, with a single supplier on one side and concentrated, increasingly
hostile customers on the other (p. 12)?
Case · p. 6
Horizontal: four businesses, one dominant
Revenue by segment, second quarter of 2024
Data centre$30 bn
Gaming$2.9 bn
Visualisation (workstations)$454 m
Automotive$346 m
Data centres were almost 90% of revenue in the quarter. For the
fiscal year to January 2024 they were $47.5 bn of $60.9 bn, or 78% (Exhibit 1a). Nearly half of revenue came
from four large data-centre customers (p. 6).
Case · pp. 3, 7
Vertical: a designer that sells, not a maker
Intel and Samsung designed and manufactured their own chips.
NVIDIA built its business on outsourcing manufacturing, mostly to TSMC, and kept design and sales (p. 3). A
state-of-the-art fab could cost as much as $25 billion in 2024; outsourcing avoided that capital but gave NVIDIA
less control of its supply chain (pp. 3–4). It has since pushed forward into complete systems and, from
2023, cloud rental (p. 7). The chain is mapped below.
Case · pp. 5, 7–8
Geographic: exposed at both ends
China was 14% of data-centre sales in fiscal 2024, down from 19%,
and US export controls meant some GPUs needed a licence to ship there (p. 8). Most of TSMC’s leading-edge
capacity stayed in Taiwan, although it was building fabs in Arizona and Japan (p. 7). Mellanox, bought in
2019, was Israeli (p. 5). Session 11 picks up global strategy.
Bars show the four segments the case names for the quarter, scaled to data centres. The
case also says data centres were “closer to 80%” of revenue in 2024 (p. 7); the two figures cover different periods.
Who does what in NVIDIA’s value chain, 2024
NVIDIA does it (make)Done with partners (ally)Bought from others (buy)
1 · p. 3
GPU design
NVIDIA designs its GPUs. Data-centre GPUs had over 16,000 cores and about 80 billion transistors (p. 4).
2 · pp. 3, 7
Wafer fabrication
TSMC, on which virtually all NVIDIA hardware depended. TSMC charged roughly $900 per chip for H100s that NVIDIA sold for $25,000–40,000 (p. 4).
3 · p. 3
Assembly, packaging, testing
A variety of third parties.
4 · p. 3
Graphics cards
Contract manufacturers put GPUs on cards for consoles, PCs and servers.
5 · p. 5
High-bandwidth networking
In-house since the 2019 Mellanox acquisition, for almost $7 billion.
6 · pp. 7, 13
Servers and systems
OEMs such as Dell, HP and Lenovo, and now NVIDIA’s own DGX systems, in competition with them.
7 · p. 7
Cloud rental
DGX Cloud, since 2023: NVIDIA rents servers in Google, Microsoft and Amazon data centres and re-rents them to customers such as ServiceNow and Amgen.
8 · p. 3
Sales
A direct sales force to OEMs and data-centre customers, plus retailers and direct-to-consumer.
Scope moves in the case: NVIDIA’s and its rivals’
Rivals were buying scope too. The right-hand list is the
context for every NVIDIA move: each one changes what the others can copy.
NVIDIA
2006CUDA introduced; later a platform with over 150 SDKs serving industries from gaming and design to life sciences, AI and robotics (p. 4).
2012–20Attempts to diversify into chips for cars and into cloud gaming; revenue outside gaming stays low (p. 5).
2017A GPU line designed for crypto mining, to protect the gaming market from shortages (p. 5).
2019Buys Mellanox for almost $7 billion (p. 5).
2020–22Agrees to buy ARM for $40 billion; abandons the deal in February 2022 after UK and FTC objections (pp. 6–7).
2021Launches Grace, an ARM-based CPU for high-performance computing and AI (p. 7).
2023Starts DGX Cloud (p. 7).
2024Sells DGX H100 systems, hoping for half a million by year-end; many customers find them premium-priced and costly to maintain (p. 7).
Rivals and customers
2006AMD buys ATI for $5.4 billion to enter GPUs (p. 9).
2015Amazon buys Annapurna Labs, the start of its Graviton, Trainium and Inferentia chips (p. 10).
2019Intel buys Habana Labs for $2 billion; its Gaudi2 targets the H100 at about half the price (p. 9).
2023Google announces TPU v5p; it reportedly paid about $15,000 for an NVIDIA chip against $2,000–3,000 for each in-house TPU (p. 10).
2024AMD buys ZT Systems for $4.9 billion to compete with NVIDIA’s DGX with more customisable systems (p. 9).
EarlierWhen Intel tried selling systems against its OEM customers, its largest OEMs defected to AMD and Intel backtracked (p. 13).
Are NVIDIA’s competencies shared across its businesses?
The professor’s teaching plan names three pillars:
GPU architecture and parallel processing, the CUDA ecosystem, and strategic vision. Porter would ask of each whether it is
an activity the businesses share, a skill they transfer, or neither. How CUDA became a platform is Session 8’s
story; here the question is only what it does for scope.
Analysis
GPU design: a shared activity
One design capability supplies chips for gaming consoles, PCs and
servers (case, p. 3), and parallel processing found uses in science, crypto and AI (pp. 3–5).
Technology development is one of Porter’s value activities (p. 19), so this is sharing, and it passes his
condition that the shared activity matter to advantage.
Analysis
CUDA: shared, and proprietary
CUDA ran only on NVIDIA GPUs and its SDKs spanned industries
(p. 4). That meets Porter’s strictest test, expertise beyond competitors’ reach. The threat is that it
stops being proprietary: OpenAI’s Triton let developers run code on other chips (p. 11).
Analysis
Networking: bought, not grown
Mellanox brought chips for fast links between GPUs and a place deep
inside data centres, a new route to customers (p. 5). That is the better-off test passed through a shared channel.
It serves one business, data centres, so it deepens scope there more than it widens it.
NVIDIA’s moves on the core competence–market matrix (Rothaermel, 2024, p. 324). Placements are this page’s analysis.
New
Build to protect
Networking (Mellanox) and the Grace CPU, for its existing data-centre customers
Mega-opportunity
DGX Cloud: renting compute is a service business NVIDIA had not run
Data-centre AI GPUs; crypto-mining GPUs; automotive chips
Existing market
New market
Rows: core competence, existing or new. Columns: market. The
textbook calls the upper-right quadrant the most challenging, because it needs new competencies for new markets
(p. 325).
Analysis
What kind of diversified firm is NVIDIA?
By the textbook’s revenue rule, a firm earning 70–95%
of revenue from one business is a dominant-business firm (p. 321). With data centres at 78% of fiscal-2024 revenue
and almost 90% in the second quarter of 2024, NVIDIA fits that band, even though the case calls it diversified.
That reframes Huang’s worry about cycles. Porter’s
premise says spreading risk is something shareholders can do for themselves (p. 7). So the argument for adding
systems, cloud or cars has to be the better-off test, not a smoother revenue line.
Applying transaction costs · case pp. 3–13; Rothaermel (2024), pp. 301–317
Make, buy or ally at NVIDIA
The case puts the boundary question twice: should NVIDIA hedge its dependence on TSMC, and should it keep moving forward
into its customers’ businesses (p. 1)? Before taking either, work through the activities one at a time.
Activity · set the boundary
Nine activities, three factors, one recommendation
Done
Pick an activity, then set the three factors as you read
the case. The recommendation follows a simple rule built from the textbook: what draws on a core competence is made
in-house (p. 301); what is not core is bought when specific assets and contracting hazards are low, and handled
through an alliance or a second source when they are high (pp. 305–307, 317). It is a teaching heuristic,
not a formula from the book. Then compare with what NVIDIA actually did.
0 of 9 activities set.
The two boundary decisions the case leaves open
Upstream · the TSMC question
One supplier, no in-house competence
Case TSMC made
over 90% of the world’s leading-edge logic chips and could demand that big customers prepay billions to reserve
capacity (p. 4). It had raised prices in past booms, its production sat in Taiwan, and the one alternative,
Intel, was behind in process technology and also a competitor (p. 12). TSMC took 18–24 months to build a fab;
Huang wanted a new product every year (p. 7).
“In the event that we have to
shift from one fab to another, we have the ability to do it. We won’t be able to get the same level of performance
or cost, but we will be able to provide supply.”
Jensen Huang, quoted in the case (p. 12)
The options the
case leaves open (pp. 1, 12): stay with TSMC alone; move part of the volume to Intel, the most likely
alternative, to build a second source; or integrate backward into making chips.
Analysis This is
the textbook’s hold-up problem in its own example industry, and its remedy is a second source (p. 305). Making
chips would mean integrating into an activity NVIDIA has always left to others: it built its business on design, with
TSMC fabricating (case, p. 3). The three tests on building
a fab show why supply security alone is not a reason.
Downstream · partner or competitor?
Selling systems and cloud to the customers who buy the chips
Case DGX systems put
NVIDIA against Dell, HP and Lenovo, whose margins on NVIDIA-based servers were already thin because the chips cost so
much (p. 13). DGX Cloud put it against the hyperscalers that rented out its GPUs profitably, and that were
designing their own chips to replace it (pp. 1, 10). Amazon’s goal was to reduce the “Nvidia
tax” (p. 10).
The choice, as
the case puts it (p. 1): keep integrating forward and go head-to-head with its biggest customers, or stay a
component supplier. Between the two lie stopping points along the chain mapped above: chips only; chips and systems;
or chips, systems and cloud.
Analysis On
Porter’s terms, the systems business may pass the better-off test and still fail his first premise, if it weakens
the chip business that has to compete for those same customers. Transaction costs add a second warning: DGX Cloud runs on capacity
inside the data centres of the firms it competes with, an alliance whose partners have every reason to limit it. How far forward to integrate is the
Vertical integration extension debate.
Go deeper
Three earlier decks on corporate strategy
None of these was built for this session. The first two use the
Amazon case and go further into vertical integration and diversification than today does, which makes them useful extensions of this session. The third is the parenting-advantage deck. Their company figures were not all checked against this
session’s readings; use this page for the frameworks and the NVIDIA facts.
Fictional board exercise, not a reported NVIDIA proposal: a chip company considers buying a small AI application business. The application team could use the parent’s engineering support, but it must sell through rival hardware ecosystems to grow. Which recommendation can you defend?
Consider the consequences
Engineering fit may create benefits, but mandatory hardware exclusivity could damage the application’s market access. Quantify both. Ownership should not be justified by a benefit while ignoring the parenting behavior that could destroy it.
Consider the consequences
A contract can test whether engineering support adds value while preserving distribution options. It may also leave coordination or intellectual-property problems unresolved. Set a measurable pilot and specify the evidence that would justify ownership later.
Consider the consequences
A categorical ban ignores potentially valuable skill transfers or shared activities. Separate the industry, entry cost and better-off tests, then ask whether this parent is better than a specialist owner. Relatedness is neither proof nor disproof.
Parenting studio · 12 minutes
Would the unit be better off in your hands?
Test a hypothetical ownership proposal. Move the fit judgments and examine net annual contribution. All figures and cut-offs are illustrative teaching assumptions. Matrix placement is a prompt for evidence, not an acquisition recommendation.
Currency units per year, relative to standalone operation. Net contribution = gross benefit − coordination and harm cost. Parent advantage = your net contribution − alternative owner’s net contribution. Acquisition price, financing and discounting are excluded.
Low misfit → high misfit / opportunities increase upwards
Heartland
High opportunity fit · low misfit
Value trap
High opportunity fit · high misfit
Ballast
Low opportunity fit · low misfit
Alien territory
Low opportunity fit · high misfit
Scores near 50 are a boundary requiring investigation; a promising partial fit can be “edge of heartland.” This simplified grid is a teaching adaptation.
Your one-minute recommendation
Write a claim, cite one case exhibit or reading page, and name evidence that would change your mind. This note stays on the page until you leave; download it to keep a copy.
For class
Six questions to argue
Come with a position on each and the page of the case or reading
that supports it.
01
Porter says diversified companies do not compete; only their business units do. Name two decisions in the NVIDIA case that only the corporate level can take, and say what each does to the data-centre unit’s competitive position.
02
Mellanox cost almost $7 billion. Which of Porter’s tests does the case let you settle, and what evidence would you need to show it passed the better-off test rather than simply bought revenue?
03
NVIDIA did not get ARM but did get an ARM-based CPU. What does that say about the cost-of-entry test, and about where on the make-or-buy continuum a capability is best held?
04
NVIDIA can now afford a $25 billion fab. Using transaction costs and core competence, argue for and against backward integration, then choose: stay with TSMC alone, add a second source, or build.
05
Partner or competitor? Should NVIDIA keep selling DGX systems and DGX Cloud when OEMs and hyperscalers are its largest customers? Use Intel’s earlier experience with its OEMs.
06
By the textbook’s revenue rule NVIDIA is a dominant-business firm in a cyclical industry. Is diversifying to smooth the cycle a valid corporate strategy on Porter’s premises? Which of his four concepts should guide its next move?
Knowledge check
Ten questions
Answer before you look. Each answer explains itself, and the explanation matters more than the score. The full practice bank for this session is at /teaching/quiz/1yr-08.
0 / 0 correct
Retrieval practice · 3 minutes
Can you defend the mechanism?
Connect → explain → challenge → revise
Session 9 practice lab
An original diagram and three contexts help you test the mechanism before committing to a recommendation.
A promising business is not automatically a good scope move
Porter’s three tests are separate gates. Parenting advantage adds a comparison with other owners. Passing a gate supports further investigation; it does not settle price, execution or the ownership counterfactual.
The diagram in words
1 · Attractiveness: Industry economics.
2 · Cost of entry: Room for a return.
3 · Better off: A specific activity link.
Then · Parenting advantage: Compared with other owners.
First test industry attractiveness, then entry cost, then a concrete better-off mechanism. Separately compare this parent’s net contribution with alternative owners and with a contract.
Identify the weakest link, the evidence needed and a condition that would change your recommendation.
Hypothetical · original scenario
Nila Foods: buy the packaging factory?
Nila Foods relies on a specialist package that cannot easily be reused by another buyer. Owning the factory could reduce repeated bargaining over changes, but Nila has never operated that production technology. A longer contract, joint investment or second source might protect supply with less operating burden. The board should compare those routes before treating supply anxiety as sufficient evidence for ownership.
Discuss: Which investment is specific to Nila, what hazard follows, and what capability would ownership require that a contract does not?
State a claim, explain its causal mechanism and point to evidence. Then identify a credible alternative explanation and one observation that would distinguish it. These prompts invite defensible interpretations; they have no automated score.
Global · NVIDIA / September 2024 case
A designer’s boundary is a choice
The assigned NVIDIA case describes a chip designer dependent on manufacturing partners while expanding into networking, systems and cloud services. These moves change different parts of the value chain. Teaching inference: owning fabrication might address a supply concern but introduce a manufacturing capability gap; moving downstream might add system value but create customer conflict. Neither benefit proves the net value of ownership. Read the existing case analysis and its page references; keep the case’s 2024 horizon.
Discuss: Choose one activity. Compare in-house production with a protected contract and identify the cost that could overturn the recommendation.
State a claim, explain its causal mechanism and point to evidence. Then identify a credible alternative explanation and one observation that would distinguish it. These prompts invite defensible interpretations; they have no automated score.
Indian · ITC / 21 September 2001
Paperboards and packaging: turn “synergy” into a test
ITC’s board announcement proposed amalgamating ITC Bhadrachalam Paperboards, already a subsidiary, with ITC. It expected capacity investment and coordination with ITC’s specialty-paper and packaging divisions. Source: ITC announcement, 21 September 2001.Teaching inference: a specific paperboard-to-packaging activity link is more testable than “a bigger group.” The announcement is a proposal and management expectation, not proof that the synergies materialized or that no alternative parent could do better.
Discuss: What operational measure would establish a benefit beyond a financial portfolio, and what cost of coordination would you deduct?
State a claim, explain its causal mechanism and point to evidence. Then identify a credible alternative explanation and one observation that would distinguish it. These prompts invite defensible interpretations; they have no automated score.
Four diagnostic questions
Choose an answer, explain it, then use the feedback to revise your reasoning. These are original teaching scenarios and simplified calculations, separate from the existing practice bank.
Practice score: 0 / 4
Make the alternative owner real
For the NVIDIA systems or cloud activity, name the capabilities an alternative owner would need. Compare net operating contribution and harmful behavior before discussing an acquisition price.
Write the better-off sentence
For Nila or ITC, complete: “Activity A improves activity B through mechanism C, measured by D, after cost E.” Then challenge whether the same benefit could be obtained contractually.
Use the supplied resources by topic
The folder names reflect an earlier sequence. This map follows the current session topic. Locate these files in your instructor-provided resources or obtain the reading through the licensed course pack, library or publisher; the originals are not rehosted here.
From Competitive Advantage to Corporate Strategy.pdf — Porter (1987). Separate the three tests and compare the value mechanisms of alternative corporate strategies. Publisher or source record.
Corporate Strategy_The Quest for Parenting Advantage.pdf — Campbell, Goold & Alexander (1995). Match improvement opportunities to this parent’s capabilities; challenge harmful misfit and compare another owner. Publisher or source record.
Competitive Advantage Creating & Sustaining.pdf — Porter, Competitive Advantage (1985) · supporting text. Supporting file as named in the resources; consult the assigned chapter in your course pack. The publisher’s book record explains the activity/value-chain lens. The local scan is not a replacement for the corporate-strategy articles. Publisher or source record.
Nvidia, Inc..PDF — NVIDIA case reused from the platform resources. Revisit the same September 2024 case for scope and boundaries, rather than assuming a newer company position. Publisher or source record.
Preparation
Readings for Session 9
The case and the two articles are
copyrighted Harvard Business School Publishing and Harvard Business Review materials, and the textbook is in copyright.
Get them through the course pack or the library; use the publisher links below to locate the readings.
Required · the case
Yoffie, D. B., & von Bargen, S. (2024). Nvidia, Inc. in 2024 and the future of AI (HBS No. 9-725-360). Harvard Business School Publishing.
Pages 1–7 and 12–13 matter most today, with Exhibit 1a. The same case serves Session 8.
Required · the outline’s reading
Porter, M. E. (1987). From competitive advantage to corporate strategy. Harvard Business Review, 65(3).
The three tests, the four concepts and the action programme. Page numbers on this page follow the 32-page HBR.org print.
Required · the outline’s reading
Campbell, A., Goold, M., & Alexander, M. (1995). Corporate strategy: The quest for parenting advantage. Harvard Business Review.
Use the reading lenses and the parenting studio to compare useful intervention, harmful misfit and alternative owners.
Further · course textbook
Rothaermel, F. T. (2024). Strategic management (6th ed.). McGraw Hill.
Chapter 8, “Corporate strategy: Vertical integration and diversification”. Today: sections 8.1 and 8.2 (pp. 298–310) and the core competence–market matrix (pp. 323–326). Use the chapter’s remaining sections with the vertical-integration and diversification extensions linked above.
A note on the figures on this page
NVIDIA figures are the case’s, as of late 2024, and some are
analysts’ estimates it reports. The 78% data-centre share and the price-to-revenue comparisons in the scoring
activity are this page’s arithmetic on Exhibit 1a. Porter’s figures describe 1950–1986 and his
examples are his, not current descriptions of those firms.
Required · Porter (1987)
From Competitive Advantage to Corporate Strategy. Harvard Business Review. Prepare separate judgments for attractiveness, cost of entry and better-off; do not average away a failed test.
Reading lens: Publisher record. Applications and model assumptions below are learning exercises.
Required · Campbell, Goold & Alexander (1995)
Corporate Strategy: The Quest for Parenting Advantage. Harvard Business Review. Prepare an opportunity, a harmful misfit and a credible alternative owner for one scope proposal.
Reading lens: Publisher record. Applications and model assumptions below are learning exercises.