How many stages of the chain to own, and what owning them buys
Session 10 asked where a firm’s boundaries should sit. This session takes one of the three dimensions of
corporate strategy, the industry value chain, and asks how many of its stages a firm should own. Vertical
integration is ownership of the inputs a firm needs or of the channels that carry its output to customers
(Rothaermel, 2024, p. 311). It can lower costs, raise quality and secure channels. It can also raise costs,
lower quality and lock the firm in (pp. 315–316).
Amazon began in 1995 by selling books online. By 2024 it produced films, sold reading devices and voice
assistants, designed AI chips, promised two-hour delivery in some cities, ran grocery stores and rented computing
to other companies (case, pp. 1–3). Some of those moves are vertical and some are not. Today you sort them,
weigh them, and test Osegowitsch and Madhok’s (2003) claim that firms now integrate forward mainly to learn from
their customers.
01When does owning the next stage of an industry’s value chain beat buying from it or allying with it?
02What does integration buy, what does it cost, and what lies between full ownership and the open market?
03Which of Amazon’s moves are vertical at all, and which of them created value rather than destroyed it?
Facts from the Gupta and Rodriguez case (revised July 2024) are marked
Case; where this page reads them through a framework, the card is
marked Analysis. Where the case is silent, the page says so and
asks the question instead. Textbook pages refer to Rothaermel, 6th edition;
Osegowitsch and Madhok pages to the Business Horizons print (pp. 25–34).
Carried forward · Session 10
From the boundary of the firm to the stages of one chain
Session 10 built the
tools: Porter’s tests for whether a business belongs in the company, and transaction costs for whether an activity
belongs inside the firm. Today narrows the second tool to one question. Along a single industry’s value chain,
which stages should the firm do itself?
Session 10 · transaction costs
Make when the hierarchy is cheaper
If an activity costs less to organise inside the firm than to buy
through the market, the firm should integrate it; between the two ends sit contracts and alliances
(Rothaermel, 2024, pp. 303, 306). Today that rule is applied stage by stage, with the specific benefits and risks
the textbook lists for vertical moves.
Session 10 · the three dimensions
One axis at a time
Corporate strategy sets the firm’s boundaries along vertical
integration, horizontal diversification and geographic scope (p. 298). This session takes the vertical axis.
Session 12 takes
geography and Session 13
takes diversification, both with the same Amazon case.
Sessions 9 and 10 · NVIDIA
A firm that chose not to integrate
Session 10 framed NVIDIA as a designer that does not make its chips:
it keeps design and sales and outsources manufacturing, mostly to TSMC. That avoided the capital of a fab but gave it
less control of its supply. Amazon has the opposite habit. It keeps building stages for itself.
The trap this session sets
Not every move into something new is vertical. A move is vertical only if it takes the firm into a stage that supplies
its existing business or carries that business’s output to customers. Entering a neighbouring industry, such as
advertising or healthcare, is diversification, which is Session 13’s subject. Even the textbook counts AWS as
both a backward vertical move and an example of horizontal diversification (pp. 297, 301). Keep the two questions
apart before you judge any move.
Vertical integration along the industry value chain · Rothaermel (2024), pp. 310–314
Upstream, downstream, and how far a firm reaches
The industry value chain, also called the vertical value chain, is the sequence of stages that turns raw materials
into finished goods and services. Each stage is usually a separate industry with its own competitors, so a firm that
integrates into a new stage takes on a new set of rivals (Rothaermel, 2024, pp. 311–312). The degree of
integration can be measured by value added: the share of a firm’s sales generated inside its own boundaries,
which tends to track the number of stages it works in (p. 311).
A generic industry value chain, redrawn from Rothaermel (2024, Exhibit 8.6, p. 311), with the textbook’s mobile-phone examples (p. 311)
Stage 1 · upstream
Raw materials
Commodities such as chemicals, ceramics, metals and oil for plastic.
Contract manufacturers such as Foxconn assemble phones for brand owners.
Stage 4
Marketing and sales
For phones, wireless carriers such as AT&T, Verizon and Airtel that sell the service.
Stage 5 · downstream
After-sales service and support
Service and support after the sale, the stage closest to the end user.
← Backward vertical integration
Moving ownership upstream, toward the inputs at the start of the chain (p. 313).
Forward vertical integration →
Moving ownership downstream, closer to the end customer (p. 313).
Degrees of integration: from grape to glass, or design only
Few firms are fully integrated. Most concentrate on a few
stages and some on just one (p. 312). The textbook’s three examples span the range.
Fully integrated · p. 312
E&J Gallo Winery
Owns its vineyards, bottling plants and distribution and logistics
network, and sells online where the law allows. Its value added approaches 100%. Gallo cites scale economies,
coordinated scheduling and better control of quality. The price is a different set of competitors at every stage.
Largely disintegrated · p. 312
Apple
Concentrates on design, marketing and retailing, and outsources
everything else. The textbook adds that the stages differ in profit potential: Apple captures about 95% of the value
created with an iPhone and Foxconn, which assembles it, about 5% (p. 314).
Moving in both directions · p. 313
HTC
Began as a no-name contract manufacturer, then integrated backward
into designing phones and forward into branded sales. It learned the adjacent skills from the brand owners it built
for, which had to share their technology to get phones made at scale.
Analysis
Where is Amazon on its own chain?
Backward and forward only mean something once you fix the firm’s
home stage. Amazon started as an online bookseller with a virtual store, unlimited inventory and low fixed costs
(case, p. 1). This page takes retail, the online store, as its home stage. Upstream of the store sit the
things it sells and the inputs it runs on: goods, films and shows, computing. Downstream sit the steps that get a
purchase to the customer and the devices through which the customer buys.
Fix a different home stage and the labels change. That is why two
careful readers can call the same Amazon device backward or forward, as the next activity shows.
Two questions to keep apart
1. Is the move vertical? Does it take the firm into a stage
that feeds its existing business or carries that business’s output to customers?
2. Is it a good move? Do the benefits beat the risks, and beat
what a contract, an alliance or a partial stake would deliver?
The first question is classification. The second is strategy,
and it is the subject of the rest of the session.
Classify before you judge · case pp. 1–3; Rothaermel (2024), pp. 297, 300–301
Eleven Amazon moves: backward, forward, or not vertical?
Jassy’s question in the case is whether Amazon should keep going into new businesses or focus on a few core ones
(case, p. 1). Before anyone can answer it, the moves have to be sorted. Some move Amazon along its own value chain;
others take it into new industries, which is diversification and belongs to
Session 13. Taking
the online store as Amazon’s home stage, classify each move. Feedback gives the page’s reading and its source;
three items have two defensible answers.
Activity · value-chain placer
Place each move on Amazon’s chain
Done
Choose a direction for each card. Your answer is pinned to
the map, marked with a tick if the page accepts it and a cross if not. The map is a text list, so a screen reader
reads it in order.
Amazon’s chain, as you have placed it
← Backward · upstream of the store
Inputs the store runs on: goods, content, computing
Home stage · the online store
Books from 1995, then music, movies, electronics and general merchandise (case, p. 1)
Forward · between the store and the customer →
Fulfilment, delivery, the devices customers buy through
Not vertical · same stage, contracts or new industries
Buying through contracts, or diversifying (Session 13)
0 of 11 placed.
What the sort shows
Most of Amazon’s vertical moves run forward, toward the
customer: delivery, and devices that sit in the customer’s hand or home. That is the pattern Osegowitsch and
Madhok described in 2003, when they observed that forward integration was far more prevalent (p. 31).
What it does not show
Classification says nothing about value. The Fire Phone and
the Echo are both forward moves into devices; one was written down within months and the other sold over 500 million
units (case, p. 2). The rest of the session is about telling such moves apart in advance.
Where the labels blur
AWS, Kindle and Whole Foods each have two readings. In each case
the disagreement is about which stage is Amazon’s home, or about whether the move serves the store or starts a
new business. Say which reading you use before you argue about value.
Benefits and risks of vertical integration · Rothaermel (2024), pp. 315–317
Five things integration can buy, four ways it can cost
The textbook asks leaders to weigh the benefits against countervailing risks at the same time, and warns that the
risks can outweigh the benefits (Rothaermel, 2024, pp. 315–316). Several risks are the mirror image of a
benefit: integration can lower costs or raise them, improve quality or reduce it. Which one you get depends on the
conditions, not on the move.
Benefits · p. 315
1. Lowering costs
Gallo’s full integration gives it scale economies and
efficiency from coordinating its stages (p. 312).
2. Improving quality
HTC bought One & Co., a San Francisco design firm, to secure
scarce design talent and make better phones (p. 315). Gallo says integration lets it control quality (p. 312).
3. Facilitating scheduling and planning
Forward integration into sales lets a firm plan for and respond to
changes in demand; HTC sells to carriers and unlocked phones on its own site (p. 315).
4. Facilitating investments in specialised assets
Assets worth far more in their intended use than in the next best:
site-specific (bauxite mining next to aluminium smelting), physical-asset-specific (moulds for Gallo’s trademarked
bottles) or human-asset-specific (skills in one organisation’s routines). Such investments invite opportunism;
backward integration is often the answer (pp. 315–316).
5. Securing critical supplies and distribution channels
Delta bought an oil refinery from ConocoPhillips in 2012. It
estimates the refinery supplies 80% of its fuel and saves about $300 million a year; fuel is some 40% of its
operating cost (p. 316).
Risks · p. 316
1. Increasing costs
An in-house supplier always has a buyer, so it has less reason to cut
costs. Outside suppliers serve a larger market and reach scale the captive unit cannot. Complexity adds administrative
costs, such as setting transfer prices between in-house supplier and buyer.
2. Reducing quality
The same guaranteed buyer weakens the incentive to improve
quality or innovate. Outside suppliers, with more customers, learn faster.
3. Reducing flexibility
When steelmaking technology changed, integrated mills such as
U.S. Steel and Bethlehem Steel were tied to their models and could not switch; many went bankrupt, while
non-integrated mini-mills such as Nucor and Chaparral took share.
4. Increasing the potential for legal repercussions
US regulators tend to allow vertical integration because it
usually makes firms more efficient, but monopoly concerns mean it has not gone unchallenged.
The common thread
The first two risks come from removing market discipline:
a captive unit no longer has to win its buyer. The third comes from commitment: assets that suit today’s
technology can trap the firm when it changes. The fourth comes from the market power that controlling several
stages can bring.
When does vertical integration make sense?
The textbook reads the question through history, then
gives one main reason to integrate and one warning (Rothaermel, 2024, p. 317).
Ford · p. 317
Owning the whole chain
Short of materials and parts, Ford came to run mines, rubber
plantations, freighters, blast furnaces and glassworks. At River Rouge, raw materials went in at one end and cars came
out the other. Over time the costs undid the savings and cost Ford its flexibility.
Waves · p. 317
In, out, and back in
In the 1970s Texas Instruments moved downstream into watches and
calculators, while Bowmar tried to move upstream into chips and failed. In 2009 GM tried to reacquire Delphi, a parts
maker it had sold in 1997. PepsiCo and Coca-Cola bought bottling plants in the 2010s and later divested them.
The main reason · p. 317
Vertical market failure
McKinsey advised clients that the main reason to integrate is
that transactions along the chain are too risky and the alternatives to integration too costly or hard to
administer. The textbook notes that this is the same answer transaction cost economics gives.
The warning · p. 317
Distance from the core
Research summarised by The Economist suggests that a company
that integrates two or more steps away from its core competency fails two-thirds of the time. Few clothing makers
want to own textile mills, with their pollution risks and slim profits.
“An innovator who can
figure out how to eliminate annoyances and poor interfaces in the chain can build an incredible advantage, based on the
customers’ desire for that unique solution.”
Rita McGrath, quoted in Rothaermel (2024, p. 317), on why
integration tempts firms that want to change the customer’s whole experience
Analysis
Integration tends to create value when…
the market for the stage fails: specialised assets, few alternative partners,
and contracts that cannot specify or enforce what matters (pp. 305, 317);
the stage sits next to a core competence, one step away rather than two
(p. 317);
owning it yields coordination or learning a contract cannot carry, which is
Osegowitsch and Madhok’s argument for the customer interface (2003, p. 32).
Analysis
…and tends to destroy it when
the stage is served by competitive, contractible suppliers who have more scale
and learn faster than a captive unit would (p. 316);
technology or demand is volatile, so a fixed commitment becomes a trap
(p. 316; Osegowitsch & Madhok, 2003, p. 27);
the in-house unit is shielded from comparison with the market, so nobody can
tell whether it is still competitive (pp. 316, 318).
The two lists are this page’s synthesis of the cited
pages, not a checklist from either source. The textbook’s own summary sets the same benefits as sources of value
and the same risks as sources of cost (Exhibit 8.14, p. 330).
Why firms still integrate · Osegowitsch & Madhok (2003), pp. 25–33
“Vertical integration is dead, or is it?”
Osegowitsch and Madhok wrote against the mood of their time. Equity markets pushed firms to focus and outsource; the
academic view was that integration shared the poor record of unrelated diversification; consultants urged firms to
“unbundle” (2003, p. 25). Yet integration had proved resilient in pharmaceuticals, defence, utilities,
computers and cars, and in services such as media and engineering. Their explanation: the old motives had weakened, and
new ones, above all learning from customers, had taken their place (pp. 25–26).
Traditional motive 1 · p. 26
Strategic: power and position
Integration to win the power play against actual and potential
rivals:
foreclose input or output markets to competitors, or raise their costs;
cross-subsidise one stage from another to squeeze focused rivals;
raise barriers to entry;
keep proprietary knowledge from suppliers and customers who might turn competitor.
Partial, “tapered” integration can be enough: it makes the
threat of full integration credible and improves terms (pp. 26–27).
Traditional motive 2 · p. 27
Governance: transaction costs and agency
Integration to cut exposure to opportunism where contracts fail.
It pays under:
uncertain demand or prices;
uncertain quality of inputs or point-of-sale service;
poor coordination between stages;
failure in markets for knowledge and technology;
a partner’s performance that is hard to specify or measure.
Contracting problems become acute under small-numbers bargaining,
which arises when assets are specific to the transaction (p. 27).
The case against · pp. 27–28
Commitment in a turbulent world
Critics point to higher performance risk where technology is
volatile and demand uncertain. Integration can be a premature commitment, costly to undo when exit barriers are high.
It also loses market incentives, raises production costs and breeds bureaucracy and loss of focus. On flexibility
grounds, the critics say, firms should form alliances instead.
Five contemporary motives
From observation across many industries, the authors list
the motives they saw driving integration at the time (pp. 28–31). Most of their examples move downstream.
01 · p. 28
Value migration
With large installed bases, value shifts from making products to
servicing them. GE’s energy division stopped seeing itself in a $12 billion turbine market and started
competing for $40 billion of utilities’ operating and maintenance spending.
02 · pp. 28–29
Differentiation
When the core product is good enough for most buyers, firms add
downstream services and manage the whole brand. Ford and GM’s bids to enter car retailing aimed partly to remove
the unpleasant dealer experience.
03 · p. 29
Integrated solutions
A pull from customers who now outsource and want one supplier for
the whole job. Bechtel and Fluor Daniel widened into finance, design, construction and even plant operation.
04 · pp. 29–30
Synergies
Close links between stages: feedback from marketing to product
development, suppliers inside customers’ inventory planning, design-build firms whose builders advise designers.
05 · pp. 30–31
Emerging industries
A pioneer may lack the credibility to attract suppliers or
customers, as early Ford did, or may need to own complementary stages to make a system work, as Time Warner did by
owning both films and cable.
Why the old motives weakened · pp. 31–32
Most upstream benefits can now be had without ownership
Scheduling: just-in-time supply, vendor-managed inventory and
electronic links to the point of sale optimise production across firms as easily as within one.
Co-location: savings once credited to integration came from
putting stages side by side, which independent suppliers and buyers now do.
Technology markets: brokers, consultants and experience with
alliances have eased the failures of buying technology from outsiders.
The Internet: it is cheaper to search for, compare and switch
suppliers or distributors.
Globalisation: converging industries and abundant capital erode
entry barriers; liberalised trade makes foreclosure futile in many industries.
Benetton and Nike, they note, achieved tight integration with
suppliers without owning them (p. 32).
The thesis · p. 32
Downstream integration is about learning, and learning cannot be outsourced
Data such as the level of demand are easy to collect. Knowledge of
why customers value an offering, how best to provide it and what they will want next comes only from close learning
relationships with them. Much of that knowledge is tacit and contextual, so it cannot be delegated to channel partners
and fed back in a spreadsheet.
“It is this
‘untradeability’ of learning that explains the vitality of downstream VI.”
Osegowitsch & Madhok (2003, p. 32)
So the right questions about scope become: do the current vertical
boundaries let the firm acquire and process knowledge, and capture its value? If not, would extending them help?
The Internet prediction · p. 26
Some predicted that the Net would break firms up by cutting
transaction costs. The authors saw a contingent picture instead: some industries de-integrated, others re-integrated,
and the Net had helped broad-based incumbents, the “clicks-and-bricks” operators, more than the pure plays.
Their caveats · p. 33
Integrating firms must keep strong market incentives and
flexibility. Where they taper, they must keep existing buyers and suppliers on side. A partial move into retailing may
capture the learning without abandoning independent dealers (note 2).
Amazon, in 2003 · p. 33
In a footnote the authors already name Amazon, with Charles Schwab,
as a portal that could find out within hours whether customers accepted new offers, prices or page designs (note 3).
The customer interface was the asset they had in mind.
Analysis
Two lenses that do not always agree
The textbook’s checklist asks whether integration lowers costs,
raises quality, secures channels and protects specific investments, against higher costs, lost flexibility and legal
exposure. It is a governance lens: integrate where the market fails.
Osegowitsch and Madhok add a reason the checklist barely names: owning
the customer interface to learn. On their lens a forward move can be worth making even when a contract would deliver
the goods more cheaply. Hold both when you reach Amazon’s devices: the Kindle, the Fire Phone and the Echo were
all bids for the customer interface.
Alternatives to vertical integration · Rothaermel (2024), pp. 306–310, 318–319
Between owning a stage and buying from it
Firms would like the benefits of integration without its risks. The textbook names two alternatives, taper integration
and strategic outsourcing (Rothaermel, 2024, p. 318). The hybrids on
Session 10’s make-or-buy continuum,
from long-term contracts to joint ventures, are the third family.
Taper integration · p. 318
Make some, buy the rest
The firm is integrated backward but also buys some supplies
outside, or integrated forward but also sells through outside distributors. Apple and Nike own retail outlets and
also sell through other retailers, physical and online.
In-house units face market comparison, and the firm keeps its competencies rather than hollowing them out.
Flexibility: when demand falls, cut deliveries to outside retailers and keep stocking your own stores.
Internal and external knowledge combine, which can lead to innovation.
In a study of 3,500 product introductions in the computer industry,
firms that tapered outperformed both the more integrated and the more outsourced on innovation and financial
results.
Strategic outsourcing · pp. 318–319
Move an activity out
Moving one or more value chain activities outside the firm to
other firms in the industry value chain. It lowers the degree of integration. Firms hand human-resource systems to
specialists that have deep competencies and scale.
Outsourcing to another country is offshoring: Infosys serves many
Fortune 100 firms from Bangalore, and US law firms send routine legal work to India. Where to locate activities is
Session 12’s question.
Contracts and alliances · pp. 306–309
Commit without owning
Short-term contracts give suppliers no reason to invest in
specific assets. Long-term contracts such as licensing, equity alliances and joint ventures support those investments
without the internal costs of ownership (pp. 306–307). An equity stake is a credible commitment: difficult
and costly to reverse (p. 309).
Osegowitsch and Madhok add that partial integration can make the
threat of full integration credible and win better terms (2003, pp. 26–27).
Where Amazon sits between make and buy
Amazon rarely chooses pure ownership or pure purchase. Four
of its stages, read against the alternatives above.
Four Amazon stages: what it owns, what it buys or contracts, and the reading
Stage
What Amazon owns
What it buys or contracts
Reading
Video content
Amazon Studios, from 2011, producing original content; in 2015 it announced that the Studios would make feature films (case, p. 2).
Multi-year licensing agreements with Viacom for hundreds of shows; over $200 million paid to HBO in 2014 for select shows (p. 2).
Taper integration backward into content: some made, the rest licensed.
Selection in the store
Its own retail sales, which began with books and widened to music, movies, electronics and general merchandise (case, p. 1).
Independent sellers on the marketplace since 2000, with more than 60% of the store’s 2023 sales (case, p. 1).
Taper-like (this page’s reading): third-party merchants sell their own products through Amazon’s platform (case, p. 1), so part of the store’s selection comes from firms Amazon does not own.
Access to mobile shoppers
The Fire Phone, launched July 2014, written down by $170 million three months later (case, p. 2).
Reaching mobile shoppers on phones whose platforms other firms control; the case notes that Amazon’s own AppStore rivalled Google’s and Apple’s (p. 2).
Chose to make. Question: what could a phone of its own do that reaching shoppers on other firms’ phones could not? The case reports only disappointing sales (p. 2).
Computing
AWS, offering cloud computing to many companies since 2002 (case, p. 1). The textbook treats creating it as backward integration (p. 301); the decks under Go deeper describe it as having started as Amazon’s internal IT.
The case is silent on what AWS buys from outside suppliers. It reports only that Amazon has built its own AI training and inference chips (p. 3).
Integrated, and selling outside. Outside customers expose the in-house unit to market comparison, the benefit taper offers (p. 318).
Facts from the case as cited; the readings in the last column are this page’s analysis using Rothaermel (2024, p. 318).
The case · Gupta & Rodriguez (2024)
Amazon in 2024: the vertical moves, one at a time
Amazon’s revenue grew 12% in 2023, from $514 billion to $575 billion (case, p. 1). Andy Jassy, CEO
since 2021, has to decide whether to keep entering seemingly unrelated businesses or focus on a few core ones
(p. 1). The case closes on the charge that Amazon was “defying the basic principle of business –
focus” (p. 4). This session answers only part of that charge: the moves along Amazon’s own value chain.
Whether the portfolio as a whole is spread too thin is
Session 13’s
question, and its international reach is
Session 12’s.
ForwardCase · pp. 1–3
Fulfilment and delivery: the promise behind Prime
Prime launched in 2005: free two-day shipping for a $79 annual fee. The fee rose to $99 in 2014, $119 in 2018 and
$139 in 2022, and membership kept growing (pp. 1–2).
By 2023 Amazon had more than 200 million Prime members, who spent more than four times what other customers spent
(p. 2).
Prime Now offered free delivery of tens of thousands of daily essentials within two hours, in several US cities
(p. 2). Amazon Business added 24-hour delivery to business buyers (p. 3).
Textbook
To deliver millions of products in two days or less in the United States, Amazon built a large network of fulfilment
centres: a large-scale commitment that lets it exploit economies of scope (Rothaermel, 2024, p. 302).
Analysis
This is forward integration one step from the store, into the stage that carries a purchase to the door. The
benefits that apply are quality the customer sees (speed), scheduling stock against delivery, and a secured channel
(p. 315). The risk is fixed facilities that cannot shrink if demand falls (p. 316).
Where the case is
silent. It does not say who carries parcels the last mile or what delivery costs Amazon. Question: if Amazon
owned that last stage too, which benefits would grow, and which risks?
Forward, contestedCase · p. 2
Devices: Kindle, Fire Phone, Echo
Kindle launched in 2007 at $399 and sold out in under six hours; after the iPad arrived in 2010, its price fell to
about $100. It gave customers a convenient way to shop for and download books and movies from Amazon’s store.
The Fire Phone, July 2014, had a “Firefly” button that identified objects to buy on Amazon. Three
months later Amazon announced a $170 million write-down.
Echo’s assistant, Alexa, answered questions and placed orders. Over 500 million Echo devices were sold by May
2023, yet the Wall Street Journal reported in 2022 that the devices unit had an operating loss of more than $5 billion a year.
The Dash Button, for reordering household items, was discontinued in 2019 for weak demand.
Case · p. 2
Amazon also built an AppStore that rivalled Google’s and Apple’s. While the devices unit was reportedly under
review in 2022, a spokesperson said Amazon was “as optimistic about Alexa’s future today as we’ve ever
been”, and Amazon planned to bring generative AI to Alexa.
Analysis
Every device is a bid for the customer interface, which is Osegowitsch and Madhok’s strongest case for
integrating forward (2003, p. 32), and for a sales channel Amazon controls (Rothaermel, 2024, p. 315). The
Fire Phone shows the limit: owning the interface did not make customers buy the phone. The case reports
disappointing sales and is silent on why.
Question.
Kindle’s price fell from $399 to about $100, and Alexa places orders on Amazon. Should a device be judged on its
own loss, or on what customers buy through it? What evidence would settle it?
Backward, then outwardCase · pp. 1, 3–4
Computing: AWS and its own chips
AWS launched in 2002, offering companies elastic IT infrastructure that could grow at low cost as needed. Bezos kept
investing although analysts questioned it. In 2023 AWS earned over $90 billion in revenue and over
$24 billion in operating income, in a fast-growing but highly competitive market (p. 1).
For generative AI, Jassy described a stack of three layers, “each of which is gigantic, and each of which
we’re deeply investing”: custom chips for training and inference at the bottom, the Bedrock managed
service in the middle, and applications such as the Rufus shopping assistant and Alexa on top (pp. 3–4).
Textbook
Rothaermel counts the creation of AWS as a backward integration move (p. 301) and also cites AWS as an example of
horizontal diversification (p. 297). The two decks under Go deeper describe AWS as having started as
Amazon’s internal IT; the case itself does not say how AWS began.
Analysis
On the backward reading, Amazon owns the computing its own businesses run on. Selling that computing to other firms
made the input a business in a new industry, which is diversification and goes to Session 13. The chips repeat the
pattern one stage further up, a backward move by the cloud business itself. Contrast NVIDIA in Session 10, which
designs chips but leaves fabrication to others; the case does not say who makes Amazon’s chips.
Question. Does
selling to outside customers make the in-house unit stronger, because it must match the market (Rothaermel, 2024,
p. 318), or does it pull the unit away from serving Amazon?
Backward, taperedCase · pp. 1–2
Content: Amazon Studios behind Prime Video
Amazon launched Unbox for instant video in 2006; Prime Video later offered a large selection of films and shows.
Amazon Studios started in 2011 to produce original content, while licensing deals brought in Viacom shows and, in
2014, select HBO shows for over $200 million (p. 2).
In 2014 Amazon also launched Prime Music, ad-free music streaming for Prime subscribers (p. 1).
In 2018 the Studios earned 10 Golden Globe nominations and two wins (p. 2).
In 2023 Amazon spent over $19 billion on TV and music content, including licensing, against Netflix’s
$13 billion, as the streaming war intensified (p. 2).
Analysis
Tapered backward integration: Amazon makes some of the content its video service needs and licenses the rest
(Rothaermel, 2024, p. 318). The risk is the textbook’s first, higher costs (p. 316), here in a bidding
war for content.
Where the case is
silent. It reports content spending and awards but no profit for video, and it does not separate what content
and what delivery each do for Prime. Question: should Studios be judged as a film studio, or by its effect on Prime
membership and spending (pp. 1–2)? What evidence would tell them apart?
Mostly not verticalCase · p. 3
Whole Foods: a store, not a stage
In June 2017 Amazon paid $13.4 billion, a 27% premium, for an
upscale chain of 460 stores with high-income customers. Whole Foods had a 1.2% share of a US grocery market of about
$800 billion, against Walmart’s 14.5%; AmazonFresh had achieved limited success, and Amazon had not passed a
0.2% share in ten years. Many experts expected the deal to help Amazon into grocery; others wondered whether a
technology company with a soaring share price should enter so thin-margin an industry.
Analysis
Retail is the stage Amazon already occupies, so buying a
grocer is mainly horizontal: a new category and a new format. The textbook offers the forward reading: Amazon’s
physical stores, including those within Whole Foods, move it closer to the end customer (Rothaermel, 2024,
pp. 300–301). Weigh the acquisition as diversification in
Session 13.
Activity · benefits-versus-risks weigher
Weigh one vertical move on the evidence in the case
Done
Pick a move. For each of the textbook’s five benefits
and four risks (Rothaermel, 2024, pp. 315–316), plus Osegowitsch and Madhok’s learning motive (2003,
p. 32), read the evidence line and score how strongly it applies: 0 for no evidence, 1 for some, 2 for strong. The
beam tips toward the heavier side. Then compare your weights with the page’s reading. Evidence lines cite the case
or the textbook; where the case is silent, the line says so.
Benefits 0 · Risks 0
The balance
Benefits
0
Risks
0
Score the factors to tip the beam.
A teaching device, not a valuation: the factors are not equally
important and the scores are not money. Its purpose is to make you say which evidence you are relying on.
What the case lets you conclude
The case reports results for only one vertical stage, computing:
AWS revenue and operating income (p. 1). For devices it gives a reported loss and a write-down; for content, only
spending; for logistics, nothing. Most judgements of value here are inferences, and you should say so when you make
them.
Can a stage be judged on its own profit?
Inside an integrated firm, what one stage earns depends partly on
the transfer prices set between the in-house supplier and its buyer, which the textbook counts among the
administrative costs of integration (Rothaermel, 2024, p. 316). Before you cite the devices loss (case,
p. 2), decide which measure fits a captive stage: its own profit, or what the rest of the chain would lose
without it.
The pattern across moves
Amazon’s forward moves reach for the customer: delivery and
devices (case, pp. 1–2). Its backward moves secure inputs: content and chips (pp. 2–3). Owning a
stage guaranteed nothing: the Fire Phone sold poorly and was written down within three months (p. 2).
Go deeper
Two earlier decks on vertical integration
Neither deck was built for this session, and both cover
diversification as well. They follow the same textbook chapter and use Amazon throughout. Some of their Amazon facts come
from outside this session’s case and were not checked for this page; use this page for the case facts.
Come with a position on each, and the page of the case or reading
that supports it.
01
Is AWS backward integration, diversification, or both? The textbook says both (pp. 297, 301). Show how your evaluation of AWS changes depending on which reading you adopt, and say what you would need to know about how AWS began, which the case does not tell you.
02
The Fire Phone and the Echo were both forward moves into devices. What separated the one written down within three months from the one that sold over 500 million units? Could Amazon have told them apart in advance?
03
The devices unit reportedly lost more than $5 billion a year, yet Alexa places orders on Amazon and Kindle is a way to shop its store (case, p. 2). As Jassy, what evidence would you need before cutting or keeping devices?
04
Prime’s fee rose from $79 to $139 while membership kept growing. Is that evidence that integrating into fulfilment and delivery created value? What else could explain it?
05
Amazon makes some of its video content and licenses the rest. Using the benefits of taper integration, argue whether it should make more or less, given $19 billion of content spending in 2023.
06
Osegowitsch and Madhok say most upstream benefits can be had through closely aligned suppliers. Do Amazon’s own AI chips contradict them? Which benefit or motive would justify them, and what would you need to know that the case does not say?
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
Preparation
Readings for Further study
The case and the articles are copyrighted Harvard
Business School Publishing, Harvard Business Review and Business Horizons materials, and the textbook is in
copyright. Get them through the course pack or the library; this page does not host or link them.
Required · the case
Gupta, S., & Rodriguez, M. L. (2024). Amazon in 2024 (HBS No. 9-514-025). Harvard Business School Publishing.
Four pages of text; read all of it. The same case serves Sessions 12 and 13, so note which moves are vertical as you read.
Required · the outline’s reading
Osegowitsch, T., & Madhok, A. (2003). Vertical integration is dead, or is it? Business Horizons, March–April, 25–34.
The traditional and contemporary motives, and the learning argument for downstream integration. Course-pack reprint BH089.
Required · the outline’s reading
Buzzell, R. D. (1983). Is vertical integration profitable? Harvard Business Review, January.
Read for class. Its text was not in the course-pack folder this page was built from, so the page does not summarise it.
Further · course textbook
Rothaermel, F. T. (2024). Strategic management (6th ed.). McGraw Hill.
Chapter 8, “Corporate strategy: Vertical integration and diversification”. Today: section 8.3 and the alternatives (pp. 310–319), with section 8.2 (pp. 302–310) as the Session 10 recap.
A note on the figures on this page
Amazon figures are the case’s, as of mid-2024, and some
are press reports it cites, such as the devices loss. The case dates AWS to 2002; the textbook gives a different
year, and this page follows the case. Textbook and article examples describe those firms when the books were
written.