# The build vs. buy vs. partner decision framework
In May 2024, Cisco closed its $28 billion acquisition of Splunk, the largest deal in Cisco's 40-year history. Six months earlier, the company had quietly killed an internal observability project that had consumed $400 million and four years of engineering effort. CFO Scott Herren made the trade-off explicit on the post-deal call: "We could have spent another three years and another billion dollars trying to build what Splunk already has. By then, the market would have moved past us."
That single sentence captures the central tension every CFO faces in corporate development: the cost of building is rarely the cost of building, it's the cost of being late.
Yet for every Cisco-Splunk, there's a Quaker-Snapple. Quaker Oats paid $1.7 billion for Snapple in 1994 and offloaded it for $300 million 27 months later, a $1.4 billion incineration that ended CEO William Smithburg's career. Microsoft paid $26.2 billion for LinkedIn in 2016 and now generates over $15 billion in annual revenue from it. Both deals were modeled by intelligent finance teams. Both passed board scrutiny. Only one created shareholder value.
The difference wasn't luck. It was framework discipline.
Most CFOs intuitively reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it. for a comparison: model the of building internally, the of acquiring, and pick the higher number. This is the wrong starting point, and it's how disasters happen.
The reason: NPVNPVNet Present Value is the sum of an investment's future cash flows discounted to today, minus the initial outlay. A positive NPV signals value creation.View full definition → comparisons assume the three paths produce equivalent assets. They don't. Building yields a capability tightly coupled to your culture and roadmap. Buying yields a capability with embedded customer relationships, regulatory approvals, and, critically, talent that may walk out the door 18 months post-close. Partnering yields *optionality* but rarely *ownership* of the economics.
McKinsey's 2024 study of 1,200 transactions over $500M found that deals where the acquirer had previously attempted to build the same capability internally outperformed by 14 percentage points in three-year TSR. Why? Because failed build attempts produce *informed buyers*. You understand the technical complexity, the talent scarcity, and, most importantly, your own organization's capacity to execute.
Before any financial modeling, run the candidate through three filters:
1. The Time-to-Capability Test. What is the half-life of competitive advantagecompetitive advantageA lasting edge over competitors: a resource, capability or position they cannot easily replicate, letting a firm earn above-average returns over time.View full definition → in this category? In generative AI infrastructure (2026), capabilities depreciate in roughly 18 months. Building a Tier-1 LLMLLMA Large Language Model is an AI system trained on vast text data to predict and generate language, enabling tasks like writing, summarizing, and answering questions.View full definition → team from scratch takes 36+ months. The math is brutal, you cannot build fast enough. This is why Salesforce paid $1.9 billion for Own Company in September 2024 rather than rebuild data protection internally, and why ServiceNow has executed 11 tuck-in AI acquisitions since 2023.
2. The Strategic Adjacency Test. Is the target capability a *core* extension of your existing business, or a *new* business altogether? Disney's $71.3 billion Fox acquisition (2019) was a core adjacency, content for an existing distribution machine. Quaker-Snapple was not. Quaker's distribution genius lay in warehouse-supplied grocery channels; Snapple lived on direct-store-delivery routes through small distributors. The acquired asset required a *different* operating system, and Quaker tried to force-fit it. The same trap killed eBay's $2.6 billion Skype deal (2005).
3. The Cultural Absorption Test. Can your organization metabolize this team within 24 months? Microsoft kept LinkedIn operationally independent under Jeff Weiner, then Ryan Roslansky, a deliberate choice CFO Amy Hood codified into the deal structure. Compare this to HP's $11.1 billion Autonomy acquisition (2011), where forced integration destroyed the asset (alongside, allegedly, accounting fraud) within 14 months.
Once the strategic filters are passed, the CFO's job is to build *three comparable* models, not three different models with three different assumption sets.
The single most common error in build models is underestimating opportunity cost. The cost of building isn't the engineering payroll, it's what those engineers *aren't* doing. At a company like Adobe, with a fully-loaded senior engineer cost of approximately $450K annually, redirecting 80 engineers to a new initiative for three years carries a direct cost of $108M. But the *opportunity cost*, the features not shipped on the core product, typically runs 2-3x the direct cost.
Your build model must include:
That last line item is where most CFOs are too optimistic. If the addressable market is growing at 30% annually and you're three years late, you're not entering the market you modeled, you're entering a market with entrenched competitors and 60% lower margins.
The acquisition price is the down payment. The full cost includes:
Under Pillar Two, deals that previously generated tax synergies through IP migration to low-tax jurisdictions no longer do. CFOs who haven't rebuilt their M&A tax models since 2024 are working from obsolete assumptions.
The partner model is the hardest to quantify because it's not really one model, it's a probability tree. A well-structured partnership typically includes:
The classic case: Roche's 2009 acquisition of Genentech for $46.8 billion came after a 19-year partnership during which Roche held a controlling but non-consolidating stake. Roche paid a premium, but they bought an asset they had de-risked over two decades. Pharma rediscovered this playbook in 2024-2025, with Merck, Pfizer, and Lilly all increasing partnership-with-option structures for early-stage biotech.
Knowledge check
1. According to Cisco CFO Scott Herren's rationale for the $28B Splunk acquisition, what was the true cost driver that justified buying over continuing to build?
2. What was the financial outcome of Quaker Oats' 1994 acquisition of Snapple?
3. Why does the lesson argue that a simple NPV comparison across build, buy, and partner paths is the wrong starting point?
4. Select ALL correct answers about what an acquisition delivers that internal building typically does not, according to the lesson.
Sélectionnez toutes les réponses correctes.
5. Select ALL correct answers about the strategic characteristics of a partnership path (versus build or buy).
Sélectionnez toutes les réponses correctes.
In February 2024, Novo Holdings (the controlling shareholder of Novo Nordisk) announced a $16.5 billion acquisition of Catalent, the contract manufacturer. Three Catalent fill-finish facilities were then sold to Novo Nordisk for $11 billion to expand Wegovy and Ozempic production capacity.
This deal is a masterclass in framework application. Novo Nordisk faced an existential capacity constraint, they were rationing GLP-1 supply globally while Eli Lilly's Mounjaro was eating market sharemarket shareThe percentage of total industry sales your company captures in a given period. It measures competitive position relative to rivals in a defined market.View full definition →. The Build vs. Buy vs. Partner analysis:
CFO Karsten Munk Knudsen described the strategic logic as "buying time we cannot manufacture internally." The deal closed in December 2024. By mid-2025, Wegovy supply constraints had materially eased and Novo Nordisk's market sharemarket shareThe percentage of total industry sales your company captures in a given period. It measures competitive position relative to rivals in a defined market.View full definition → recovered.
The framework worked because all three tests were passed: time-to-capability was decisive, the adjacency was core (manufacturing for their own products), and cultural absorption was minimal (the facilities operated semi-independently under a manufacturing services model).
In August 2020, Teladoc acquired Livongo for $18.5 billion in stock, the largest digital health deal in history. By 2022, Teladoc had written down $13.4 billion of the goodwill. The stock fell 95% from peak to trough.
Where did the framework break? Three places:
1. Strategic adjacency test failed: Teladoc was a synchronous telehealth platform (live doctor visits). Livongo was an asynchronous chronic care management platform (data and coaching). Leadership called this "diabetes meets video" but the customer acquisition channels, billing models, and clinical workflows had near-zero overlap.
2. Build option was never seriously modeled: Teladoc's internal analysis assumed Livongo's $4.5 billion revenue run-rate was uniquely unreplicable. In retrospect, multiple competitors (Omada, Virta, even CVS through Aetna) built comparable capabilities for under $500M each.
3. The COVID-era valuation embedded permanent assumptions about temporary behavior: The deal was modeled assuming sustained 60%+ virtual care penetration. By 2023, virtual visits had reverted to 25% of total volume.
The Teladoc-Livongo deal violated a fundamental rule: never build your M&A model on top of a macro assumption you cannot independently verify. CFOs who modeled COVID-era growth as permanent, in any sector, paid for it.
When the CEO walks into your office Monday morning with a target name, here's the disciplined sequence:
1. Demand the build alternative model before any deal discussion. If your corporate development team cannot produce a credible internal-build model within two weeks, you do not have enough information to evaluate the acquisition. The build model is your bargaining anchor and your sanity check.
2. Apply the three strategic tests as a gating mechanism, not a tiebreaker. Time-to-capability, strategic adjacency, and cultural absorption are veto items. A deal that fails any one of them should require board-level override, not financial engineering to make the NPVNPVNet Present Value is the sum of an investment's future cash flows discounted to today, minus the initial outlay. A positive NPV signals value creation.View full definition → work.
3. Model partnership-with-option structures for any deal above $1B in volatile categories. In AI, biotech, and climate tech, the optionality premium of waiting 18-36 months to acquire often exceeds the price escalation. Structure commercial agreements with embedded conversion mechanics.
4. Rebuild your tax synergy assumptions under Pillar Two. If your last M&A tax framework predates 2024, every cross-border deal model is wrong. The 15% global minimum has