# kengelbach-bcg-non-ai-companies-ai-dealmaking-2026-09-21

## Veille

Article by **Jens Kengelbach**, **Daniel Friedman**, **Lianne Pot**, **Georg Keienburg** and **Tobias Söllner** (**Boston Consulting Group**), published on **September 21, 2026** in the series *The 2026 M&A Report* (~12 min). The displayed title is "How Non-AI Companies Win at AI Dealmaking"; the page and URL title is "How Non-Tech Companies Should Buy AI in 2026."

## Titre Article

How Non-AI Companies Win at AI Dealmaking

## Date

2026-09-21

## URL

https://www.bcg.com/publications/2026/how-non-tech-companies-buy-ai

## Keywords

AI M&A, M&A, AI dealmaking, Adopt Partner Build Buy, build or buy, competitive differentiation, market availability, six signals, lightest structure, reversibility, lock-in, optionality, alliances, joint ventures, announcement return, non-tech buyers, proprietary data, intellectual property, deploy-reshape-invent, Schneider Electric, Cognite, L'Oréal, Nvidia, Shell, SparkCognition, 2026 M&A Report

## Authors

Jens Kengelbach, Daniel Friedman, Lianne Pot, Georg Keienburg, Tobias Söllner — Boston Consulting Group ; série *The 2026 M&A Report*.

## Ton

A strategy consulting publication, analytical and prescriptive in register, aimed at executives and M&A leadership. The text moves from quantified findings (deal and alliance volumes, announcement returns, five *exhibits*) to a decision framework, then to sector cases. It tempers its own framework: the signals rarely all point in the same direction, a lighter structure is not necessarily easier (nine out of ten industry-tech partnerships miss their objectives according to a BCG study), and a company unable to reshape its workflows will underperform in every quadrant. Announcement returns are presented as an early indication, not as proof of value creation. The text refers to BCG's *deploy-reshape-invent* progression.

## Pense-betes

- **Instinct to buy**: paying a control premium on capabilities that quickly become commoditized; the real question is *where ownership matters*.
- **Exhibit 1**: 1,355 AI M&A deals in 2025, non-tech buyers **53%**, majority on software and applications.
- **Exhibit 2**: 1,340 AI alliances and JVs in 2025, almost as many as acquisitions.
- **Exhibit 3**: announcement returns **0.3%** (hardware), **0.4%** (cloud, data centers), **0.5%** (software); non-tech buyers of AI software **0.8%+**.
- **Two questions**: does the capability differentiate? (for a non-AI company, rarely the model, more often data, workflows, IP, expertise); is it available on a competitive market, or tied to a specific provider, team, or asset?
- **Exhibit 4**: Adopt (general-purpose copilot, standard document processing), Build (forecasting on proprietary data, business copilot on in-house IP), Partner (maturing sector capability, co-development with a provider), Buy (proprietary external data, rare team). Placement varies by company.
- **Exhibit 5**: differentiation; urgency and commoditization; market capacity; integration, governance and control; target economics; optionality and lock-in.
- **Trend**: the decision most often tilts toward Partner or Build. Before signing: data and IP ownership, portability, benchmarking rights, exit clause.
- **Sector rule**: buy when AI becomes a competitive advantage, partner when it is shared infrastructure. Corollary: oilfield services companies are becoming AI sellers.
- Neighbor of [[padhi-mckinsey-ai-operating-system-ceos-2026-10-01]].

## RésuméDe400mots

BCG devotes a section of its 2026 M&A Report to how companies outside the technology sector should acquire AI capabilities. The instinct to buy is common, but models and tools are advancing so fast that one risks paying a control premium for capabilities soon outdated or commoditized. Many are available off the shelf or through a partner: the challenge is knowing where ownership truly matters.

The market has transformed. AI-related acquisitions surged in 2025 across every layer of the stack, with 1,355 deals, and non-tech buyers accounted for the majority of acquisitions of software and AI application companies. Alliances and joint ventures more than doubled between 2024 and 2025, exceeding 1,300, and now rival M&A. Investors welcome these deals: announcement returns rise from 0.3% for hardware to 0.5% for software, and exceed 0.8% when the buyer of AI software is not a tech company, while tech buyers trigger almost no reaction. This premium rests on an expectation, which raises the bar for results.

The authors propose evaluating each capability according to its ability to differentiate the company and its availability on the market. This yields four options: adopt a market solution, partner with a provider, build from one's own assets, or buy a rare asset. The guiding principle is to choose the lightest structure that provides the necessary access or control while preserving the freedom to change course. Six signals refine the choice: differentiation, urgency and commoditization, market capacity, integration and governance, target economics, and optionality and dependency. Most often, the decision tilts toward partnering or building. A lighter structure is not necessarily simpler: nine out of ten industry-tech partnerships miss their objectives, hence the importance of clarifying data, intellectual property, portability, and exit terms.

Three sectors illustrate the framework. In industry, Schneider Electric acquires Cognite to control the data layer of its products. In consumer goods, L'Oréal partners with Nvidia, its advantage lying in its biological data on skin and hair. In energy, Shell allies with SparkCognition for exploration. The common rule: buy when AI becomes a competitive advantage, partner when it is shared infrastructure. Whatever the structure, the company must absorb the capability and reshape its workflows. The winners will make ownership the exception and adaptability the rule.

## GrapheDeConnaissance

- Jens Kengelbach —publie→ article How Non-AI Companies Win at AI Dealmaking (DOCUMENT, 0.97)
- Jens Kengelbach —travaille_chez→ Boston Consulting Group (ORGANISATION, 0.95)
- article How Non-AI Companies Win at AI Dealmaking —fait_partie_de→ 2026 M&A Report (DOCUMENT, 0.95)
- Boston Consulting Group —a_créé→ grille Adopt Partner Build Buy (METHODOLOGIE, 0.96)
- grille Adopt Partner Build Buy —utilise→ six signaux de sécurisation d'une capacité IA (METHODOLOGIE, 0.94)
- grille Adopt Partner Build Buy —affine→ deploy-reshape-invent (CONCEPT, 0.82)
- Boston Consulting Group —recommande→ choisir la structure la plus légère qui donne l'accès ou le contrôle nécessaire en gardant la liberté de changer (AFFIRMATION, 0.96)
- Boston Consulting Group —affirme_que→ pour une entreprise non-IA, l'avantage vient rarement du modèle mais des données, workflows, PI et expertise qui l'entourent (AFFIRMATION, 0.94)
- Boston Consulting Group —mesure→ 1 355 acquisitions IA en 2025, dont 53 % par des acheteurs non-tech (MESURE, 0.93)
- Boston Consulting Group —mesure→ 1 340 alliances et coentreprises IA en 2025, plus du double de 2024 (MESURE, 0.92)
- Boston Consulting Group —mesure→ rendement à l'annonce de 0,8 %+ pour un acheteur non-tech de logiciel IA, ~3× la moyenne (MESURE, 0.93)
- Boston Consulting Group —mesure→ environ neuf partenariats industrie-tech sur dix manquent leurs objectifs (MESURE, 0.88)
- Schneider Electric —utilise→ Cognite (TECHNOLOGIE, 0.85)
- L'Oréal —collabore_avec→ Nvidia (ORGANISATION, 0.95)
- Shell —collabore_avec→ SparkCognition (ORGANISATION, 0.94)
- Boston Consulting Group —affirme_que→ acheter l'IA quand elle devient avantage concurrentiel, s'associer quand elle est infrastructure partagée (AFFIRMATION, 0.93)
- Boston Consulting Group —prédit→ à mesure que modèles et outils se banalisent, l'avantage résidera dans les données propriétaires, workflows et expertise (AFFIRMATION, 0.9)
- article How Non-AI Companies Win at AI Dealmaking —affirme_que→ make ownership the exception and adaptability the rule (CITATION, 0.95)

---
Canonical: https://www.thekb.eu/en/fiches/kengelbach-bcg-non-ai-companies-ai-dealmaking-2026-09-21/