Innovation

How Big Companies Build Innovation Engines That Keep Delivering

By Jason Kumpf, Strategy Advisor · November 16, 2026 · 5 min read

How Big Companies Build Innovation Engines That Keep Delivering

The most consistent corporate innovators do not rely on lucky breaks. They build repeatable systems, from long-term research labs to venture studios, that turn ideas into new businesses year after year.

Some of the most important technologies of the past decade came from inside large companies. The Transformer architecture behind today's language models was introduced by Google researchers in 2017. AlphaFold, which earned a share of the 2024 Nobel Prize in Chemistry, came from Google DeepMind. Waymo began as a self-driving car project inside Google's moonshot lab, now known as X.

None of these was an accident. Each came from a deliberate structure designed to give ambitious ideas time, talent and room to grow. The encouraging news for enterprise leaders is that these structures can be studied, adapted and built at any scale.

The payoff from building new businesses

McKinsey's research makes the case in numbers. In a February 2025 survey of 1,039 companies, top performers ranked innovating new offerings as their top investment priority for growth. They were also 78% more likely than other companies to build new businesses in other industries. Of the top 20 global companies by five-year average economic profit, 14 accelerated growth through major innovation investments.

McKinsey's October 2025 survey on corporate venture building, covering 715 senior executives, adds a second layer. It found that 61% of new corporate ventures had passed $10 million in annual revenue, up from 45% in 2023. Respondents said ventures launched in the past five years contribute about 12% of enterprise revenue, and they expect that share to reach about 19% within five years.

Momentum is building on several fronts. Nearly half of respondents, 47%, said ventures launched in the past five years met or exceeded expectations. The average age of ventures above $10 million in revenue fell to about 31 months from 38 months, a sign that companies are reaching scale faster. In a separate analysis, McKinsey notes that half of surveyed CEOs rank new-business development among their top three strategic priorities.

Innovation investment can also sustain growth over decades. McKinsey points to Taiwan Semiconductor Manufacturing Company, which grew revenue by 17% a year from 1995 to 2023.

Model 1: The long-term research lab

Google's research culture shows what patient investment can produce. In 2017, Google researchers introduced the Transformer, a neural network architecture built on a mechanism called self-attention. The team reported that it outperformed earlier models on English-to-German and English-to-French translation benchmarks while training up to an order of magnitude faster.

Just as important, the team released open-source tools so others could build on the work. The architecture went on to underpin many of today's large language models.

The lesson is that a research group with a long time horizon, published work and freedom to explore can create value well beyond its original mandate.

Model 2: The moonshot factory

X, Alphabet's moonshot factory, describes its mission as developing and de-risking early-stage technologies and turning them into products that can support large, sustainable businesses. In X's words, its teams "tackle ideas that have the riskiness of research, and run at them with the focus and speed of a startup."

The track record is long. X's work includes Google Brain, Verily, Wing and Waymo, whose self-driving car project graduated from X in December 2016. Chronicle followed in 2018, Tidal graduated in July 2024 and Taara spun out as an independent company in March 2025.

Google Brain is a good illustration of how far a single project can travel. The deep learning research effort that began at X became a core part of Google's AI work, contributing to the research culture that produced the Transformer.

The moonshot model works because it combines two things that rarely sit together: the ambition of a research lab and the operating discipline of a startup.

Model 3: The venture studio

A corporate venture studio, sometimes called a venture factory, builds new businesses in a repeatable way. McKinsey's May 2025 analysis identifies three building blocks:

  1. Incubation capabilities: a shared pool of expertise, operational support, technology infrastructure and curated data that every venture can draw on.
  2. Parent company assets: structured access to the corporation's customer and supplier networks, milestone-based financing, internal systems and senior leaders.
  3. Portfolio management: a leadership mandate, a data-based approach to choosing ventures, funding released at stage-gate reviews and deliberate talent allocation.

ENGIE Factory, the startup studio of the French energy group ENGIE, shows the model at work. Operating from Singapore, it has launched 12 decarbonization startups. Saudi developer ROSHN Group built a home-buying super-app through its venture factory that helped sell more than 1,500 housing units in its first year.

Investor support is strong as well. McKinsey reports that 90% of investors want new-business-building investment increased or maintained, and that new ventures are expected to drive more than half of companies' growth in the coming years.

Experience compounds. McKinsey found that the most experienced venture builders are twice as likely to succeed as newcomers and see, on average, 12 times more revenue in a venture's fifth year.

AI is speeding up the engine

AI is now part of how the strongest builders work. In McKinsey's 2025 venture survey, advanced AI users reported ventures with about twice the revenue of basic users. Serial builders, companies that launch ventures repeatedly, were more likely to use AI for complex venture-building tasks: 72%, compared with 60% of companies with fewer ventures.

Looking ahead, 84% of respondents expect to integrate AI or automation into their venture-building process within five years, and 56% plan to build data, analytics or AI-driven businesses.

Common traits of durable innovation engines

  • A clear aspiration from the top. McKinsey finds that top performers commit to an innovation aspiration and build a culture around it.
  • A culture that rewards experimentation. Among leaders whose culture rewards risk-taking and experimentation, 68% report ventures that meet or exceed expectations.
  • Building from strengths. About six in ten respondents now build ventures within their primary industry, up from about half in 2024.
  • Investment in people. More than half of successful ventures offer upskilling programs, compared with 42% of less successful ones.
  • Openness. Google's decision to publish the Transformer and share AlphaFold multiplied the value of both.

Building your own engine

An innovation engine does not require starting at Google's scale. It requires a structure: a home for long-term ideas, a repeatable path from concept to venture and a portfolio mindset that backs what is working.

A practical first step is to pick one model, whether a small research group, an internal accelerator or a venture studio, and launch a single pilot venture built on an asset your company already owns. Track it, share what you learn and build the second one faster. As the data shows, each new venture makes the next one more likely to succeed.

Sources

  1. Google Research Blog, "Transformer: A Novel Neural Network Architecture for Language Understanding," August 31, 2017.
  2. Google DeepMind, "AlphaFold: Five Years of Impact," November 25, 2025.
  3. X, The Moonshot Factory, press page, accessed October 2026.
  4. McKinsey Quarterly, "How top performers use innovation to grow within and beyond the core," February 12, 2025.
  5. McKinsey, "The way to win in corporate venturing: Serial building and AI," October 28, 2025.
  6. McKinsey, "The three building blocks of a successful venture factory," May 27, 2025.

More insights from Enterprise Innovation Center