Of the 87 large U.S. public companies that ran with co-CEOs between 1996 and 2020, the average delivered annual shareholder returns of about 9.5%, versus 6.9% for the indexes they sat in—nearly 40% better. That comes from a Harvard Business Review analysis by Marc Feigen, Michael Jenkins and Anton Warendh.
For decades that finding was a curiosity, because almost nobody did it: fewer than 100 of 2,200 companies in the S&P 1200 and Russell 1000 tried the model over that stretch. This year, the curiosity became a pattern. Oracle, Spotify and Comcast all moved to two chief executives. The question worth asking isn't whether it's a trend. It's what these companies know about the job that their org charts haven't caught up with.
The Job Got Too Wide
The clearest explanation came from Axios C-Suite this week, reviewing 2026's leadership changes. All seven Fortune 100 companies that completed a planned CEO transition this year promoted long-tenured insiders — and several hedged by elevating more than one person. Jim VandeHei put the logic in one line: "One person can't keep sales numbers and AI enablement balanced in their brain."
That's the real driver. The CEO role now spans two jobs with different clocks. One is running the existing business — quarterly numbers, customers, operations. The other is rebuilding how the company works while it's still running. Look at how the new pairs split: at Oracle, founder Larry Ellison moved to executive chair while two operators took the top seat. At Spotify, Daniel Ek stepped up to executive chair and handed the company to Gustav Söderström and Alex Norström — a product leader and a business leader.
Fortescue, the Australian mining group, made the split explicit: Dino Otranto runs metals and operations, Agustin Pichot runs growth and energy. "What I love about it is you can't have an ego," Otranto told The Irish Times, "and the ego of a CEO is a really dangerous thing."
Why the Returns Aren't Automatic
The 9.5% figure is real, and it's also the most misused number in this conversation.
For every Netflix — which has run two CEOs since 2020 — there's a cautionary tale. The HBR authors flag co-CEO structures that buckled under stress at SAP, Chipotle and Research In Motion, the company behind BlackBerry. INSEAD's José Luis Alvarez and Copenhagen Business School's Silviya Svejenova, who have studied shared executive power for years, add Unilever, Airbus and Deutsche Bank to the list of arrangements that didn't hold.
BlackBerry is the instructive one. Its co-CEOs, Jim Balsillie and Mike Lazaridis, built a category together. "I don't know how we could have done it without the two of us," Balsillie has said. Then the iPhone arrived, and a structure designed for steady growth had to make a fast, existential call. Two heads can be better than one. Two heads that need to agree before moving can also be slower than one.
The data also show a selection effect. Companies that choose co-CEOs are often large, complex, and already strong enough to experiment with governance. The outperformance tells you the model can work. It doesn't tell you it will work for you.
The Conditions That Separate Them
What's useful is that the research converges on the same short list of conditions — and they're specific enough to check.
- Split the work, not the title. Both INSEAD and HBR put clear, predetermined decision rights first. Co-CEOs who share everything paralyze the organization; co-CEOs who own separate domains move faster than a solo leader could. Gensler, the architecture firm, runs co-CEOs Jordan Goldstein and Elizabeth Brink on exactly that principle:
"We each lead different areas with clarity and stay aligned through frequent check-ins, operating with a one-voice philosophy."
- Fight in private. Every failure case shares one symptom: disagreements that reached employees, the board, or the press. Once staff learns which leader to lobby, you don't have two CEOs — you have two factions.
- Pick complements, not clones. The winning pairs combine different expertise with compatible temperament. Two brilliant salespeople are a coin flip. A builder and an operator who trust each other is a structure.
- Write the ending first. The HBR authors are emphatic about an exit plan — what happens if one leaves, if they deadlock, or if the board decides the arrangement has run its course. Pairs without one tend to end badly and publicly.
What This Means
For most Zilck readers, the takeaway isn't about Oracle's boardroom. It's about the partnership you may already be running.
If you co-founded your business, you're already in a co-CEO arrangement whether you named it or not — and the failure modes are the same ones that sank much larger companies. Write down who owns which decisions. Agree now how you'll break a tie. Decide what happens if one of you wants out. Those three conversations are cheap today and ruinous to have for the first time during a crisis.
If you're a solo leader feeling stretched between running the business and reinventing it, you've identified the real problem Oracle and Spotify were solving. The answer may not be a second CEO. It may be a second-in-command with genuine authority over one of those two clocks — so you can actually hold the other.
Two at the top isn't the future of leadership. It's a sign of how wide the job has become. The companies getting it right aren't doubling the boss. They're dividing the work.
Discussion