mistake.club2nd Opinion
Back to the archive

The encyclopedia · R&D & Science · Product decision · 2022

IBM spent $4 billion on Watson Health and got a $1 billion exit

IBM's Watson was supposed to revolutionize cancer treatment — instead it became a $4 billion lesson in overhyping AI.

IBM · 2022-01

What happened

IBM's Watson won the quiz show Jeopardy! in 2011, and the company bet its future on the technology. CEO Ginni Rometty declared healthcare would be a $10 billion annual business for Watson. IBM spent $4 billion building Watson Health, including $2.6 billion to acquire Truven Health Analytics in 2016. The pitch was revolutionary: an AI that could read medical literature and recommend personalized cancer treatments better than human doctors.

The system never worked as advertised. Watson for Oncology was trained on a narrow set of patient cases from a single institution — Memorial Sloan Kettering — and simply reproduced the preferences of a handful of U.S. doctors. It struggled with basic tasks like reading medical records with acronyms and errors. In Denmark, local doctors agreed with Watson only a third of the time. The MD Anderson Cancer Center spent $62 million on a Watson pilot before cancelling it in 2017 amid allegations of overspending and mismanagement.

IBM published no peer-reviewed studies proving Watson improved patient outcomes. The company's acquisitions created a patchwork of products that were rebranded as 'Watson' without real AI integration — a practice insiders called 'Blue Washing.' In January 2022, IBM sold the Watson Health division to private equity firm Francisco Partners for approximately $1 billion. The unit was renamed Merative. IBM had invested roughly $4 billion and got back a quarter of that.

Why it happened

  • IBM overhyped Watson's capabilities, marketing it as revolutionary AI when it was essentially a rules-based system trained by a handful of doctors at a single hospital.
  • The system was trained on synthetic data and a narrow set of patient cases, making it unreliable outside the U.S. context where it was developed.
  • IBM's leadership treated Watson as a 'generational meal ticket' and pushed for rapid commercialization before the technology was ready for real clinical use.
  • The acquisition-heavy strategy created integration problems — IBM rebranded existing products under the Watson name rather than building genuine AI capability.
What it cost$4bn invested; sold for $1bncostly

The lesson

A company that sells AI before it works is not doing R&D — it is selling a story about R&D, and the bill comes due when the customer tries to use it.

Sources

spotted an error? The club wants to know.

Comments · 0

    Sign in to join the comments.

    More like this

    2010–2025 · R&D & Science

    NASA's arsenic life press conference was a PR disaster — the science never held up

    Reputational damage; researcher left science; retracted 2025
    2014–2015 · R&D & Science

    The smoking gun of cosmic inflation was a smudge of galactic dust

    Ten months of Nobel talk for a galactic dust signal
    2010 · R&D & Science

    Harvard's star psychologist faked data — and eight counts of misconduct ended his career

    Harvard reputation damaged, papers retracted, career ended

    Somewhere, someone solved the problem this company failed at. 2nd Opinion →