The Game Chip and the Scientists

Learning objectives

1. Understand why a programming platform needs an installed base before developers commit.
2. Weigh exploiting a known market against exploring an unknown one.
3. Recognise survivorship bias in famous platform bets.

Session plan

1. 0–5 min Read the situation, what is known and unknown, and the voices in the room. No discussion yet.
2. 5–13 min Each participant chooses an option on the case page, states a confidence and writes a one-line pre-mortem: if this choice fails, the most likely reason is…
3. 13–28 min Debate by option. Ask each group what it would tell investors about the cost of its option.
4. 28–33 min Reveal the outcome.
5. 33–50 min Discuss why the process is graded Sound rather than Strong: what made the bet reasonable, and what made it depend on luck and on the company's ability to wait?
6. 50–60 min Transfer: participants identify a capability that could become a platform in their organisation if it were put where users already are.

The case page records each participant’s choice, confidence and pre-mortem in their own browser, and scores the confidence once the outcome is revealed. Nothing is sent anywhere, so it can be used in any classroom.

The key facts on one page

• The company's first chip with the unified architecture launched on 8 November 2006; a free public beta of the programming toolkit followed in February 2007.
• A dedicated computing card launched in mid-2007 at about $1,499.
• In 2008 the company took a $196 million charge for defective notebook chips, and its share price fell 76 percent over the year, for several reasons.
• In 2012 a deep neural network trained on two of its gaming cards won the ImageNet competition.
• In fiscal 2026 its data-centre revenue was $193.7 billion.

Where discussions go wrong

• Reading the outcome back into the decision. In 2006 the size of the market was unknowable.
• Attributing the whole 2008 share-price fall to the bet; a product defect and the financial crisis also struck that year.

The forces at work

• The chicken and the egg: Programmers would write software for the chips only if many people owned them, and buyers would value the capability only if useful software existed. Putting the hardware in every gaming PC was a way to build one side of the market before the other existed.
• Exploring against exploiting: The gaming business was large, competitive and profitable. Every square millimetre of chip spent on computing for unknown customers was taken from a known market, which is the classic tension between exploiting what works and exploring what might.
• Lock-in, for good or ill: The first programming model to win developers' habits, libraries and code would be very hard to dislodge later. That made being early valuable, and made the cost of being early worth bearing.
• An unknowable market: There was no data on how large general-purpose computing on graphics chips might become. The decision could not be settled by forecasting, only by judgement about what an install base might make possible.
• The story we tell afterwards: This bet is famous because it worked. Many companies made costly platform bets that never found their market, and they are rarely studied.

The lesson to carry out of the room A capability that might become a platform has to be put where people already are, and paid for long before anyone can prove it was worth it.

Pairs well with: The idea Markets With Two Sides and the essay on compute.