Deep Blue Defeats Garry Kasparov in Six-Game Rematch
In May 1997, IBM's Deep Blue chess-playing system defeated reigning world champion Garry Kasparov over a six-game match by a score of 3½–2½, becoming the first computer system to defeat a reigning world chess champion under standard tournament conditions.

Background
Chess had been a test case for machine intelligence almost since the field began. By the early 1990s, computers could beat club players and most amateurs without much difficulty, but the gap between strong programs and world-class grandmasters remained wide. The best players could sense danger across the whole board and plan many moves ahead in ways that were hard to reduce to rules.
IBM’s chess project had been building toward a serious challenge for years. Feng-Hsiung Hsu, Murray Campbell and Joe Hoane, working at IBM’s Thomas J. Watson Research Center in New York, developed Deep Blue from an earlier system called Deep Thought, which Hsu and Campbell had built at Carnegie Mellon University in the late 1980s. By 1989 the project had moved to IBM and taken on the name Deep Blue. The system combined a conventional alpha-beta search, which works by exploring possible sequences of moves and pruning the ones that cannot improve on what has already been found, with custom VLSI chips designed specifically to evaluate chess positions at high speed. Those chips gave Deep Blue the ability to examine far more positions per second than any software-only approach of the time could manage.
In February 1996, Deep Blue faced Garry Kasparov, the reigning FIDE World Chess Champion, in Philadelphia. Kasparov won that match 4–2. It was the first match of its kind under standard tournament conditions, and Kasparov’s victory left the question open.
What happened
IBM brought the match back in May 1997, again under standard tournament conditions. Kasparov and an upgraded Deep Blue played six games. Joel Benjamin, a grandmaster, worked with the IBM team as a chess consultant, alongside Jerry Brody who contributed to the project’s broader work. The rematch ended 3½–2½ in Deep Blue’s favour. Kasparov won the first game, Deep Blue won the second, and three games ended in draws. Deep Blue then won the sixth and final game to take the match.
The hardware behind the 1997 system was considerably more powerful than the 1996 version. According to Campbell, Hoane and Hsu’s 2002 paper in Artificial Intelligence, the machine could evaluate around 200 million positions per second. The evaluation function, which judged the quality of any given board position, had been refined substantially between the two matches, drawing on analysis of grandmaster games. The search itself remained an alpha-beta approach, but the chess-specific knowledge built into the evaluation was deep enough that the system could weigh positional factors that earlier programs had ignored entirely.
The match attracted attention well beyond the chess world. IBM organised it publicly and broadcast the games widely, which meant that millions of people followed the result in real time. For many observers it was the first moment when a machine had visibly outperformed the best human alive at something demanding sustained, complex thought. Kasparov disputed some of what had happened during the match and called for a rematch with full access to Deep Blue’s game logs; IBM declined and retired the system shortly after. That decision left the result standing, and the debate about what it meant running for years.
Why it mattered
Deep Blue's victory demonstrated that dedicated hardware combined with efficient search algorithms and handcrafted evaluation functions could match and surpass the best human performance in a domain long considered a benchmark of human intellect. The result shifted debate about machine intelligence from whether computers could defeat grandmasters to what such victories did and did not reveal about cognition. It also marked a turning point in industrial investment in AI research, with IBM's public framing of the match drawing sustained attention from governments, corporations, and the research community.
People
Garry Kasparov, Feng-hsiung Hsu, Murray Campbell, Joe Hoane, Jerry Brody, Joel Benjamin
Organisations
IBM, IBM Thomas J Watson Research Center
Sources
- Deep Blue.Artificial Intelligence (Elsevier), vol. 134, pp. 57–83.Primary source
- Deep Blue, Campbell, Hoane, Hsu (2002).ScienceDirect / Elsevier.Primary source
- Deep Blue: IBM's Chess-Playing Computer.Computer History Museum.Secondary
Cite this page
AI Achievements. (1997). Deep Blue Defeats Garry Kasparov in Six-Game Rematch. Retrieved 2026-08-22, from https://achievements.ai/milestone/deep-blue-computer-program-ibm
@misc{achievements_deep_blue_computer_program_ibm,
title = {Deep Blue Defeats Garry Kasparov in Six-Game Rematch},
author = {{AI Achievements}},
year = {1997},
url = {https://achievements.ai/milestone/deep-blue-computer-program-ibm}
}