Deep Blue beat Kasparov with 11.38 GFLOPS in 1996. A B200 GPU today is 4,500,000× more powerful — and rents for $3.75/hr. Here’s what that compression means.
On 10 February 1996, IBM’s Deep Blue became the first computer to defeat a reigning world chess champion. It used 11.38 GFLOPS. A single B200 GPU today delivers 4,500,000× more compute — and you can rent one in five minutes.
Key takeaways
On this date in 1996, Garry Kasparov — the greatest chess player in history — sat across from a 1.4-ton machine and lost. It’s worth pausing to understand what Deep Blue actually was, and what it tells us about where we are now.
Deep Blue wasn’t intelligent. It was a brute-force calculator: 480 special-purpose chess chips evaluating 200 million positions per second. Total compute: 11.38 GFLOPS — gigaflops, not teraflops.
It had one skill. It couldn’t write text, generate images, or reason about anything other than the 64 squares in front of it. IBM built a machine the size of a refrigerator that could beat humanity’s best chess player, and it could do nothing else.
In 1996, it took a purpose-built supercomputer to do one thing well. Today, a single GPU runs large language models, trains neural networks, generates images, simulates molecular dynamics, and serves real-time inference — all as different workloads on the same hardware.
The B200 that runs a 70B Llama model could equally run a protein folding simulation, a diffusion model, a reinforcement learning agent, or a video generation pipeline. The same chip, the same drivers, the same CUDA stack. That’s what general-purpose GPU compute means, and it’s the shift that made the current AI wave possible.
Deep Blue required years to design and cost millions of dollars. A developer anywhere in the world can now launch a B200 on packet.ai in under five minutes for $3.75/hr — and that B200 delivers roughly 4,500,000× the compute of the machine that beat the world chess champion.
That compression — from years and millions to minutes and dollars — is the real story. It’s not just that compute got faster. It got accessible.
Kasparov’s response to losing to Deep Blue is worth remembering. Rather than concluding that computers had made human chess players obsolete, he pioneered “Advanced Chess” — the idea that a human working with a computer is stronger than either alone.
Thirty years later, that principle scales to everything. The teams building the most capable AI products aren’t replacing human judgment with models — they’re building systems where models handle the scale and humans handle the direction. The GPU infrastructure that makes that possible is now accessible to anyone.
Last reviewed: 10 June 2026. Rent a B200 on packet.ai →
Same models. Same API. Fraction of the cost. Start free — no credit card required.
Start Building →