Supercomputers will achieve one human brain capacity by 2010, and personal computers will do so by about 2020.

Kurzweil is making a concrete, date-stamped prediction typical of his “law of accelerating returns” framework: that computing power (and, by implication, the feasibility of brain-level AI) will grow exponentially and reach a rough equivalence with the human brain on consumer hardware within a few decades. The claim hinges on treating “brain capacity” as something that can be approximated in computational terms (e.g., operations per second and/or memory), and it functions rhetorically to compress the future—turning a speculative idea into an imminent engineering milestone. Read today, it also illustrates the risks of precise technological forecasting: even if trends continue, definitions (what counts as “brain capacity”?) and bottlenecks (architecture, energy, software) can make calendar predictions miss while the broader trajectory remains influential.