It's all in the work units

ArmsControlWonk discusses estimates that Iran is years away from the nuclear bomb. The calculations raise questions about how estimates are made. Regardless, we shall proceed with some of his cited numbers.

Overall, Iran is probably a little less than a decade away from developing a nuclear weapon...
Each of Iran’s centrifuges has an output between 2-3 SWU/year...
Iran probably only has about 700 centrifuges, as well as components for another 1,000 or so...
So, the real question, however, is how quickly Iran could assemble and operate 1,500 centrifuges in a crash program to make enough HEU for one bomb (say 15-20 kg).

It seems key to such estimates as to how efficient Iran's centrifuge technology is. The estimate of 2-3 SWU/year appears to be low. By contrast, the Wisconsin Project estimates a higher SWU, as well as noting that there may be smaller enrichment facilities in existence besides Natanz.

Rather than the Iranian estimate of about six or seven separative work units (SWU) per centrifuge per year, the IAEA estimated that the throughput of Iran's centrifuges could be as high as 12 to 14 SWU per machine per year, according to the media report.

Taking low end of the Iranian estimate, that would result in half the production time required for a crash nuclear program. With the lower bound IAEA estimate, that would result in a quarter of the time. SWU efficiency is key. Why are the estimates so high?

GlobalSecurity has further numbers for comparison.

A single centrifuge might produce about 30 grams of HEU per year, about the equivalent of five Separative Work Unit (SWU). As as a general rule of thumb, a cascade of 850 to 1,000 centrifuges, each 1.5 meters long, operating continuously at 400 m/sec, would be able to produce about 20-25 kilograms of HEU in a year, enough for one weapon. One such bomb would require about 6,000 SWU.
With current technology, a single gas centrifuge is capable of about 4 separative work unit [SWU] annually, while advanced gas centrifuge machines can operate at a level of up to perhaps 40 SWUs annually.

The higher Iranian and IAEA estimates, assuming they are sound, indicate that Iran's centrifuge technology is more advanced than the single gas centrifuge, due to cascading. Assuming an estimated 700 assembled centrifuges are kept operational at the level of performance claimed by Iran, with the spares being used to replace worn parts, they would be able to produce 4,200 SWU per year, yielding enough material for a single weapon in approximately 1.5 years. IAEA estimates would result in this time being reduced to under 9 months. Weaponization would take longer.