"It wasn't a conscious decision to emulate the Twin Towers," Foster said on the phone from New York in December, still audibly on a high from his presentation of the day before. "At first we designed something completely different - a single, supertall tower with a pinnacle top. Then we tried a cluster of three towers, low, medium and tall. But it was when we were discussing the project with the sculptor Anish Kapoor that the concept emerged of two towers that came together. Afterwards, you could see the connection with the old Twin Towers, but that wasn't how we started."

The Foster practice has had its share of fierce criticism, not least from me, over some of the less-than-masterful buildings that are sometimes churned out from what is one of the biggest architectural production-lines in the world. His City Hall for London Mayor Ken Livingstone has had mixed reviews, to say the least. But when Foster takes full command of a project personally, as he has here, he is able to summon The Force. Top talents are wheeled in to help - in this case ranging from Kapoor to the consultancy Space Syntax, which analyses pedestrian movements.

From his own office, no fewer than 41 other architects were credited on the design, including three of his partners who normally run huge jobs on their own - like the British Museum's Great Court, or his new Zeppelin-shaped Swiss Re tower in the City of London. Then there were the independent computer-graphics wizards who created the virtual-reality world of the new World Trade Center, set alongside Foster's own hand-drawn sketches of a little girl wandering through the new complex with her parents. All this for what was officially meant to be a broad-brushstroke, concept design. The entrants were paid just $40,000 each, and for this reason Frank Gehry refused to take part, on the grounds that they could not be serious. This all-singing, all-dancing entry cost Foster big. It was his biggest gamble yet. "And do you know," remarked a New Yorker friend of mine in mid-January, "he's not going to get it." My friend had been listening to the comments of the public at the exhibition of the schemes, had detected the shift among the experts towards Libeskind and Vinoly, and reckoned the Foster scheme was wrongly packaged for the pragmatic Manhattan way of going about things.

To succeed, the winning architect - or group of architects, in the case of several entries - must first satisfy the families of the victims of 9/11 by incorporating an appropriate memorial. Next they must get everything working much better at ground level, from pavements to subway lines (Foster points out that the transport interchange beneath his 20-acre memorial park, stretching down to the river, is bigger even than the towers). Then they must provide the right-looking shape on the skyline - a symbol of rebirth and national defiance. Finally - and this is always how it has worked in Manhattan - they must make sure that everything stacks up financially. If the numbers don't work in the world capital of capitalism, then it won't get built. This is where Foster may have miscalculated. His new twin towers look great, but are just not needed at present.