Colchester

I'm puzzled. I am standing in Firstsite, a new building in Colchester, designed by a celebrated architect, and achieved mostly with public money, plus a certain amount of blood, sweat and tears. Its purpose is to display and communicate visual art, and educate about it, yet the more I look, the more it seems designed to make it unusually difficult to mount an exhibition.

A great wall, which might be a nice place to put pictures, not only curves but also slopes outwards as it rises. Other gallery walls also curve or are made of glass. Some spaces are very high, to no purpose. On the rare occasions when a plain, blank piece of wall presents itself, it usually gets punctured by doors. Firstsite will show temporary exhibitions of contemporary art, and say that "art practice has changed so much in recent years; artists are creating work in so many different media", so the idea seems to be that flat surfaces for fuddy-duddy paintings would not be needed as much and there would be installations and sculptures instead. Except the slope of the walls narrows the space at ground level, precisely where you would most want room to circulate around large objects. Oh well, perhaps they can project some video pieces. Or would, if a profusion of windows at many levels did not make much of it almost impossible to black out.

The art gallery that is tough on art is not a wholly new experience. Frank Lloyd Wright's Guggenheim in New York demands that you view art from a continuously sloping, curving and none-too-wide ramp. Zaha Hadid's Maxxi in Rome has its share of tricky angles and hard-to-fill spaces. But both these museums have a splendour and conviction which might, and do, inspire curators to rise to the creative challenge of animating the spaces. Firstsite's curves and slopes and variegated volumes have a certain intrigue, but it's hard to see them working like the Guggenheim, not least because Colchester is neither New York nor Rome, and so will find it hard to mount the big-budget productions that could transform these spaces. From time to time, curators will find interesting things to do with a slope or a curve. Their problem will be that the building requires that they do it every time.

My puzzlement is a cousin of a mystification prompted by the extraordinary success of its architect, Rafael Viñoly, especially in this country, over the last decade. He has masterplanned the Radcliffe Observatory Quarter in Oxford, the largest single building project the university has undertaken, plus the design of the Mathematical Institute.

He has created the masterplan for the eternally deferred development around Battersea power station and he seduced Peter Rees, chief planner of the City of London, into enthusiastic support for his "Walkie Talkie" skyscraper at 20 Fenchurch Street, which is now under construction. He has designed Curve, a theatre in Leicester, and Firstsite, and was commissioned to produce an unrealised plan for the South Bank.

This impressive portfolio – more than Zaha Hadid or David Chipperfield, much more than the perennially shunned Frank Gehry has achieved over here – suggests a truly exceptional talent or great powers of organisation and delivery. If you speak to his clients, a tone of awe comes into their voices, as if they are truly privileged to have secured his services, and it is their honour and delight to clear up such technical glitches as have accompanied his projects. The RIBA's awards panel called Curve "genuinely iconic… a new level of ambition in theatre design" and his practice has been called "blazingly successful". Yet, although he has flair, skill and energy, and an international array of completed works, I don't quite understand the extent of his popularity with commercial and cultural clients.

Viñoly was born in Uruguay, then built a successful practice in Argentina in the 1970s, before moving to New York, where he built up a portfolio of substantial, well-received projects. His big break came with the Tokyo International Forum, a $1.5bn complex of auditoriums and exhibitions, completed in 1997. He also led the team chosen to masterplan the rebuilding of Ground Zero, until Governor Pataki overruled his advisory committee and chose Daniel Libeskind instead. It was probably a job it was good not to win.

He is charming and charismatic, and a talented musician whose Steingraeber und Söhne, one of several grand pianos he owns, sits prominently in his New York office. His homepage opens with a film of a magical hand producing a fluent sketch of a tower, to reinforce his creative aura. He makes no claim, as some architects do, that his practice is about teamwork: he is the one and only designer of his buildings. He passionately advocates the importance of spending an extra "20-25%" over a basic building budget to achieve good architecture: "25% is what you need… people don't understand how important good architecture is."

His projects have had blips, although precise responsibility for these, as always in building projects, is a complex subject. Curve cost £61m against an original budget of £26m, opened late, and was roundly condemned as poor value for money by the Audit Commission. Viñoly was sued by the Kimmel Centre in Philadelphia over cost overruns and delays, and by the Boston Convention and Exhibition Centre over technical defects, both cases being settled out of court.

At Battersea, he proposed a 1,000ft-high glass funnel, claiming that it was fundamental to achieving a zero-carbon development, when in fact it would have required an immense amount of energy to build, and taken decades to get payback.

In Colchester, the complexity of Firstsite's curves and angles proved too much for its contractor to handle. Building work ground to a halt, creating an embarrassment to the then Tory-run borough council. The Liberal Democrats, with the help of pictures of the half-finished shell on their election literature, then won control of the council. They now say that, "with realigned baselines", the project is on time and budget, which means that it has cost £28m against an original target of £18m and taken five years to build.

There is a pattern to his projects. They have whoosh and sparkle and make direct appeal to the glitter-loving magpie inside us. The exterior of Firstsite is a long, gold-clad crescent, which has inevitably had it nicknamed "the golden banana". When asked to explain his choice of colour, Viñoly shrugs winningly and says: "Why not?" And it is no bad thing if the first view of a building cheers you up.

He also aims to achieve what he calls "the idea that justifies the extra 25%". At Firstsite, this idea is his decision to relocate the building away from the smaller plot set by the brief, so that it could spread over the site and enable interconnection between the education, exhibition and other functions of the building. With the Walkie Talkie, the idea is to bulge outwards at the top – although the alleged beauty of the resulting shape eludes me – so as to maximise floor space where it is most valuable.

But the closer you look at his projects, the less sense they tend to make. At Oxford, he chose a radiating plan of straight avenues, like Parisian boulevards, which have nothing much to do with the existing patterns of the city and the university. At Firstsite, the galleries feel like leftover spaces in a conversion of an existing building, which is quite an achievement when it is a single-storey new structure on an open plot. These designs typify what another architect calls "the view from 30,000ft", by which glamorous gesture triumphs over detail.

This article might seem unduly personal, about an architect who is very far from being the worst. The reason for writing it is not any animus against Viñoly, but for what his rise reveals about the culture of recent British architecture. There has been too much faith in the idea of "iconic" or "world-class" architecture and too much fascination with the big name that will excite funders into giving money, or planners into granting planning permission and too little with the things, such as vertical walls in a gallery, that actually make spaces successful. The result is Firstsite, whose slightly grating name now makes sense: great at first sight, but then less so.