The splintering of the global aluminium price is becoming ever more acute. Physical premiums on both sides of the Atlantic have just sailed through the $500 per tonne level, widening the disconnect between the global reference price set on the London Metal Exchange (LME) and the "all-in" price paid by manufacturers.

There is accumulating evidence that first-stage aluminium users, who have up to now taken the unhedgeable premium pain, are looking to pass it down the supply chain to their own customers.

If they do, what was once considered a temporary aberration risks becoming embedded throughout the aluminium industry.

The LME is racing to catch up with its US competitor CME Group in offering premium contracts, a potentially valuable hedging tool but one that may act to cement the fracturing of the previous pricing model.

And all the time the premiums keep rising. Why? And what will halt this seemingly unstoppable premium machine?

IN THE BEGINNING THERE WAS DETROIT

Aluminium manufacturers have blamed the LME's warehousing network for causing premiums to rise. The exchange has in turn pointed to the extraordinary combination of financial crash in 2008-2009 and the subsequent period of negative real interest rates.

What's not in doubt is that huge amounts of unwanted metal flowed into LME warehouses, particularly in Detroit, at the end of the last decade as financial crisis became manufacturing crisis.

That transformed the LME forward aluminium curve into "super contango", attracting the attention of stocks financiers, who realised they could earn a return by buying spot and selling forward. "Free" QE money both facilitated the trade and, by obliterating any positive return in fixed-income markets, incentivised it.

The main cost of what LME veterans still call the cash-and-carry business is storage. LME rent rates are much higher than off-exchange costs, so it was no surprise that there was a collective rush to move metal out of LME sheds.

A delivery system designed to handle industrial-scale flows couldn't cope with the load-out requirements of financiers, who were dealing in the hundreds of thousands of tonnes.

Metro, the LME warehousing operator that found itself in the eye of this aluminium storm in Detroit, decided to capitalise on its good fortune of being in the right place at the right time by using the rental income from its load-out queue to bid for more metal.

The more metal that flowed in, the more metal that was hoovered up by financiers, who then cancelled it with a view to moving it to cheaper sheds elsewhere in the city. The load-out queue grew and grew, a virtuous circle for Metro, a vicious circle for anyone else trying to buy physical aluminium.

It was inevitable that such a great money-making scheme would attract the attention of bigger players. Goldman Sachs snapped up Metro in 2010 and finessed the revolving-door warehousing strategy. Glencore, one of the power players in the aluminium market, bought its own LME warehousing operator, Pacorini, and rapidly built its own load-out queue at the Dutch port of Vlissingen.

Whatever the disputed drivers of the original disconnect between LME basis price and premium, the linkage between lengthening queues and rising premiums was indisputable.

It seemed logical then that if the LME could reduce the queues, premiums would fall.

Logical but, with hindsight, wrong.

THE INVISIBLE FLOOR

The LME's new load-in-load-out formula has brought to an end the revolving-door warehouse model. Even though it only comes into effect next February, both Metro and Pacorini have preemptively changed their behaviour.

The load-out queue at Vlissingen has been trending lower, now at 637 calendar days, compared with 748 at the end of April. That at Detroit has been capped at around 700 days in recent months and should now start falling, given there is hardly any metal left there to cancel.

Premiums, it follows, should have topped out, if load-out queues were the primary driver.
But they haven't.

The premium for Midwest delivery as assessed by Platts, a leading global energy, metals and petrochemicals information provider, currently stands at 23.5 cents per lb, equivalent to $518 per tonne. The cost of getting metal out of Detroit, factoring in the length of queue, the corresponding rent and the load-out charge, is $382 per tonne.

The physical premium, in other words, has itself disconnected from the queue "premium".

That doesn't mean that the queues are not still in the mix. It's just that they have changed from primary driver of rising premiums to invisible floor.

There is still a "hard" arbitrage between getting metal out of the LME system and the physical market, even if it is not currently profitable.

PHYSICAL DRIVERS

But, quite evidently, other factors are now driving physical premiums over and above anything that could be explained by the length of LME load-out queues.

And, in many ways, they are the traditional drivers of physical premiums, reflecting both the cost of delivering metal to a manufacturer's yard and the nuances of regional supply-demand dynamics.

Right now, for example, at a very local level trucking rates are rising in the US, increasing the costs of delivery.

At a regional level, the US and Europe are evolving into ever deeper deficit markets, reflecting the closure of smelter capacity over the last few years.

Annualised primary aluminium production in North America has fallen from 5.9 million tonnes at the end of 2007 to a current 4.5 million tonnes.

Consumption is booming and the US is sucking in ever greater quantities of imported metal. And, because traditional sources of supply such as South America have also seen smelter capacity slashed, those imports are coming from ever further afield.

Imports from both Russia and the Middle East, in particular, are rising but the freight is significantly higher, directly impacting the cost of physical delivery.

Physical premiums, in other words, are reflecting what they are supposed to reflect, just off a higher base.

That's one half of the pricing disconnect.

The other is the LME basis price, which remains low at least in part due to the hedge short position associated with financing all the stock overhang, both that still in the LME system and that now being stored outside it.

The stocks financing trade is still profitable, even if the returns are much diminished from the days of "super contango".

Some 1.77 million tonnes of aluminium have left the LME system this year and most of it hasn't gone anywhere near an industrial buyer. After all, that was why the metal was queuing in the first place, to be shifted to lower-cost storage which would make financing it more profitable.

Physical premiums are also about pricing accessibility of metal. And while they have risen because of rising import and transportation costs, the simple fact is that much of the metal in the US and Europe is still locked down in financing deals.

It is no more accessible to a spot physical buyer than when it was sitting in an LME load-out queue.

UNSTOPPABLE?

So what can stop premiums rising further and will aluminium pricing revert to its previous simple state or forever be fractured into two parts?

The currently active combination of physical market drivers is unlikely to change any time soon, given the tectonic migration of aluminium production away from the US and Europe.

The LME queues will fall and in theory the floor price of premiums will fall with them. But this is going to be a long, long process and one that will remain highly sensitive to warehouse behaviour. There remains plenty of wiggle-room in the LME's new rules for queues to be stabilised at a high level.

Any normalisation of interest rates, the bedrock of the stocks financing trade, is also going to be a long, long process.

The US Federal Reserve may be preparing to start lifting rates, but it has signalled it will be a gradual incremental process. Central banks in both Europe and Japan are moving in the opposite direction, suggesting that "free money" in one form or another is going to be with us for a good time yet.

Changes in the shape of the LME forward curve may pose a more immediate challenge to the stocks financiers, particularly given an increased tendency of the front part of the curve to flip from contango into backwardation.

But that only affects new financing deals not existing ones, where the profit has already been realised on paper.

Volatility in the premium itself may cause an unwind of financing deals, given it has become the main unknown in the trade.

But an unknown that is already hedgeable on the CME and will from the second quarter of next year, be hedgeable on the LME.

It would be ironic if a measure intended to serve the interests of industrial users actually serves to cement the disconnect that has caused them so much anguish in the first place.

Maybe, it will take a combination of all these factor to stop the premium machine. But it doesn't look as if it's going to happen tomorrow. Or the day after that.

And in the interim the market will keep evolving.

The history of the premium itself is one of evolution. Metro had no idea when it first came up with the idea of using rental from its load-out queue to lure more metal into its sheds that it would trigger such a fundamental split in the aluminium industry's pricing model.

But that's what's happened.

And as high premiums have become the new normal, players are evolving their own ways of adapting to that reality, whether it be passing through the premium to end-users or learning how to hedge it.

The premiums can't keep rising forever. But that doesn't mean to say that there will be a return to how things were before. Every day that passes makes that more unlikely. - Reuters

* The opinions expressed here are those of the author, a columnist for Reuters.