A couple of days ago, Seagate Technology PLC (NASDAQ:STX) reported its financial results for its fiscal fourth quarter and year-end 2011. Seagate is one of the world’s leading manufacturers of computer hard drives, and according to the company, in its last fiscal year it shipped over 199 million units worldwide. These shipments generated $11 billion in revenues and net income of $511 million, with a gross margin of 19.6%.
While these numbers are pretty impressive, they would have been otherwise unremarkable to the denizens of the upstream rare-earths sector, if it weren’t for some comments made by Seagate’s CEO, Stephen Luczo, on a conference call with analysts this past Wednesday. According to Barron’s, Mr. Luczo said that
The cost of many upstream materials, especially rare earth elements, which have increased significantly, these cost are expected to adversely impact gross margins by at least 200 basis points. In regards to the increase in cost of upstream materials, Seagate has historically been able to observe these cost increases and insulate our customers. However, the recent increase in the cost of rare-earth elements, combined with the pre-existing upward trend of other commodities, far exceeds our ability to offset cost reductions.
This is just about the first time that I’ve seen a prominent end user within the technology supply chain, actually publicly quantify the effects that increased rare-earth prices have had on their margins. Some commentators have claimed that rising dysprosium (Dy) prices have caused these loss of margins, but this is not so (though not a million miles off the mark). Let’s take a look at what IS the likely root cause here.
First, let’s ask ourselves a question: just how, exactly, are rare earths used in computer hard drives? The answer is that they are found in two sub-assemblies that contain rare-earth permanent magnets (REPMs), which are critical to the functionality of the hard drive – specifically:
- The spindle motor – this is the sub-assembly that causes the platens within the hard drive, on which the data is stored, to spin at high speeds to enable access;
- The voice-coil actuator – this is another sub-assembly within the hard drive, that moves the read / write head(s) to the correct position.
Both motors and actuators operate on the principle of converting electrical energy into motion, through the manipulation of magnetic fields associated with coils and with permanent magnets. In the case of the spindle motor, the motion is rotary; with the voice-coil actuator, the motion is from side to side. Both sub-assemblies are controlled by sophisticated electronics, which coordinate all movement of sub-assemblies within the hard drive.
As I said, each sub-assembly contains REPMs – however, the magnets in each sub-assembly are distinctly different from each other. Those in the spindle motor are usually so-called polymer-bonded permanent magnets, based on an isotropic (the properties are the same in all directions) compound that principally contains neodymium, iron and boron (the so-called Nd-Fe-B composition).
Compression-molded, polymer-bonded Nd-Fe-B permanent magnets, used in spindle motors. Source: TDK.
These polymer-bonded magnets are made by mixing the Nd-Fe-B powder with a special binder, which allows the magnets to be pressed easily and precisely to a final shape, with little-to-no additional machining required. More specifically, the typical polymer-bonded magnet used in spindle motors is made by magnet manufacturers using powders produced by the Magnequench division of Neo Material Technologies Inc. (TSX:NEM). Magnequench produces a range of different powders for a variety of applications; variations on its MQP powder family, some with additions of cobalt (Co), and occasionally niobium, will be used for these magnets. For those who like to get into the technical nitty gritty, typical magnetic properties of these materials are Br = 6.5 – 7.5 kG (650 – 740 mT); Hci = 5 – 9 kOe (415 – 715 kAm-1;) and BHmax = 8.5 – 11 MGOe (70 – 90 kJm-3).
You might have noticed that I have yet to mention Dy here. Virtually every man and his dog in the rare-earths sector will tell you that all Nd-Fe-B magnets need Dy as an additive to make them work at higher temperatures. And I’m here to tell you that every man and his dog would be wrong.
Dy is not used in the aforementioned magnet materials found in hard-drive spindle-motor magnets. Instead, improvements in performance of these materials at higher temperatures, come from increasing the Curie temperature of the materials by those additions of Co, which substitutes for some of the Fe present. The Curie temperature is the point at which the magnet will lose all of its magnetism. And since we’re dispelling half-truths here, let’s dispense with another one; again, contrary to popular belief, Dy does NOT raise the Curie temperature of the magnets to which it is added (at least not to any significant degree). Instead, Dy improves the higher temperature characteristics of a magnet, by producing a compound that is more resistant to being demagnetized. A byproduct of this is an increased ability to resist the effects of increased temperature.
The downside to adding Dy to a REPM, however, is that while it increases the resistance to demagnetization, it actually decreases the magnetic output of the magnet, or the so-called residual induction of the magnet. So there is a trade-off in the use of Dy. This is why the second set of magnets in the computer hard drive, those to be found in the voice-coil actuator, generally contain no Dy either. The typical computer hard drive does not use any components that contain heavy rare earths, such as Dy.
Voice-coil actuator magnets (the curved components) - the scale is in cm. Source: TDK.
Because the voice-coil actuator sees very little temperature variation beyond the ambient room temperature, there is no need to add Dy to the permanent magnets used in this sub-assembly. Unlike the spindle motor, the voice-coil actuator uses anisotropic (has a preferred intrinsic orientation) sintered Nd-Fe-B permanent magnets, i.e. they are produced without a binder. The magnet material grade used for these magnets is selected to maximize the residual induction of the magnet, which in turn gives greater sensitivity to the overall functionality of the voice-coil actuator, i.e. to move the read/write heads to precisely the right place at the right time. Again, for the techies, the typical magnetic properties of these sintered materials are Br = 14.2 kG (1420 mT); Hci = 14 kOe (1120 kAm-1); and BHmax = 50 MGOe (400 kJm-3).
In some material grades we may see a small amount of the Nd substituted with praseodymium (Pr). This is primarily to allow for the use of didymium (a mixture of Nd and Pr) in the magnet manufacturing process, reducing the price a little. However, anything more than 20-25% substitution and the performance of the Nd-Fe-B-based magnet will degrade – Pr can NOT be used as a full substitute for Nd, for this reason.
Since the escalating price of Dy has nothing to do with the reduction on gross margins for Seagate, and by implication other hard-disk drive manufacturers (since these hard drives don’t contain any Dy), what is causing this effect? The answer of course, lies in the rising costs of the other rare earths present – i.e. Nd, and possibly Pr. More specifically, however, the timing of Seagate’s remarks would perhaps indicate that they are procuring most of their magnets (at least the sintered magnets for the voice-coil actuators) from Chinese magnet manufacturers.
How do we know this? Take a look at the following charts, showing the spot prices of oxides of Pr and Nd over the past 13 months, for both exported and non-exported materials originating in China:
Average spot prices for praseodymium oxide over the past 13 months.Source: Technology Metals Research, metal-pages.com.
Average spot prices for neodymium oxide over the past 13 months.Source: Technology Metals Research, metal-pages.com.
The usual tales of woe concerning price increases for rare earths, center on the export prices for these materials. As we can see from these two charts, the imposition of ever-increasing export quota surcharges by the Chinese traders, caused a significant divergence of pricing between internal and export product. The less-reported detail to this, however, is that for these and other light rare earths, the price in China remained effectively flat until around January or February of 2011. At that point, the price of these materials started to increase, and soon even the Chinese REPM manufacturers began to feel the pain, as did their customers.
Such price increases would have manifested in the associated downstream products after some period of delay. If Seagate has not previously reported challenges with their margins, caused by rare-earth prices, then it is reasonable to assume that they have only recently started to feel the effects, which would strongly indicate that they are sourcing their sintered magnets from China. The cost of the magnet powders for the polymer-bonded spindle-motor magnets has also increased, but it is less clear from where Seagate buys the magnets made from these powders, since Magnequench powder is used by a variety of magnet manufacturers around the world.
In case you’re interested, even though there is pretty much no Dy in computer hard drives, here is a chart showing the recent prices for Dy oxide:
Average spot prices for dysprosium oxide over the past 13 months.Source: Technology Metals Research, metal-pages.com.
Here we can see that the effects of the quota surcharges, while still significant, have not caused as strong a decoupling of the prices associated with internal and exported product. Interestingly, for part of June, the internal price in China appeared to be higher than that of exported materials, though that has since changed.
Finally, it’s an interesting exercise to do a rough calculation to see the quantity of rare earths used in computer hard drives each year. If we focus on the sintered magnets alone, given the fact that such magnets typically weigh around 6 g, and there are two magnets per assembly, and given the fact that Seagate shipped 199 million units in their last fiscal year, that amounts to 199,000,000 x 6 x 2 = approximately 2,388 t of sintered Nd-Fe-B magnets last year. Those magnets would contain approximately 715 t of Nd metal, equivalent to the amount of Nd in 845 t of Nd2O3. Recent data-storage industry figures estimate Seagate’s share of the market to be somewhere around 28%, so a rough estimate of total rare-earth usage in sintered magnets for the hard-disk drive industry, would be over 3,000 t of Nd2O3 equivalent, each year. That would account for as much as 12-15% of the currently annual global supply of Nd2O3, a pretty significant fraction – and that’s not even taking into account the Nd content of the spindle motors!
The hard-disk drive industry is therefore an important end-use market for rare-earth materials, and the components which contain them. Both ends of the overall supply chain would do well to keep a closer eye on each other…
Update ( 07/23/11):
As part of a response to a comment below, here is the price chart for samarium oxide:
Average spot prices for samarium oxide over the past 13 months.Source: Technology Metals Research, metal-pages.com.
I recently bought a solid state flash drive to replace my hard disk and love the speed. Since it’s a better technology that seems likely to continue a downward price trend, I’d be fascinated to know whether flash drive will increase or reduce the REE dependency of the computer industry.
Gareth, and excellent informative article, thanks. To round it out, do you have any idea of the magnets input cost percentage, relative to the production cost of the hard drive? Perhaps this is trade secret of each manufacturer. Presumably if Scandia’s new brushless motor heat sinks for CPUs works out, there will be even more NdFeB needed.
How cost-sensitive is demand for various rare earths? Given that for many products there are no easy substitutes would that make demand relatively inelastic? And what percentage of the total cost of a product is the cost of the rare earths used in it? (for various key products) And for example also for the Seagate example you provide. If the cost of all rare earth were to suddenly double what would happen to demand? If the cost of gasoline at the pump were to double demand obviously would drop considerably because gasoline makes up a significant portion of a household budget. If rare earths only make up a small percentage of the cost of a product (say 1%) presumably if this cost rose to 2% it might not affect the overall cost of the product that much nor demand for its rare earths? Any light you can shed on the above relationships I think would be interesting and useful. Thanks and regards
Very nice article and calculation, to add the bottom line though, we can also say that 845tonnes/200 million PCs at $20/kg (approximate from your chart) for Nd203 costed $.08 per PC, and at $300/kg it’s $1.2.
This is a huge increase, especially with other magnets and other rare earth prices increasing, however for a PC, even 10-20 bucks increase may seem very managable looking at end user prices (although from production margin point of view, it is a hit).
Great article, good to know.
Only one comment: there are 4 magnets used in each arm motor, two on both sides of the coil that is shaped like a piece of pie.
The 2 on one side of the coil are magnetized up, the two on the other side magnetised down and the (magnets-) supporting steel plates guide the flux from the up-side to the down-side and circulating back to th up side by the lower piece of steel. Magnetic flux has to follow closed lines and the more iron is guiding the flux the more flux density will be in the air-gap.
Flux-density in the air-gap times total current (n windings times current per winding) times wire length inside magnetic flux area is giving force.
RHABE
I would like to know about rare earths in flash drives, too, like John above, especially since Seagate went down at the end of the week while Sandisk, a flash manufacturer, went up — both dramatically.
The price of Nd is the issue now. It adds to the cost of products. Will availability become the issue? If Nd is in short supply then production is curtailed. In other words, a shortage of material limits growth. While a disk drive requires only a small amount of Nd; nevertheless, it is a key material which is indispensable. Nd is also a key material in the production of a car. For example, a Prius requires 1 kg. Will supply and demand determine whilch product is produced–a disk drive or a car?
@John Petersen & @Bob M: Flash drives are clearly a significant factor in the future of hard-disk drives (HDD). I’ve read reports though, that the rise in cloud computing is giving the HDD industry new impetus in its fight against flash, since the use of off-site systems in this manner negate the “local” advantages that flash can provide.
@Graeme Irvine: Eamon Keane did a nice calculation here, though he’s assumed that there is a little Dy in the magnet (not an unreasonable assumption, but my take is that most of these magnets don’t use Dy).
@max iocono: for complex, engineered end products, the percentage of the overall cost associated with rare earths is relatively small, even with the dramatic price increases of late. The HDD industry is one of the most cut-throat businesses on the planet, in terms of squeezing every last penny. It is likely that in this case, Seagate has absorbed the costs of increased rare-earth prices, but they won’t be the only ones facing this issue in the HDD industry. Some price increases may be masked by the introduction of new products with higher areal densities for data storage.
@daniel: agreed – a reflection of the complexity of the end-use product.
@Ruediger Haberland: indeed – I have seen 4-magnet designs; I’ve also see 1-magnet designs. In each case the magnetic circuit is generally the same as you’ve described it; with the 2-magnet configuration the magnets are magnetized so that they have two poles on each face, instead of a simple straight-through magnetization. Using 4 magnets makes that a bit easier, but I am guessing that the result is around the same in terms of mass. I’m also presuming that physically small hard drives will use physically smaller voice-coil actuators and thus magnets, too, but that 13 g / voice-coil actuator is a good rough estimate.
@D. Carlton Rossi: it’s a good question; I suspect though, that the HDD guys have long-standing arrangements with the magnet manufacturers with whom they do business. I am not sure that there will be a “guns or butter” situation though, at least not in the high-tech applications like HDDs, because to a point, the buyers in those cases can afford to “outbid” the users of materials for lower-tech applications.
All of this discussion focuses on NdFeB. Who makes these devices with SmCo which is another super-magnet with higher Temp capability, and is Mfg. in Pennsylvania USA.
How About SmCo ?
@Philip A. Studer: I presume you’re referring to Electron Energy Corp? They make excellent samarium-based (Sm-Co) magnets. Early HDDs did use Sm-Co magnets, before Nd-Fe-B came along; but at room temperature, the Nd-Fe-B magnets have significantly higher residual induction (which is more important than increased coercivity, or resistance to demagnetization), put out more magnetic field, and are thus superior to Sm-Co magnets. They’re also still cheaper, and for applications that don’t need Dy, will remain so. I’ll add a price chart for samarium oxide to the article above: the relative price increases for Sm have been very significant too.
Garett, while cloud computing may represent another alternative to Hard disk drives within computing devices in addition to flash drives, the question raised by John Petersen, above, which I share, has still not been answered. It’s my understanding that flash drives use even more rare earth elements than hard disk drives, and the more expensive ones at that, but I’m not entirely sure. Can you clarify this?
@Mike Holt: rare-earth elements are NOT used in flash memory or solid-state drive systems. Flash is based on electrostatics, not magnetics. Polycrystalline silicon is the primary material used, with silicon nitride, silicon dioxide and related compounds featuring in variations of the main technology.
@ John Petersen
There’s a good report from IBM comparing solid state drives (SSD) and hard disk drives (HDD) here (www-03.ibm.com/systems/resources/ssd_ibmi.pdf). The summary is:
“While the cost of SSDs is trending downward, the $/GB for SSDs is still substantially higher than that of HDDs. Thus, it is not yet cost-effective for most applications to replace all HDDs with SSDs. Fortunately, it is often not necessary to replace all HDDs with SSDs. For instance, infrequently accessed (cold) data can reside on lower cost HDDs while frequently accessed (hot) data can be moved to SSDs for maximum performance. Many applications have a high percentage of cold data compared to hot data allowing SSD usage to be leveraged very effectively. The appropriate mix of SSDs and HDDs is then used to strike a proper balance between performance and cost.”
This is also what Dell recommend in a May 2011 white paper.
dell.com/downloads/global/products/pvaul/en/ssd_vs_hdd_price_and_performance_study.pdf
SSDs are currently 13 times more expensive than HDD for “sequential access applications” e.g. media streaming, while SSDs are half the price of HDDs for applications that use “random data access and small payload sizes” e.g. email.
There’s a further good presentation here:
snia.org/sites/default/education/tutorials/2011/spring/solid/AnilVasudeva_Are_SSDs_Ready_%20Enterprise_Storage_Systemsv4.pdf
A good stat is that 65% of the I/O data requests originate from just 2% of corporate data.
The total cost of ownership (TCO) of SSDs improves if the side benefits of increased reliability & worker productivity and reduced maintenance & power consumption are included. The retail $/GB cost of SSDs is still around 10 times the cost of HDDs, although there are indications that HDDs may be hitting the “commodity brick wall”, while SSDs continue their Moore’s Law descent. Were this trend to continue a crossover would occur which might be an event horizon where HDDs rapidly lose market share. On the other hand some forecasts still show HDD costs declining exponentially, and another forecast showing PCs with SSD rising from 5m in 2010 to 65m in 2014, or to around 10% of the market.
The data universe is exploding, with seers expecting it to go from 1 zettabyte in 2011 to 35 zettabytes in 2020, so there should be plenty of demand for both types of storage.
@ Gareth Hatch
Great informative article. Thanks for the reference, I’ll concede to your point on dysprosium :-) !
Good idea on working out bottom up Nd demand, but the figure for 0.815 kt Nd2O3 demand from HDDs. The recent USGS report (based on 2008 estimates) gives as 24 kt the amount of Nd (18) and Pr (6) used in NdFeB magnets. pubs.usgs.gov/sir/2011/5094/pdf/sir2011-5094.pdf
Separately slide 45 of Christian Hocquard’s presentation last year gave HDDs a 30% share of the NdFeB market. (pdf direct download): ifri.org/downloads/comptes_rendu/fichiers/51/hocquard.pdf
This would suggest demand of 7.2 kt of Nd + Pr demand (24*0.3). Assuming the figures are right for your home PC market, perhaps the much more significant demand comes from the corporate/server market, which is where most information is stored. Indeed the current approach of server farms to increase the speed of data retrieval from HDDs is to use short-stroking (storing data on the outside tracks of the HDD to limit the distance the hard disk drive has to travel). This will use more NdFeB per GB stored and will tend to be reduced by hybrid data storage which can use SSD for the most frequently accessed data and use “full-stroking” for cold data.
@Eamon Keane: excellent data points, as always :-) I’ve read estimates of anything from 22-25 kt Nd + Pr required for Nd-Fe-B-based magnets, so the USGS numbers sound about right. Note that my rough estimate of 3,000 t Nd2O3 did not include the use of Nd + Pr in the spindle motor magnets. To be honest I’m not sure how much they weigh, but if they weighed the same as a 6 g sintered magnet, then we’re at 6,000 t Nd2O3 (actually (Nd,Pr)2O3) required, which would put us at around 25% of total usage per the USGS report. That’s not far off the 30% estimate in the Hocquard PDF.
Speaking of which – what do you make of the tonnages in slides 45 and 46 of that PDF, for REOs in Nd-Fe-B magnets? They’re off, aren’t they?
@Gareth Hatch: in a desktop PC a Matsushita spindle motor magnet was 30/26×5 (all mm). So with density of 6 => 5,27 g. Assuming that spindle motors for PC’s do not differ sizewise by much, your 6 g estimation fits.
Nd used in wind turbines installed during 2010.
The major manufacturer of wind turbines with low-speed permanent magnet generators (LS-PMG) is Goldwind from China which installed 3.735 MW in 2010 (Source: Chinese Wind Energy Association, http://www.cwea.org.cn/upload/2010?????????.pdf). Assuming 600 kg of Nd/MW this results in 2.650 t of Nd oxide for Goldwind only.
Other significant LS-PMG wind turbine manufacturers include GE Energy (its 2.5MW machine mainly), Clipper, and licensees of the Vensys technology. I don’t have 2010 data but I roughly assume that these machines could add a 30% extra, i.e. 800 t Nd.
Other PMG used in the wind industry (e.g. by Siemens) are medium- and high-speed (i.e. from 100 to 1800 RPM), and require 1/10th of the magnet content per MW than LS-PMG require. Again a rough estimate is that in 2010 MS/HS-PMG could add 350 t of Nd demand.
The total 2010 Nd demand for the wind industry could be therefore 3.800 t.
I must correct the figures above: the 600 kg/MW is not Nd but magnet content.
Accordingly:
Goldwind – 2650 t of PM = 800 t of Nd content
Other LS-PMG manufacturers – 240 t of Nd content
Other PM, not LS – 110 t of Nd content
Total 2010 wind industry permanent magnet generators – 1250 t of Nd
I realised of this mistake as I reviewed Arnold Magnetic’s presentation mentioned by Eamon. There Mr Constantinides states that 95% of the Chinese wind turbine installations in 2010 were using PM generators. This is not correct as e.g. the only one of the top-4 manufacturers using PMG is Goldwind. Sinovel (AMSC design), Dongfang (REpower design) and United Power (Aerodyn design) use electromagnets.
@RLacal: Would you know if magnets used in LS-PMG’s are high temperature ones (dysprosium containing, some 5 to 7 wt.%)? If so, then LS generators would use some impressive 25% of annual Dy metal production.
@Boris
I cannot tell exactly the magnet grade used in LS-PMG, but the figures that we are using, from Shin Etsu and TMR, suggest 2- 3% Dy.
In wind turbines electricity generators can have a cooling system. I suppose that there is a trade-off here: magnets with higher Dy content are more expensive but dispose of the a cooling system resulting in system savings (also efficiency savings), but if the price of those magnets goes up strongly then the cooling system is cost-effective.
Roberto
@ Rlacal & Boris
The current issue of Renewable Energy Focus has a review of 2010 wind production (online.qmags.com/REW0511). Total installations last year were 37.64 GW of which offshore represented 1.16 GW. Half of all wind turbines were installed in China (18.9 GW) with the US in second place with 14.9% (5.6 GW) and much of the rest in European countries.
The technology trends section is dominated by permanent magnets, although Winergy trumpets its ‘HybridDrive’ tech which combines a gearbox and permanent magnets which only uses 20% of the REs that a full direct drive turbine uses.
In its Wind Energy Outlook section on the subject of direct drive the magazine states:
” Most turbines use a gearbox to increase the generator speed, but they are prone to failure and increase the mass of the turbine. A number of manufacturers have replaced gearboxes with a single, large-diameter generator which, combined with a converter, can be connected to the grid but this is not without its own challenges. Both the cost and the weight of this option currently exceed the more conventional gearbox designs. However, this approach is very promising, particularly if permanent magnets become cheaper and more powerful, lighter materials are introduced, and converters become more versatile. Direct drive options will become cost-competitive towards 2020, and are likely to become dominant.”
According to the REN 21 Renewables Status Report 2011, direct drive turbines had 18% of the market in 2011: (ren21.net/Portals/97/documents/GSR/REN21_GSR2011.pdf)
“Direct-drive turbine designs captured 18% of the global market, led by Enercon (Germany), Goldwind (China), and Hara XEMC (China). Preferred turbine sizes were 2.5MW in the U.K., 1.4 MW in China, and 1.2 MW in India. Globally, the average turbine size increased to 1.6 MW, up from 1.4 MW in 2007. Vestas launched the largest commercial turbine thus far, the dedicated offshore V164 7 MW turbine, targeting North Sea opportunities.”
Enercon (7% market share 2011) use RE-free electromagnets in their direct drive, it’s less clear what the breakdown of the remaining 11% of direct drive installations were, but a reasonable guess for 2011 RE containing installations might be 10%. If an average Nd2O3 content of 0.2 kt /GW is assumed (slide 48 of Christian Hocquard, adjusted from Nd content), although you will find higher estimates, then 2010 RE demand from wind turbines was 0.75 kt Nd2O3 (37.64*0.2*0.1). This would give wind about a 3% share of Nd + Pr demand. As RLacal states 2-3% seems a reasonable guess as Dy content, so that would equate to about 62.5 tonnes Dy or 4.2% of Dy output (1.5 kt in 2010).
Great discussion here. Does anyone have any predictions of other products – either within the consumer electronics industry or outside – which are next to see significant margin headwinds due to rising RE prices?
@ Gareth
I confused the 0.8 kt figure with your 3 kt figure, yeah the figures get very close to the USGS figures including the spindle magnets. The NdFeB figures on slides 45 and 46 are indeed off, the figures on slides 38-40 however are v. close to the USGS though.
Comments on this entry are closed.