Does it actually cost manufacturers more to make a 256gb vs 128gb SSD?

Does it? What if it were HDD?

Just curious.

 

Yeah, it does cost more - even if it were a HDD.

The reason? Simple: as long as there is a demand (at any price...) then each point in the chain will want what is 'theirs'.

So, it costs manufacturers a little more in each step of the chain, and it costs us (much) more as end users.

 

The truth is, it probably only costs them something like £50 to make the 256GB models and then probably only a fractional increase for the higher end ones. It's that way with all electronics, depending on how many stages it goes through to make.

It's like with apple, they just charge that much because they know they can and they know people will still buy it. It's completely stupid and extortionate imo, but that's just how the world is.

 

Most of the cost of a contemporary SSD is in the raw materials. Actual assembly costs are minimal. The margins on completed/assembled SSDs are quite minimal.

For instance:

DRAMeXchange- Market research, price and trading of DRAM, NAND Flash, SSD and Memory Card

64 gigabits of NAND flash (ie: 8 x 8) or 8 gigabytes go for $9.92. To build a 128gb drive, you need ~$160 worth of NAND flash alone. Add in another $10-$20 for fabrication/test, and another $10-$20 for a chip, and another $10-$20 for distribution costs/distributor/dealer markup/warranty returns, and its pretty easy to see that a $220 128gb SSD really isn't making a lot of money for the people involved in building them.

Of course, to go to a 256mb configuration, you basically have to tack on another $160 worth of NAND flash. Plus, depending on the design, you might be forced into using more costly higher-density chips. If you look at the above website, and run a few calculations, you can see that there is an increased cost per gigabit for the 64-gigabit parts versus the 32-gigabit parts.

 

Not even close. Just the NAND flash itself can't be procured for anything near as cheap as £50. Just the NAND alone for a 256GB SSD costs £200, in wholesale quantities, from the Asian semiconductor fabs.

Apple has high costs in other places in their business, ie: they have to expend massive resources developing an operating system that not many use, plus they have to do much of their own hardware engineering. Plus, unlike some other companies in the computer business (ie: Intel), Apple at least tries to create shareholder value instead of giving their products away for free. Considering how important computers are to daily life these days, there's nothing extortionate about Apple trying to charge more for their products. Some people find more value in them compared to cheaper alternatives, some don't. That's the risk Apple takes as a business.

 

Citation, please......

 

Scroll up, I just gave you one, the DRAM exchange, and its listing of spot prices on various NAND flash parts that are the raw components of a SSD....

 

Oh yeah, I forgot that SSDs need to be 'overprovisioned' as well, ie: a 128GB SSD may very well have an extra 16-24gb worth of chips in it. So in light of that, the numbers given were actually quite conservative in terms of the cost of building one....

 

so... i think my actual question should be...

what is the difference in cost for an asian semiconductor fab to produce 128gb of NAND flash that fits into a hard drive box from raw materials versus producing 256gb of denser NAND flash that fits into the same box?

 

That's essentially a cost of labour, electricity, and capital question. There are some chemicals involved, and the cost of plastic, gold, and silicon, but these costs are relatively minor.

Semiconductor fabs tend to be $1-$10B factories, with insanely expensive equipment, highly paid engineers, and relatively quick obsolescence. Therefore, they only have a very limited window to recover their investment before they have to toss it all in the garbage and build a replacement factory.

You might want to take a look at the stock of a company such as Taiwan Semiconductor Manufacturing (TSMC, ie: TSM on the NYSE, or 2330.TW on the Taipei stock exchange), and see just how much of a return such a firm has provided to shareholders over the past decade.

TAIWAN SEMICON MAN TWD10 Stock Chart | 2330.TW Interactive Chart - Yahoo! Finance

As you can see, its an enormously cyclical business, and they don't exactly make a lot of money. TSMC isn't a big NAND producer, but they make all of Nvidia's and ATI's graphics chips (some of the highest margin products in the business!) -- so kind of gives you an idea of just how profitable semiconductor manufacturing is (or rather, isn't!). NAND is even more competitive/less profitable than making graphics chips....

 

Good analysis, pitz, except for profits--TSMC actually makes quite a bit of money, as its profit margin is very good (especially compared to its competitors, UMC, SMC). In fact, when working with typical semiconductor design firms (which submit the designs and work with TSMC to actually fabricate the chips), TSMC's margin is on par with or even higher than the design firm's. Obviously, TSMC does have high costs as well (particularly upfront investment costs), but at the same time, it has very high volume and quite a bit of capital. Source: father, who works as the Foundry Director of a semiconductor design firm directly interacting with TSMC.

The main raw material costs of producing an SSD are 1) the NAND memory, and 2) the controller. The former is a more significant percentage of the costs, particularly in higher-capacity SSDs. So yes, it does cost significantly more to manufacture higher-capacity SSDs. For which size SSD manufacturers have the highest profit margin, well, that's harder to say.

 

:sigh: so much for an analysis or understanding of how a manufacturing supply chain works.

Prices on the much-vaunted 'dram spot market' aren't all that reliable or representative of long-term manufacturing or sourcing prices.

I can get better prices than that from my local warehouse/jobber for a phone call. Imagine what a manufacturer can get if they commit to an order for 10 million pieces of X.

 

Sure, but at $220 for a 128gb SSD, its pretty easy to see, from the back-of-the-napkin analysis that I did, that there are not large profits associated with building SSDs. At least from the perspective of a company like Intel, Corsair, or OCZ, who buy chips from fabs, integrate, and ultimately market them. Samsung, of course, controls the entire value/supply chain for building their SSDs, but they have largely the same considerations inbetween business units of the company.

Those 'spot' prices are for significant contract quantities, and are not retail prices. NAND prices have fallen considerably in the past year, so it is likely that firms that signed to long-term supply contracts are actually paying more than spot, and may actually be losing money for every cheap SSD they deliver (because they're competing, obviously, against firms that can buy at spot and build SSDs -- SSDs not being very hard to build as Sandforce et al will basically sell their chips, reference boards, and firmware packages to anybody who wants to buy them -- and the Sandforce chips are just as good as any on the market, ie: there is little proprietary advantage that a SSD builder can achieve to push prices up!).

Sure, they may lock in a discount initially, but if prices fall rapidly, that discount can become quite a premium to spot and the buyer will have to write down their inventory in order to compete. The trend for firms like OCZ, Corsair, etc., who want to deliver low-cost products and not put their entire firm at risk by entering into long-term contracts for future delivery, is to minimize their contract exposure. This way, these players, at the margin, can shut down production of finished units if the spot price of the raw materials exceeds what they can earn on finished goods, minus incremental costs of production.

 

Committing to a huge order does NOT lock a price. Anyone who signs such a contract usually gets fired.

You've got the original delivery price, an series of reducing extended-production prices, and often further price reductions based on continuing orders/production efficiencies.

More than once I've placed orders for 18-36 month deliveries of an item at say $100/1000 initially which fell to $15/1000 at the 50% mark in the contract.

Additional incentives are often set up by the buyer and/or seller if the specific components are needed beyond the initial 18-36 month production contract. The easiest place to look at this is Intels extended availability matrix where they guarantee to have certain CPUs and other semis available for a min of 70 months and in some cases, up to 94 months.

Again, 'spot' prices for any volume commodity are cute and make headlines, but they have little relevance for production or BOM planners.

 

No, nand is very profitable - (the whole world is your customer).

When you have a >$1B factory and are in the processes of building another $10B factory (or two) in the budget; you're not in the poor house.

The prices, as I've stated before... are what they are because they can be what they are. If the demand (from us...) was not there at that price, it would be much cheaper (all of a sudden... ).

Charging what the market will bear - has no bearing with what the manufacturing costs actually are. Do you really think they're doing this (paraphrasing @pitz: 'scraping by'... ) from the goodness of their souls?

As long as there is only two or three nand manufacturers 'competing' then the prices for this commodity will stay unrealistically high. (Not even taking collusion, price fixing and other underhanded practices that is almost invariably happening as we speak between the handful of nand manufacturer who will do anything to protect their highly profitable interests).

 

I'm not privy to specific details, but the whole point of committing to a big order is to agree on a price that the fab can use for production planning/financing, and the purchaser can use in order to price their goods in the marketplace. Hopefully lowering risk for both sides of the transaction, if other factors are otherwise stable.

I disagree here; the spot market is quite important to many players in the SSD space who buy chips at spot or on very short term contracts (ie: near-spot contracts), and deliver those finished SSDs to market. These players, when spot prices are falling rapidly, are able to most quickly respond with inexpensive finished end-user products. When spot prices are rising rapidly, of course, those products simply dissappear in the marketplace until market prices of the finished end-user products have risen adequately to cover purchase of new raw materials at the new spot prices.

The cost of capital, and risk to the capital of a business, when entering into a long-term contract for supply, can be considerable, and these prices are passed onto customers. If Apple gets stuck in a bad contract, they have sufficiently high margins on their end-user goods that they can absorb it. If OCZ, Corsair, or some other low-capital player entered into such a contract and the prices fell or rose faster than predicted, they could very well go bankrupt.

 

Then TSMC's (and other fabs' ) stock charts should be vertical. They're not.

Actually, when they go out and build those factories, they have to borow money, and repay that money, with interest, throughout the lifetime of the factory. The profitability of the firm is wholly dependant on whether they can charge enough for the output such that the value of the output exceeds the cost of the capital (equity and debt) required to build and operate the factory.

The stock chart of TSMC, UMC, etc., is a reflection of how efficient these firms have been, in the marketplace, in being able to capture the economic surplus of investments in fabs. As you can see from TSMC, over the past decade, they've basically been treading water, although they have paid a modest annual dividend.

Look, I'm a scientist, I work on data. I also invest in TSMC (ie: own shares). And I see no evidence that there are huge profits in the industry. Quite the opposite, shipments and use of computers worldwide has risen dramatically in the past decade, yet the price of my stock in one of the leading producers of semiconductors worldwide barely rises.

Sounds like 'conspiracy theory' to me.

 

Conspiracy theory? Call it what you want. I'm not saying it is happening - just that I would be surprised if it wasn't.

With all due respect: If you invest in a company whose stock barely rises - not only are you a bad investor, but your 'scientist' background is questionable too.

I'll simply repeat myself for emphasis: do you really think they're doing this for the goodness of their souls?

The best companies are not the ones that show are the most profitable: they're the ones that profit the CEO's and other top management inside and outside the company. (Think: strategic partners).

 

Can't speak for the purer NAND manufacturers, but at TSMC, the level of salary disclosure in their annual reports is impeccable for management, and they definitely do not have the outrageous compensation packages that you see in, for example, US banking.

Go download the TSMC annual report and see for yourself. The level of transparency that TSMC has in its expenses and employee compensation is amazing and puts most US-based companies to shame.

I never denied that these companies aren't, over the long-term, profitable, but they're not highly profitable, or insanely profitable, and most certainly, they re-invest heavily into bringing people like us the latest technology as quickly and as inexpensively as possible.

Name me even one other industry that consistently doubles the performance of their products every 18 months, yet consistently asks for less and less $$$ for those products? Does your local gas station give you double the amount of gas they gave you 2 year ago and charge you half as much? Does your barber cut your hair twice as well for half as much? I could go on and on, but it borders on the ridiculous that anyone would claim that computer manufacturers have been anything but highly, if not cut-throat competitive over the past few decades.

 

Haha, that's what was going on when I was in the auto industry, except the "Big 3" would strong arm suppliers into costs for plastics for 5-10 years, and when the cost of oil skyrocketed, they couldn't make a profit. Dozens of smaller companies just folded and it left a big mess with the huge suppliers that offered more than mainly plastic components, and would more or less be shipping components with dollar bills attached to them. It was insane, and the Big 3 wouldn't budge for a long time.... but I digress.

Those costs though have to be wrong considering you can see the likes of Kingston and WD offering the 96GB SSD for < $100 and 256GB SSD for < $300 periodically. No not the fastest drives, but still use NAND. And other 120-128 GB SSD's are dropping in price to below $150.

 

Sometimes the prices get lowered to clear out inventory (at a loss). Its not a lot different than, for instance, airline tickets, which are often sold at less than the average cost of production, simply to move product instead of accepting even more severe losses. Or GM that kept building insane amounts of cars even though they were taking losses on each unit delivered.

The 96gb SSD may very well use parts that are in a sort of 'sweet spot' in terms of a price point. For instance, that link I provided to the "DRAM exchange" showed that, per gigabit, 32gbit chips were less than half as expensive as 64-gbit chips.

Also, firms that have bought chips on contract at below spot, can always turn around and sell them at spot. Rationally, they'd do this in preference to building a bunch of drives where they're not even achieving the spot price for their finished product. There appears to be a glut in capacity to turn NAND chips from raw chips, into finished SSDs, and the dirt cheap SSDs is showing this. Could be because the global economic downturn has severely dampened business purchases of laptops and SSDs generally.