How is entering the GPU industry too expensive?

[Anarchist420]

I don't think it could be, if it's done efficiently and if there are no laws that bar entry. A lot of people say it is expensive, but isn't overhead extremely high? Also, how much is IP a barrier to entry?

When I say overhead, I mean how much these companies pay their CEOs and management. Also, nvidia is in CA, which either means that CA laws are making their overhead high or they're benefitting from CA's pro-business regulations. Whole Foods is successful and they don't pay their management a whole lot more than their workers. I also think a lot of the technology sector is subsidized by law, due to IP and handouts like DoD contracts. This is not about debating whether their subsidization is just, but I'm pointing out that that's a sign that nvidia and ATi may not be very efficient and how they may be benefiting from barriers to entry.

 

[SirPauly]

I think it's the engineering prowess and patents that separates things. In the early days, there were exactly that -- lots of players entering the fray but the ones with sound management but more importantly engineering prowess and not surprising that it was nVidia and ATI/AMD that survived the onslaught of competition.

I think the answer is simple: Because it is very hard. Even Intel has had problems entering the discrete GPU arena.

 

[DominionSeraph]

wat

 

[Red Hawk]

In order to be competitive with the two current gaming GPU makers you need to hire a whole lot of hardware engineers and software engineers to make a successful GPU chip. And not just a single GPU chip, but a series of GPUs. You also need to pay a fabrication company like TSMC to make your chips. You need to partner with other companies that will manufacture boards around the chips. All that is an incredibly risky investment when you really aren't sure if it will pay off.

Just look at Intel, titan of the computer hardware industry. At least twice it's tried to break into the GPU industry, and failed both times. The first time the GPU it released failed miserably compared to its competition. The second time its GPU was so bad they didn't even bother releasing it. If Intel of all companies can't break into the GPU industry, I dare say no one can.

 

[amenx]

I think what makes it especially tough, esp with players like Nvidia and ATI, is that the product cycles are so rapid, technology and engineering have to be really quick to come up with cutting edge products at such short notice. To keep up with Nvidia and ATi would be a huge undertaking if starting from scratch.

 

[lehtv]

It's a huge risk to enter the market, that's why. It's a gamble. First you need a ton of money and time to develop a product that can potentially compete with AMD and NVIDIA in any market segment. And you need some way to make the product sell, you need name and a brand and good marketing.

If you're a big business like Intel, you already have the money and the brand. But you don't have any guarantee that it will pay off (read below). If you don't have a brand, you need to do one of two things: 1. create a strong brand in a different, more approachable market, and move into the graphics business much later (and whether it will be a good idea then is very hard to foresee) - or - 2. convince some existing strong brand to cooperate in a brand+developer relationship.

Finally, even if you do get a fully finalized competitive product into the market, there's just no way to know if it will sell before it's there. NVIDIA and AMD just have such a good foothold. You'd need to come up with something seriously original that neither NVIDIA or AMD can offer. And if it will sell, it may not sell enough to even cover the costs of developing and marketing that product, nevermind profiting from it enough to finance the development of a second generation.

Huge gamble. Any big business is more likely better off concenctrating on competing in their established area, and any new business will have lots of trouble financing the thing.

 

[DaveSimmons]

^ and as mentioned above, patents. Once you do all of that, AMD and nvidia use patent infringement lawsuits to block you from selling your card at all and/or make you give them all of your profits as licensing fees.

Microsoft is collecting $15 per phone for some Android phones, and Apple has sued Samsung to keep their Android tablet from being sold in the EU.

 

[PowerYoga]

Did you really just make an analogy of managing whole foods vs managing the R&D and product release cycles of graphics cards?

 

[Idontcare]

OP, if I gave you $6B today, told you to spend it all over the course of the next 4 yrs by creating a company from the ground-up, hiring the thousands of engineers needed to design a competitive GPU (competitive with what the competition will be fielding in 4 yrs, not good enough to be competitive with what they are fielding today), buying licenses to the IP necessary to support DX9/10/11/12 and so on, building a support team, software driver team, AIB relations and support, etc...and let's say by some miracle you actually did it, 4yrs from now you have a chip that is competitive...what is going to be the ROR (rate of return) on my $6B investment into your business venture?

Have you seen the profit margins for AMD and Nvidia? After 4yrs my $6B would be gone, paid out in salaries, depreciated equipment purchases, and licensing fees, and it would take you 15 yrs to simply recoup my initial investment (if profits are a consistent $100m per quarter) let alone to actually generate a positive return on my money.

And what would be my ROR had I simply invested that $6B into purchasing Nvidia stock, or Intel stock (no one in their right mind would purchase AMD stock, so let's not go there )...or investing the money in many other "safer bets" that can be made?

$6B * 19 yrs @ 3% = a veritable buttload more money than I'd get by starting a new GPU company.

The world isn't lacking investors with $6B to invest in such an endeavor, the problem is that the world has a plethora of more lucrative and lower-risk opportunities for investors with $6B to pursue.

The proof of the pudding is in the eating...do you see Warren Buffet investing in GPU businesses? or do you see him investing billions into banks? :hmm:

 

[Anarchist420]

OP, if I gave you $6B today, told you to spend it all over the course of the next 4 yrs by creating a company from the ground-up, hiring the thousands of engineers needed to design a competitive GPU (competitive with what the competition will be fielding in 4 yrs, not good enough to be competitive with what they are fielding today), buying licenses to the IP necessary to support DX9/10/11/12 and so on, building a support team, software driver team, AIB relations and support, etc...and let's say by some miracle you actually did it, 4yrs from now you have a chip that is competitive...what is going to be the ROR (rate of return) on my $6B investment into your business venture?

Have you seen the profit margins for AMD and Nvidia? After 4yrs my $6B would be gone, paid out in salaries, depreciated equipment purchases, and licensing fees, and it would take you 15 yrs to simply recoup my initial investment (if profits are a consistent $100m per quarter) let alone to actually generate a positive return on my money.

And what would be my ROR had I simply invested that $6B into purchasing Nvidia stock, or Intel stock (no one in their right mind would purchase AMD stock, so let's not go there )...or investing the money in many other "safer bets" that can be made?

$6B * 19 yrs @ 3% = a veritable buttload more money than I'd get by starting a new GPU company.

The world isn't lacking investors with $6B to invest in such an endeavor, the problem is that the world has a plethora of more lucrative and lower-risk opportunities for investors with $6B to pursue.

The proof of the pudding is in the eating...do you see Warren Buffet investing in GPU businesses? or do you see him investing billions into banks? :hmm:

You have a point as I probably wouldn't invest in making GPUs myself if you gave me 6 billion I would probably invest most of it in more rare metals and save them (although that would help with investing in making a GPU as investing some of the credit in rare metals and being frugal with the cash is a good way to cover for depreciation of equipment).

However, if one starts out frugally (doesn't pay themselves or their employees as much as the competition which minimizes overhead), then they can build on that and pay themselves or their employees more later; the employees (engineers, support, etc.) will work for less initially if they think they're making a competitve product, because that gives them hopes that they'll make more as their product sells.

I don't know what nvidia's profit margins are like, but I have read that the CEO got paid like $20m 4 years ago. That's none of my business, and I definitely believe he's entitled to what he earns, but that makes me think that nvidia could possibly have more overhead than we know about.

Also, as you confirmed for me (more than I actually thought as I didn't know that microsoft charged people to make DX hardware), IP is a legislated barrier to entry (it may be just one barrier, but it's still a barrier).

 

[formulav8]

OP, if I gave you $6B today, told you to spend it all over the course of the next 4 yrs by creating a company from the ground-up, hiring the thousands of engineers needed to design a competitive GPU (competitive with what the competition will be fielding in 4 yrs, not good enough to be competitive with what they are fielding today), buying licenses to the IP necessary to support DX9/10/11/12 and so on, building a support team, software driver team, AIB relations and support, etc...and let's say by some miracle you actually did it, 4yrs from now you have a chip that is competitive...what is going to be the ROR (rate of return) on my $6B investment into your business venture?

Have you seen the profit margins for AMD and Nvidia? After 4yrs my $6B would be gone, paid out in salaries, depreciated equipment purchases, and licensing fees, and it would take you 15 yrs to simply recoup my initial investment (if profits are a consistent $100m per quarter) let alone to actually generate a positive return on my money.

And what would be my ROR had I simply invested that $6B into purchasing Nvidia stock, or Intel stock (no one in their right mind would purchase AMD stock, so let's not go there )...or investing the money in many other "safer bets" that can be made?

$6B * 19 yrs @ 3% = a veritable buttload more money than I'd get by starting a new GPU company.

The world isn't lacking investors with $6B to invest in such an endeavor, the problem is that the world has a plethora of more lucrative and lower-risk opportunities for investors with $6B to pursue.

The proof of the pudding is in the eating...do you see Warren Buffet investing in GPU businesses? or do you see him investing billions into banks? :hmm:

Definitely this :thumbsup: . R&D especially is majorly high dollar stuff

 

[uribag]

May I be the CEO?

 

[DaveSimmons]

IP is a legislated barrier to entry (it may be just one barrier, but it's still a barrier).

Some IP protection is necessary. Your company would have even less chance of succeeding if someone else could just copy your chip design and drivers to sell their cards for less.

No costs for R&D and ongoing driver updates would mean they could always sell for less than you while making more money. So no one would buy your cards.

 

[Capt Caveman]

Some IP protection is necessary. Your company would have even less chance of succeeding if someone else could just copy your chip design and drivers to sell their cards for less.

No costs for R&D and ongoing driver updates would mean they could always sell for less than you while making more money. So no one would buy your cards.

The OP believes there should be no patents or IP protection b/c it will allow companies to compete more cheaply and still be profitable even with copy cat companies ready to produce the product that a company may have spent billions to develop. You're not debating an issue with a sane person.

 

[Idontcare]

You have a point as I probably wouldn't invest in making GPUs myself if you gave me 6 billion I would probably invest most of it in more rare metals and save them (although that would help with investing in making a GPU as investing some of the credit in rare metals and being frugal with the cash is a good way to cover for depreciation of equipment).

However, if one starts out frugally (doesn't pay themselves or their employees as much as the competition which minimizes overhead), then they can build on that and pay themselves or their employees more later; the employees (engineers, support, etc.) will work for less initially if they think they're making a competitve product, because that gives them hopes that they'll make more as their product sells.

I don't know what nvidia's profit margins are like, but I have read that the CEO got paid like $20m 4 years ago. That's none of my business, and I definitely believe he's entitled to what he earns, but that makes me think that nvidia could possibly have more overhead than we know about.

Also, as you confirmed for me (more than I actually thought as I didn't know that microsoft charged people to make DX hardware), IP is a legislated barrier to entry (it may be just one barrier, but it's still a barrier).

I think you are overstating the relevance of the IP licensing angle and understating the relevance that your legions of employees will have little-to-no loyalty to work for you at pennies on the dollar.

You might be willing to work for peanuts, but try finding >1,000 engineers who are technically competent and capable of the job who will work for you for peanuts versus working for Nvidia or AMD for something considerably higher than peanuts.

And coming back to the reality aspect of it, there is a simple reason you don't see it being done. Transmeta didn't die out because of restrictive or financially burdensome licensing requirements, they went bankrupt for not having enough capital to keep paying their employees. If only those greedy louts would have taken a pay-cut and worked for minimum wage instead of expecting >$60k/yr

 

[TemjinGold]

You have a point as I probably wouldn't invest in making GPUs myself if you gave me 6 billion I would probably invest most of it in more rare metals and save them (although that would help with investing in making a GPU as investing some of the credit in rare metals and being frugal with the cash is a good way to cover for depreciation of equipment).

However, if one starts out frugally (doesn't pay themselves or their employees as much as the competition which minimizes overhead), then they can build on that and pay themselves or their employees more later; the employees (engineers, support, etc.) will work for less initially if they think they're making a competitve product, because that gives them hopes that they'll make more as their product sells.

I don't know what nvidia's profit margins are like, but I have read that the CEO got paid like $20m 4 years ago. That's none of my business, and I definitely believe he's entitled to what he earns, but that makes me think that nvidia could possibly have more overhead than we know about.

Also, as you confirmed for me (more than I actually thought as I didn't know that microsoft charged people to make DX hardware), IP is a legislated barrier to entry (it may be just one barrier, but it's still a barrier).

If Joe Blow engineer was a tech god who could be making big bucks NOW working for Intel/AMD/nVidia, why the heck would he work for you for peanuts now and the HOPE of making big bucks later? The only people you will get by paying them cheaply are people not good enough to be hired by the big boys. How the heck do you plan on building a competitive product with second rate employees, no patents, and a shoestring budget?

Also, don't underestimate the power of good managers that command top dollar. You really do get what you pay for here and poor management will cost you way more than what good managers get paid.

 

[PingviN]

What IDC said: you'd be better off investing in AMD or Nvidia stocks rather than taking on a new venture with a new company starting from scratch. No sane investor would place billions of dollars in a ground-up operation when there are established alternatives with a much better chance of a good return on the investment.

Especially in R&D-intensive businesses. There will be several years before you'll actually see any form of income.

 

[krumme]

Seeing how things is done in high-tech, i would say Idontcare 4 years is an understatement

Not so much for the R&D on the architectual side, they might, if you get the best, and pay way over the profitable reasonable, get you the right design to be at least considered number 3, if you look on the design on paper.
But the more craftman side of the things, the process stuff, and the connection between production and architecture, is a mess to get so fast. Look have NV is strugling here (the vias and the pilot stuff) All your tool knowledge fx., can under no circumstances be build so fast. You will simply need more years.

And for the software side, you probably face very similar problems.

Have a look at Intel trying to catch up on the gfx side. Its not like they need cash, but they simply dont have the critical mass for major parts from idea to production.

Both GPU and CPU is textbook examples of high entry barriers. They are not only huge they are gigantic.

 

[Seero]

OP, if I gave you $6B today...

I believe the hardest part is solved already.

 

[Arkadrel]

OP, if I gave you $6B today, told you to spend it all over the course of the next 4 yrs by creating a company from the ground-up, hiring the thousands of engineers needed to design a competitive GPU (competitive with what the competition will be fielding in 4 yrs, not good enough to be competitive with what they are fielding today), buying licenses to the IP necessary to support DX9/10/11/12 and so on, building a support team, software driver team, AIB relations and support, etc...and let's say by some miracle you actually did it, 4yrs from now you have a chip that is competitive...what is going to be the ROR (rate of return) on my $6B investment into your business venture?

Have you seen the profit margins for AMD and Nvidia? After 4yrs my $6B would be gone, paid out in salaries, depreciated equipment purchases, and licensing fees, and it would take you 15 yrs to simply recoup my initial investment (if profits are a consistent $100m per quarter) let alone to actually generate a positive return on my money.

My gawd that is a beautifull post.

This is why people dont do it, 20years of hard work, that might not even be enough to re-coup the initial investment of the $6B (in all likely hood, you ll probably not just make 100mil profits fright from the get go).
People that've never designs a GPU before, dont get everything right first go.

A "generous" estimate is 20years, to actually earn a dime from it (pay off initial investment).

$6B or so (could be higher) is more than most want to gamble with
(and most people arnt willing to wait 20years for something to pay off).


The "problem" is the cost of "borrowing" $6B.
You pay intrest, that eats from your profits.

Suddenly, its not 20years, its ~30 before you earn back that $6B.
Unless you know someone who ll just give you $6B and not demand anything from it.

So you go into the GPU bussiness, you need to know, your doing a longterm investment of 30+ years.


***
Look at Intel, their swimming in $ and yet they've tried to enter into the GPU market.
They have had serious problems with it, and they arnt worried about quick returns, and they even have a advantage in process technologies that you likely wouldnt have.

IF they cant do it, what makes you think you can?

 

[Gryz]

Where did the number $6 billion come from ? Out of thin air ? I can't believe it would take that much money to build a videocard from scratch. IDC, can you explain what rough estimate you made ?

You don't get all your capital in one go. You begin with a much smaller starting capital. As the company grows and progresses, and as more money is required, you attract more investment. If your product (or design) is progressing well enough, that will be helpful in attracting more investment money.

One reason that experienced professionals might be willing to leave nVidia, AMD or Intel might be because they hate their job ? They might hate the company they work for, their boss or whatever. If you work long enough for the same company, things will happen that will make you wanting to leave.

The intellectual challenge ? Not everybody works only for money. If you are a very good professional, you might have made millions already in the last 1 or 2 decades. Money is not their main concern. What if a new company would have such an ambitious goal that you would love working on such a project ? If your mortgage is paid for, your kids in University, and you still have enough money in the bank until retirement, then why not join a startup ? Thousands of people have made that jump a decade ago. And even though many of those startups have failed, I'm sure some people had a lot of fun. At least it was a change from working in your boring secure predictable desk-job.

A very important detail is that you need the expertise to build a videocard. But how many people do you need ? If you would get 10 of the best senior GPU-developers, videocard-builders and driver software engineers, would that be enough ? And have teams of less experienced engineers around them ? Not everybody at id software is a John Carmack. You just need one to build an engine. Plus a bunch of people that work with Carmack, learn from Carmack, etc. Does the GPU-business have those type of "prima donnas" ? Would recruiting some of them be enough ?

You can buy employee loyalty with stock-options. That used to work great over a decade ago. Not so great anymore now. But still, joining a company pre-IPO could be the best thing you ever did in your life. It's a gamble. But some people will be willing to take the gamble. If the new company succeeds, you might be able to retire after a few years.

Last thing. People seem to believe that the initial investment needs to be gained back in profits, before you can call the company a success. That's not true. The initial investment needs to be made back by selling the company. Probably on the stock market. Stock market is a weird thing. If you can create the hype, that makes a huge difference. Do you really believe that Apple is worth $400B ?

Heck, you could start a company, spend $1B investment money. Then have only a half-finished product. And sell it for double the money to Intel.

Still I think it will be very hard, if not impossible, to start a new videocard company. But all the reasons I heard so far do not seem absolute.
And again, where did that $6B number come from ? Just curious.

 

[Arkadrel]

Where did the number $6 billion come from ? Out of thin air ? I can't believe it would take that much money to build a videocard from scratch. IDC, can you explain what rough estimate you made ?

Buying ATI.

You think $6B is high? why buy a company if its "easier" or "cheaper" to do it on your own?

The $6B number is probably a "low" estimate.

You want to get into GPU bussiness? I think its cheaper to just buy Nvidia or AMD <.<

 

[GoStumpy]

I think building a videocard (ASUS, MSI, Powercolor, Sparkle, EVGA) is a lot different than building a GPU (AMD/NVIDIA/Intel).

Which one did the OP intend?

 

[Arkadrel]

Which one did the OP intend?

Im assumeing he ment as competition to AMD/NVIDIA.

Not just slapping together some chips he "bought" off other companies and placed on a board.

 

[Gryz]

Buying ATI.
You think $6B is high? why buy a company if its "easier" or "cheaper" to do it on your own?

The fact that AMD paid $5.4B does not mean that it takes $6B to design and manufacture a new GPU and videocard. Prices that are being paid for companies are bullocks, imho. Can't say much about the gpu industry. But in network land, I know of many acquisitions that seemed completely bogus. But billions of dollars were sometimes paid. Even when a company already had a "product x", they would still buy another company that would make a similar product.

The $6B number is probably a "low" estimate.

Again, how did you come up with that number ? Not just because of the AMD/ATI deal, I hope.

I worked for a startup once. Over a decade ago. At the time, it was the technology-startup that had the highest amount of investment money ever. The goal was to build a router from scratch. Hardware and software. The hardware had 6 proprietary chips. Not ASICs, but full-blown chip-designs. The largest chip had more transitors on it than a Pentium3. I was told that each spin of each chip would cost $1M. We had about 150 engineers working on the hardware (chip design and the rest). Still, the burn-rate was way under $100M per year. The company produced a working product. The chips worked. It was the fastest router at the time. But in those days, right after the big stockmarket crash, all ISPs had frozen their hardware expenses. And the company ran out of money. Lots of energy wasted, lots of money lost. But an interesting ride. And it didn't cost $6B.