Mercury Research on graphics market share

[Dave Baumann]

Seems that there is some doubt over how the "Performance" calculations were done.

Due to a somewhat arbitrary definition of 'performance' done according to transistor count, rather than actual performance, Mercury included Nvidia's 6800 (enthusiast), 6600 (performance) and 6200 (mainstream) products in the 'performance' category, while only including ATYT's X800 (enthusiast) and a few new X700's (performance) products and excluding ATYT's X600 and X300 products. As a result, the performance market size and Nvidia's share in it, was overstated. Based on our checks, we believe Nvidia's actual 6800 shipments were on the order of 275,000 (plus or minus 25,000) compared to around
382,500 X800's for ATYT.

 

[Bouncing Zabaglione Bros.]

Seems that there is some doubt over how the "Performance" calculations were done.

Oh dear. So it turns out the one good thing that Nvidia could point to (and has been doing so frantically) turns out to be nothing but spin and manipulation of the numbers. Why am I not surprised?

 

[Mark]

Which report is that from Dave?

 

[Mark]

I found a simialar document. This one is from a TD Waterhouse report following ATI's Analyst Day thing, so the quote Dave used was likely from an analyst at another firm who attended the same meeting...

Upbeat ATI Analyst Day; Demand Outpacing Supply

Event: ATI hosted its annual Analyst Day on October 28, with presentations from its CEO, CFO, the PC business unit, the integrated and mobile business unit, and the consumer business unit (wireless and digital TV).

Impact: Positive. We left the meeting with increased confidence that ATI is well positioned to extend itsnewly crowned desktop discrete graphics chip leadership, to maintain its mobile graphics chip leadership, and has significant green field opportunities in desktop integrated chipsets, workstations, wireless handsets, digital TV, and game consoles (ATI targets a combined addressable market of $9 billion per year). We are also encouraged by ATI’s comments that demand continues to outpace supply of its discrete chips and integrated chipsets. ATI reiterated its Q1 (November) guidance for $600 million to $640 million in revenue (we have no changes to our estimates for $629 million in revenue, up 10% sequentially, and $0.26 in adjusted EPS).

ATI is targeting F2005 revenue growth in all product lines (highlighted by greater than 2x growth for workstations, desktop integrated chipsets, wireless chips, and digital TV chips). Gross margins are expected to improve each quarter in F2005, beginning with guidance of 33% to 34% in Q1/F05, and exiting F2005 above 35% (our Q4/F05 gross margin estimate of 34.4% may prove conservative). F2005 operating expenses are expected to increase at approximately the same rate as revenue growth (we estimate operating expense growth of 26% compared to revenue growth of 28%). ATI indicated that the analysts’ EPS estimates for F2005 are in a reasonable range (we estimate $1.00 vs consensus of $1.02). ATI’s F2005 to F2007 targets are for gross margins in the 34% to 38% range, and pre-tax income in the 11% to 15% range.

ATI continues to have capacity constraints, with demand outpacing supply. We believe the supply constraint is centered on ATI’s 0.13 micron, low-k process chips, including the PCI Express desktop discrete chips RADEON X600 and RADEON X800, which are manufactured at a TSMC 8 inch wafer fab. We do not believe ATI is capacity constrained for its 0.11 micron, FSG process chips, namely the PCI Express RADEON X300 and RADEON X700. The company indicates it has not experienced a slowdown in desktop discrete chip orders, and commented that its desktop PC business unit is more bullish now than it was six weeks ago. Management is furthermore not concerned over the risk of its channels placing double orders, emphasizing that it closely monitors the channel inventory and compares historical ordering patterns to avoid shipping into such situations. Although Mercury Research expressed concerns over a potential inventory build in mobile integrated chipsets, ATI indicates it has not experienced a slowdown in this segment, with particular strength in the AMD K7 mobile
PC platform.

Much discussion has arisen this week with respect to the Q3 market share shifts in the DirectX 9 desktop discrete graphics chip market, catalyzed by NVIDIA’s comments on October 25 that it had share gains in the DirectX 9 performance sub-segment, defined by Mercury Research as chips having over 100 million transistors. We believe NVIDIA’s share gains in the DirectX 9 performance sub-segment are misleading. The DirectX 9 performance sub-segment as defined by Mercury Research is not segmented based on retail board prices, but instead includes a mix of both high and low price points. Clearly a lower priced graphics board will drive relatively more unit volume than a higher priced offering. We therefore do not believe investors should focus on the two sub segments and believe it is most important to review the market share gains in the DirectX 9 segment as a whole. In the entire DirectX 9
segment, ATI gained 12% of market share sequentially to 46% (5.1 million chips) from 34% (3.2 million chips), and NVIDIA lost 11% of market share to 54% (6.0 million chips) from 65% (6.0 million chips).

ATI is focused on growing its desktop integrated chipset business, targeting $150 million in revenue in F2005. We estimate desktop integrated chipset revenue was $26 million inF2004, and forecast $108 million in F2005, which may prove conservative relative to ATI’s internal target. We believe ATI is gaining significant traction with its upcoming PCI Express desktop chipset portfolio expected in H1/F05, with a goal of exiting F2005 with a number three market share position and exiting F2006 with the number two position (up from only 2% market share in Q3/04).

ATI’s digital TV chip growth is expected to exceed 2.5x in F2005. We estimate digital TV chip revenue was $63 million in F2004, and forecast $190 million in F2005. Digital TV’s have a 12 to 18 month design win cycle, giving ATI significant visibility for F2005. ATI estimates its front-end XILLEON (HD MPEG-2 decoder) has a 40% ATSC market share of all integrated digital TVs in North America at the end of 2004. ATI estimates its back-end Theatre/NxtWave (terrestrial and cable demodulator) has an 85% market share of the North American cable and terrestrial integrated digital TV market. ATI expects its wireless handset chip business to grow 2x in F2005, triggered by a volume ramp with other tier one handset OEMs in Q3/F05 (management specifically mentioned Samsung and Siemens as possibilities), and by broadening its partnership with Motorola (currently 14 GPRS design wins and one Nextel iDEN design win). We estimate wireless handset chip revenue was $134 million in F2004 (over 5x year-over-year growth), and forecast $270 million in F2005. ATI is focused on the top 12 handset OEMS (Nokia, Motorola, Samsung, Siemens, Sony Ericsson, LG, Panasonic, Kyocera, NEC, Sagem, Ningbo Bird, and TCL), which represent a combined 90% of the market. The QUALCOMM MSM7xxx baseband chipset royalty agreement is expected to ramp in F2006.

Justification of Target Price: ATI trades at 17x our C2005E EPS, at a discount to NVIDIA at 25x, and at a discount to Intel trading at 20x. We believe a fair valuation for ATI is in the 20x to 25x C2005E EPS range, and apply a 22x P/E, but recognize that Intel’s contracted valuation may continue to be a governor for ATI’s P/E multiple.

Key Risks to Target Price: Risks include a slowdown in PC sales; competition from NVIDIA and Intel; inability to secure PCI Express design wins; and unexpected delays in shipping new products.

Conclusion and Recommendation: We have increased confidence that ATI is well positioned for the holiday season and believe that competitive concerns over NVIDIA’s positive pre-announcement are overblown. We expect ATI’s F2005E growth will be driven by sustained market share leadership in PC graphics (we strongly believe ATI’s PCI Express roadmap of the X300, X700, X800, and upcoming R480 enthusiast chip will sustain its leadership position with both OEMs and white-box channel manufacturers), triple-digit growth in ATI’s consumer business (guidance is for wireless chip revenue to more than double in F2005 and for Digital TV chip revenue grow over 2.5 times), and ATI’s three royalty streams from QUALCOMM, Microsoft, and Nintendo (expected to begin in late-2005, adding a potential $0.25 in combined annual earnings power). We maintain our BUY recommendation on ATI and $23.00/C$28.00 target (22x our C2005E EPS of $1.05).

 

[digitalwanderer]

Seems that there is some doubt over how the "Performance" calculations were done.

Due to a somewhat arbitrary definition of 'performance' done according to transistor count, rather than actual performance, Mercury included Nvidia's 6800 (enthusiast), 6600 (performance) and 6200 (mainstream) products in the 'performance' category, while only including ATYT's X800 (enthusiast) and a few new X700's (performance) products and excluding ATYT's X600 and X300 products. As a result, the performance market size and Nvidia's share in it, was overstated. Based on our checks, we believe Nvidia's actual 6800 shipments were on the order of 275,000 (plus or minus 25,000) compared to around
382,500 X800's for ATYT.

So ATi has the high-end now too as well as the low-end? Or am I reading this wrong? :|

 

[Pete]

Direct links, Dave and Ratchet? And are we talking chips or cards, AGP or PCIe? Is Dave's link including 6800s and LEs and X800SEs and Mobility Radeon 9800s (based on chips)?

 

[webmedic]

Seems that there is some doubt over how the "Performance" calculations were done.

Due to a somewhat arbitrary definition of 'performance' done according to transistor count, rather than actual performance, Mercury included Nvidia's 6800 (enthusiast), 6600 (performance) and 6200 (mainstream) products in the 'performance' category, while only including ATYT's X800 (enthusiast) and a few new X700's (performance) products and excluding ATYT's X600 and X300 products. As a result, the performance market size and Nvidia's share in it, was overstated. Based on our checks, we believe Nvidia's actual 6800 shipments were on the order of 275,000 (plus or minus 25,000) compared to around
382,500 X800's for ATYT.

Yes I would like to see some links if we can. Other than that I know as well as you that sometimes some of us know things that there is no link for.

At any rate to quote a guy I used to work for" Thats an owie"

 

[pcchen]

Hmm... I think there are only PCIe version of 6600, and 6200 is still not on the market. So even if 6600 are included, it can't be really offset that much since PCIe is a realtively small market.

 

[Dave Baumann]

PCIe is a relatively large market when you compare it against the very high end as a whole.

Anyway, you'll not find a link for the reports themselves since they are analyst reports. First one was from Goldman Sachs.

 

[Mark]

The one I posted was from TD Waterhouse, no direct link to it.

 

[Bouncing Zabaglione Bros.]

The one I posted was from TD Waterhouse, no direct link to it.

Well it's interesting that both they and Goldman-Sachs are telling us not to pay attention to Nvidia's recent spinning of the "top DX9" numbers. It's also worth noting that big companies like Nvidia have to be somewhat more truthful to the business community (both legally and for their own self-interest) than they are to the buying public.

I guess Nvidia are trying to make the best of a bad lot and ignoring the rest of the reports.

 

[WaltC]

NV did much the same last year by including the 5200 in its '03 sales projections for its "DX9" gpus. Then there was the unforgettable PR bulletin nV circulated as to how "100,000,000 nVidia gpu users can't be wrong," (or words to that effect), as if 100% of that 100M figure were "DX9 gpus," instead of the actual < 1/100 of 1% of that number being "DX9" gpus at the time of the statement. I think the degree to which nV feels compelled to spin and manipulate the numbers as they emerge shows the effect of the competitive pressure ATi has been applying in the market for the last couple of years.

 

[Bouncing Zabaglione Bros.]

I think the degree to which nV feels compelled to spin and manipulate the numbers as they emerge shows the effect of the competitive pressure ATi has been applying in the market for the last couple of years.

It's more disappointing when you dig under these numbers that sound great initially, only to find it's a fiction. It's a real sign of desperation if that's the best Nvidia can come out with, but I suppose it follows Nvidia's normal modus operandi of lying to all and sundry.

Nvidia really need to consider purging their upper management and reorganising the way they do business.

 

[Dave Baumann]

NVIDIA didn't actually produce the figures.

 

[Bouncing Zabaglione Bros.]

NVIDIA didn't actually produce the figures.

You think they spun the numbers without knowing how they were compiled? I'm sure Nvidia know how many 6800's they actually shipped.

 

[ChrisRay]

Here's my question. Why the heck is Mercury research even lumping midrange/low end cards into the high end category?

This includes the 6600/6200/X700.

 

[Mark]

For how much of an impact these reports make on the industry it should be made exactly clear how they arrive at their numbers.

 

[Pete]

NVIDIA didn't actually produce the figures.

They just directed.

 

[martrox]

It's just another nVidia half-truth......... typical nVidia ethics....... Why be suprised?

 

[jimmyjames123]

So Mercury Research compared 6800/6600/6200 vs X800/X700? Obviously it makes no sense to include 6200. However, realistically, how many 6200's could they have sold anyway? These products have just barely been announced.

I think it's pretty hilarious that the usual suspects are criticizing NV's ethics. A 3rd party company comes up with some figures, and NV PR shows off the positive portion of these figures just as any other public company would do, and this is unethical? Hilarious. How come no one questioned ATI's ethics in boasting about a record 3dmark05 score that was based on a ridiculously overclocked video card?