Nvidia’s financial presentation, although targeting investors, has laid out Nvidia’s plans for the entire year, and it all rests on TSMC’s shoulders to bring home the bacon with its 28nm manufacturing process.
Financially, Nvidia presented a healthy balance for its consumer products, with a year-on-year growth of 200 percent, from $197.6 million to $591.2 million. Its GPU business, both consumer and professional GPUs and GPGPU, was stable, having gone from a total $3345.1 million to $3406.7 million.
The ‘consumer products’ represent Nvidia’s Tegra business, and more specifically, Tegra 2 which went into hyperdrive in the last quarter of the year. Tegra 2 accounted for $260 million in revenues, Nvidia believes with Tegra 3 it can increase that by 50 percent this year.
We will be able to see Tegra 3 products very soon, as it will take front stage at the upcoming Mobile World Congress in Barcelona, where the company will premiere the first design wins, it was said.
Overall, margins are down, revenues are down in the last quarter, mostly due – says Nvidia- to fewer PC sales thanks to the HDD supply shortages. This means fewer machines and less discrete GPUs being shipped out, because integrators are just staying away from those pesky non-buoyant mechanical HDDs.
The conference, however, quickly turned into a debate of results-over-technical-capacity. Nvidia’s 2012 product portfolio rests entirely on TSMC’s ability to ramp up 28nm wafer production to meet the needs of the company’s GPUs (Kepler) and SoCs (Tegra 3). Simultaneously, the company touted to have won on these new products more design wins than any previous generation.
While Nvidia pointed out it has lost its special relationship with TSMC, somewhat, it has relied on wafer-based pricing like the remainder of humankind, and low-yields have a direct impact on the company’s margins.
As yields improve, costs will decrease and margins grow. Therefore, the company expects to finish the year strong.
Unfortunately, if you are relying on the 28nm process to get your entire product portfolio out the door then you rely on TSMC, which has not completely mastered the process.
“We are ramping very hard all of our 28nm products. The yield is much better than 40nm when it first came out, but it isn’t as good as we had expected in our previous projections,” said Rob Csongor, VP Investor Relations at Nvidia.
Both Rob and Jen-Hsun couldn’t stress enough that the company could achieve even greater things if only TSMC’s yields allowed it to.
In an odd way, this could be quite upsetting for Nvidia. Its success in marketing and the promised quick ramp up of its 28nm-based Tegra 3 and Geforce/Tesla products will pile up orders from clients eager to integrate them , and may put the company under pressure to meet demand.
More to the point, this is one of the main reasons Nvidia’s GPU rival, AMD, is not seen lowering prices on its flagship Tahiti chips.
The entire phone conference was overcast by the looming insufficient yields of TSMC’s 28nm process, which Nvidia’s top guns couldn’t stop mentioning, yet vesting full confidence in TSMC’s ability to overcome it.
The industry will be completely constrained throughout the whole year, said Nvidia, so it is reasonable to think that prices on GPUs will remain high, in particular if Nvidia indeed takes the performance crown from AMD with its new Kepler architecture.
Nvidia seems particularly confident about being able to do this. During the call, Nvidia said: “Our Kepler generation won design wins at virtually every single PC OEM in the world. Performance and power efficiency is surely the best we ever created. This is expected to be very high.”
Nvidia is definitely timing the launch of Kepler with Intel’s Ivy Bridge, as we’d mentioned before, trying to steal some thunder from the integrated graphics market. It wants to cash in on the upgrade or the product line refresh Intel will merrily co-market with its PC partners.
After the 28nm dust particles setled, Jen-Hsun Huang had his opportunity to fire off some replies and lay out his vision of the market for 2012 and 2013.
It turns out that smartphones are the way to go, with Nvidia positioning itself quite well as this year “quad-core” is in for SoCs, and next year it’ll be full baseband integration.
Nvidia has been “lucky enough” to pick up Icera, so future phone products will break down into very powerful, LTE-integrated devices and lower-cost dual-core phones with the processing power of the current iPhone 4S, selling for as low as $150.
It’s a huge gamble for Nvidia right now, but all things considered, by the end of the year it will have paid off. At the end of the day, the take home message for investors is: Nvidia can make a killing if TSMC lets it, so it won’t be anyone’s fault but the fab if supply is constrained, or prices are high.
If you’re a bit of a geek like us, you’ll want to see what that means product-wise. Well, just read between the lines: “TSMC”, “fault”, “fab”, “constrained”, “prices”, “high”.