A report from Doug Freedman at RBC Capital (via Barron's) says graphics chip designer NVIDIA (NVDA 2.15%) has shifted some of its chip manufacturing from longtime partner Taiwan Semiconductor (TSM 1.92%) to Samsung (NASDAQOTH: SSNLF). The analyst cited "conversations with management," as well as a recent NVIDIA 10-K filing that indicates the company uses "industry-leading suppliers such as Taiwan Semiconductor Manufacturing Company Limited and Samsung Electronics Co. Ltd." to build its chips.
Although Freedman expects Taiwan Semiconductor will build most NVIDIA chips in 2015 and "probably" 2016, his report raises a very interesting point about what could happen beyond that. Freedman said he believes NVIDIA could use its relatively sizable foundry business ($1.7 billion of revenue per year per the analyst's estimate) as leverage to "increase the likelihood of collecting future royalty revenue streams from Samsung and/or Intel (INTC 1.69%)."
I would like to take a closer look at this from the perspective of Intel as a potential foundry partner to NVIDIA.
Digging a bit deeper into this
As it stands today, NVIDIA collects $264 million per year from Intel as part of a licensing agreement between the two companies. As has been widely noted, this revenue stream (which is essentially pure profit) is expected to dry up by early 2017 if the deal is not renewed.
Let's suppose Intel were to get substantially all of NVIDIA's chip manufacturing business in exchange for renewal of this cross-licensing deal. Intel would presumably get $1.7 billion in annual foundry revenue at approximately a 50% gross profit margin, but it would have to cough up $264 million per year in licensing payments.
This would work out to about $586 million in incremental gross profit for Intel. To put this into perspective, Intel generated approximately $35.6 billion in gross profit during 2014, so this potential deal would amount to a 1.6% increase. However, given that 2015 is expected to come in weaker than 2014 for Intel, the impact should be more pronounced relative to 2015 numbers.
What if Intel does not get the whole pie?
However, I am not convinced NVIDIA would move 100% of its business over to Intel. It makes sense from a supply chain management perspective to reduce its dependence on a single supplier.
So, if Intel is added to the mix, I would not be surprised if it were to build its highest-end graphics processors (where NVIDIA might be willing to a pay a premium for Intel's wafers), while Taiwan Semiconductor and Samsung split the more mainstream and cost-sensitive GeForce/Tegra products.
In fact, even if we generously assume Intel,Taiwan Semiconductor, and Samsung split NVIDIA's orders evenly, the economics of the licensing-for-foundry do not work. In that case, Intel would reap about $567 million in NVIDIA foundry revenue and roughly $283 million in gross profit. That would just about neutralize the financial impact of a potential renewal of the current licensing deal.
In the case that Taiwan Semiconductor is cut out of the supply chain completely (which I would not bet on) and Samsung and Intel split the $1.7 billion foundry pie evenly, Intel would net $161 million in annual gross profit after paying the $264 million licensing fee. This does not seem like it would be worthwhile partnership for Intel.