Posted: By Jon Peddie 07.24.20
One rumor, one report, three results
The story put out by Bloomberg was that Nvidia may be bidding to buy some or all of Arm from Softbank. That is such a delicious rumor-speculation I just can’t ignore it.
If Nvidia agreed to Softbank’s inflated price—Arm couldn’t have possibly increased in value so much from the time Softbank bought it—it would have to do it on a stock deal and maybe borrow some. Nvidia has been very cautious with its balance sheet and is building back its nest egg after the outrageous price it paid for Mellanox. Nvidia currently has almost $11 billion in the bank. Some people think Softbank will want $45 to $50 billion for Arm (they over paid at the time and bought it for $32 billion – over 32 times revenue!)
The Bloomberg story gave Nvidia’s already high-flying share price a little bump ($408 last time I checked).
If Nvidia did go for Arm, which I think would be a brilliant move, that would be a serious blow to Intel, and a potential torpedo for the open-source RISC-V adventure, of which Nvidia is a participant. Apple, Qualcomm, and others wouldn’t be too bothered as they have Arm architectural licenses and can do their own thing.
The second thing to happen this week was Bob Swan’s announcement that he might use a third-party fab for their dGPU project. Rumors have been floating around that the Xe dGPU is in trouble and Intel has a due date on the Ponte Vecchio for the Aurora supercomputer project. Intel also announced its 7nm process had slipped a year.
And, just coincidently, AMD announced a big family of APUs based on its successful Zen architecture for OEMs and SIs.
Intel’s and AMD’s announcements resulted in AMD share price jumping 16%, (to $65.6 last check) and Intel’s tumbling 16% (to $50.5).
The what if what if
If Nvidia picks up Arm that will put them right in the face of Intel as a vertically integrated data center and edge server supplier with CPU, GPU, and interconnect. Nvidia has also been investing a ton of money in the software support for the data center and AI, possibly even exceeding Intel’s own huge software investment.
Meanwhile AMD’s Arm-based server ambitions may be in question if Nvidia takes over. But, AMD is also a RISC-V member so there may another route for the company if Nvidia makes things difficult. However, it would not be in Nvidia’s interests to make life difficult for AMD and others because Nvidia needs additional Arm-based server suppliers if it is going to create a category.
Intel hasn’t been asleep, and is rolling out some amazingly powerful and low power X86 processors that will give Arm a run for the money and the FLOPS/W.
Adding Arm to Nvidia’s revenue will be nice but not great. Before Softbank swallowed them up Arm was doing about $900 million; give them the benefit of doubt and call it a Billion and you’ve added maybe 10% to Nvidia’s top line. And that’s good because Arm’s margins were about 40% and Nvidia wouldn’t want their 65% margin diluted too much.
The deal will be tricky. Nvidia can’t use all its cash because its been on an aggressive program of buying back its shares. That helps the remaining shareholders, so the rubric is going to be to balance shareholders wealth with the Arm deal. But Nvidia is clever and I have no doubt that if they do the deal it will be a sweetheart.
It does put a lot of moving parts into play and that is the wonderfully disruptive thing that make working in this industry so much fun. But then, it’s all just a rumor, right? Right?