Patrick Wang of Evercore Group, who rates Intel shares Underweight, and rates shares of Applied Micro Circuits ( AMCC), an ARM licensee developing server chips, writes today that the Bloomberg story accords with what he’s heard, and he thinks it’s bad news for Intel, but he also thinks things are a bit more complicated:
We agree (mostly) and think it’s not just a possibility but believe designs are already underway [...] This is where we differ. Server chips can be categorized as lightweight (scale-out, distributed) or heavyweight (big iron, heavy duty). Our checks indicate a focus on the latter with early designs using the POWER architecture from IBM ( IBM) and not ARM. Leveraging a feature called CAPI, we think these efforts are performance driven due to ongoing CPU interfacechallenges [...] While we believe work has not begun on ARM-based SOCs, it’s quite possible Google is currently test driving ARMv8 or may have even signed an undisclosed license already. Nonetheless, it’s clear this work would be motivated by cost. Given Google’s scale, it’s worth noting that Intel’s blended server ASP has risen from $446 to $580 over the last 5 years [...] The pursuit of a custom lightweight SOC could also drive Google to be strategic. First, timing as it would basically take ~2 years to design an SOC from scratch. Second, Google lacks certain critical IP blocks such as I/O, optical, and a fabric needed to drive massive scale-out systems. Third, they generate a lot of cash – FCF in 3Q13 was $2.8bn. There are already a number of companies investing in 64b ARM chips. Applied Micro leads the pack with X-Gene 1 shipping and X-Gene 2 samples in 1Q14. Others: Samsung ( 1H15), Broadcom( BRCM) (2H15) AMD (1H14), Cavium ( CAVM) (2H14), and Calxeda (private, 2014).
On the other hand, Credit Suisse‘s John Pitzer, who has an Outperform rating on shares of Intel, and a $30 price target, writes that the speculation is not new, and there are several reasons to believe that even if true, it won’t amount to much, and certainly not much of a threat to Intel:
We would note: (1) speculation of Cloud/Web 2.0 providers
(GOOG, FB and AWS) developing optimized ARM-based CPUs is NOT new,
(2) the report, if correct, likely signals that third-party vendors trying to
develop ARM-based CPUs are perhaps underwhelming, and customers arelooking for other avenues to minimize single source dependence, (3) thereport does NOT change our view that ARM-compatible workloads will bestcase overlap with only 10-20% of the Server CPU market that is ALREADYbeing addressed by INTC’s Avoton’s platform, (4) lack of 64-bit softwareecosystem limits the real threat to 2015 NOT 2014, (5) lack of volume scalein the Server CPU market limits leverage customers are likely to receivefrom foundries – AAPL will produce 225-250m A-class ARM-based APs thisyear, and INTC’s entire DCG business ships approximately 20m units a year(10-20% of the Server CPU market would imply 2-4m units), (6) INTC isalready working closely with at least 8 customers for x86 optimization and(7) the TCO benefits of ARM versus x86 seem uncompelling.
Market Data 
Markets 
AI
