My Authors
Read all threads
OK, $AMD financial day megathread. They are releasing their financial outlook for the next 4-5 years (!!)

I'll put my thoughts in this thread, presentation is at ir.amd.com/events/event-d…
This is AMD's estimated TAM, $79 Billion. Lisa Su says this is ~2023 timeframe
Lisa Su - "Count on us to have a very strong cadence" going forward

Zen 3 & 4, RDNA 2 & 3
Also talking a lot about packaging and interconnect being a big deal for AMD.
AMD "CDNA"
New GPU Compute Architecture for HPC and Machine Learning. Will have it's own multi-generational roadmap.
Expect a similar cadence for CDNA as with Zen (which is going for a 1-1.5 year cadence
Ooh, the juicy stuff! New financial guidance. Long term expecting 20% compounded growth rate, >50% gross margin, mid-20%s operating margin (think of how much profit that means), > 15% free cash flow. Glad I bought a bunch of calls today 😁
Really really wish I had caught $AMD at ~$41/share it hit last week, but I was traveling 😭
Interesting slide, Zen cores multiplying like tribbles 😅
Mark Papermaster: Zen 3 on track for release late this year. Next Zen core (Zen 4, I presume) will be based on 5nm.
AMD used to fighting against Intel with a worse process. Now they have (unexpectedly) a better process, and are like a kid in a candy store, Mark Papermaster failed to add.
3rd gen AMD Infinity Architecture providing "up to 8-way GPU with coherent connectivity"
Intel is doing something like this with their Xe project, I think. At least being able to combine multiple chiplets in GPUs. I wonder what Nvidia's plan is for this, or are they getting left behind?
Pretty soon the 2 most powerful computers ever built will have AMD inside. Unless China comes out with something unexpected.
Ok, looks like RDNA is already the first example of the bifurcation of the GPU strategy from general to gaming and compute optimized. So probably no RDNA for compute?
Ray tracing and variable rate shading coming in RDNA 2 and will be out "later this year"
AMD's multi-GPU solutions will scale pretty well
AMD claims the lead over Intel for single core performance in mobile chips
AMD more than doubled unit marketshare *before* 7nm, expecting more with 7nm
Hey look, they sell commercial products too! 😄 Growing share even before 7nm, getting more platform design wins, and another big boost coming with the 7nm chips.
I know AMD is crushing in the gaming market, especially consoles, but whenever AMD executives talk about gamers I get this mental image 😂
OEMs embracing AMD GPUs, if I understood correctly they state they got 3x the number of SKUs for this generation.
AMD's RDNA2 coming out "at the end of this year"
"More APUs this year than ever"
I'm sorry, I kind of like APUs, looking forward to some good ones!
They're having an intermission now, not sure when they come back
I missed this slide, RDNA2 expected to be to RDNA what RDNA was to GCN, as far as performance per watt. Which was a pretty huge jump. NVidia better be on their A game. Competition, good for consumers!
Wonder if the 50% perf per watt improvement over RDNA1 is with or without ray tracing?
And we're back. Forrest Norrod up to bat, expecting some hype! 😁 He speaks kind of monotonously but he's a bit of a showman IMO.
@anandtech already has an article up by @RyanSmithAT about RDNA2 anandtech.com/show/15591/amd…
I'm skipping a lot of stuff that has been presented in the past, for example a few of these slides explaining the Epyc Rome is the best server chip around.
Forrest Norrod saying again that AMD will hit double digit server market share in Q2. Which is pretty soon, so I think they are pretty confident about that at this point.
Expecting almost triple the platforms running AMD chips in 2020 compared to 2018
As Forrest Norrod takes a few victory laps about Rome, and explains how Milan is going to continue the carnage, shipping "later this year", let me say more investors should be following @EricJhonsa. I could recommend many, but how does he have less than 4K followers?
Forrest would like to remind you all that AMD GPUs have hardware based virtualization, and others don't
Ok this has come up several times now. Basically the innovation that AMD is doing with CDNA architecture plus inifinity fabric, it's going to reduce latency between CPU and GPU, and allow developers to treat them as one unit. Or something like that.
I think with this type of thing, plus AMD using open source standards for a lot of things, means that AMD is going to become the go-to for big government projects. Which is kind of already happening.
Devinder Kumar, CFO of $AMD is up now. Financial details incoming!
This is the last 4 years of gross margin and Opex. Imagine gross margin going up above 50%, which is what Lisa Su said. Opex will also increase but stay flat as a percentage of revenue, mid-20%
Nice data, but along with @IanCutress I hate these charts without labels on the y-axis that don't start at 0 :(
2020 Financial guidance
Long term model, same as what was posted towards the beginning of this thread
Datacenter expected to go from 15% of revenue to >30% of revenue for $AMD
$AMD planning on returning value to shareholders (dividends or buybacks?)

Look at them, gonna be a blue-chip company with investment grade credit
@LisaSu is up, going to summarize key takeaways and then Q&A. Not looking like any specific product reveals. Big Navi might be waiting for RDNA2, which they said would get up to the top of stack.
Lisa addressing coronavirus impact in short-term. Supply chain is mostly in China, Malaysia and Taiwan. Already back to near normal supply level. They are also monitoring their customers which have seen disruption, and they are making progress.....
Outside China demand as expected, inside China demand down a bit, but also increased demand from datacenters or something I missed.
Not updating guidance, impact will be modest, might be at lower end of guided range, but Q1, Q2 guidance is unchanged. This is a big positive IMO.
Tweet worth retweeting: $AMD doesn't see the need to update guidance for Q1 and Q2 when accounting for coronavirus impact. Expect to still hit the guidance range.
CDNA perf per watt will be increasing similar to RDNA improvements (which are drastic, 50% gen on gen)
Mark Papermaster, when asked about customers making their own chips, Arm competitors. Basically says there will always be custom silicon for specific workloads.

What he forgot to say was that eventually their distinctiveness will be added to the x86 ISA
CDNA adding tensor ops needed for acceleration, and optimizing architecture for compute
El Capitan supercomputer will be based on Genoa (server chip after Milan, which is after Rome, which is currently out). Seems like the coherency between CPU and GPU is fully realized in CDNA 2, so looks like that killer combo is Genoa + CDNA 2, not Milan + CDNA 1
And $AMD's financial analyst day 2020 is over!
The AMD website has been updated with the presentation materials: ir.amd.com/events/event-d…
Missing some Tweet in this thread? You can try to force a refresh.

Enjoying this thread?

Keep Current with siliconmemes

Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

Twitter may remove this content at anytime, convert it as a PDF, save and print for later use!

Try unrolling a thread yourself!

how to unroll video

1) Follow Thread Reader App on Twitter so you can easily mention us!

2) Go to a Twitter thread (series of Tweets by the same owner) and mention us with a keyword "unroll" @threadreaderapp unroll

You can practice here first or read more on our help page!

Follow Us on Twitter!

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just three indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3.00/month or $30.00/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!