( ESNUG 600 Item 1 ) ------------------------------------------------------- [12/05/25]
Subject: Jensen cleverly injects $2B into SNPS to thwart AMD/Cerebras/Intel's AI chips
"A man always has two reasons for doing anything; a good reason and the real reason."
- J.P. Morgan, American financier & investment banker (1837 - 1913)
"WHAT THE EFF IS GOING ON !?!" -- four days ago, Nvidia Jensen and SNPS Sassine had
a surprize announcement on Monday (12/01/2025) where they were bringing Nvidia CUDA
Blackwell-Vera-Rubin OmniVerse AI deeeeeeeep inside all the SNPS/Ansys multi-physics
tools promising super amazing 20X to 1000X speed-ups and through-puts:
(click pic to enlarge image)
and in their video they specifically bragged on how ANSS Redhawk-SC, ANSS IcePak, ANSS
Fluent, ANSS Mechanical, ANSS Maxwell, ANSS PercieveEM, ANSS Discovery, ANNS LS-DYNA,
ANSS HFSS and ANSS pyAnsysHeart were already getting kickass runtimes with Nvidia AI.
(click pic to enlarge image)
On top of that, in addition to already having SNPS Verilog simulation and SNPS SPICE
simulation -- they also announced how Synopsys PnR and Synopsys DRC/LVS was going
whole hog into Nvidia CUDA Blackwell-Vera-Rubin OmniVerse AI in 2026 !!!!
---- ---- ---- ---- ---- ---- ----
---- ---- ---- ---- ---- ---- ----
---- ---- ---- ---- ---- ---- ----
OF COURSE, SASSINE LOVES THIS! -- what's even wilder is NVIDIA is directly injecting
$2 Billion into SNPS (at $414 a share) to own 2.5 percent of SNPS.
THE BIG QUESTION IS WHY IS JENSEN GIVING SASSINE $2 BILLION IN CASH RIGHT NOW ???
To sleuth this out, I asked Jay Vleeschhouwer of Griffin Securities roughly how
much does NVIDIA usually spend on SNPS tools every year?
Jay said "I can't tell you that."
Me: "OK, then how much does NVIDIA spend total on all EDA tools worldwide per year?"
Jay said: "Worldwide NVIDIA's total annual spend for EDA is roughly $500 million."
Then -- using EDAC's global market share stats (45% SNPS, 40% CDNS, 15% MENT) -- this
NVIDIA $500 million total annual spend for EDA tools scales out roughly to:
SNPS products: :############################################# $225 M (45%)
CDNS products: :######################################## $200 M (40%)
MENT products: :############### $75 M (15%)
So
$2,000 M / $225 M == 8.8 X ~== 9 X
That is, Jensen is effectly fronting Sassine a whopping 9 years of sales in cash for
free (pretty much), so Sassine can make everything SNPS sells to be based on Jensen's
Nvidia CUDA Blackwell-Vera-Rubin OmniVerse AI !!!
(How is this free money? $2 billion for 2.5 percent of SNPS doesn't really dilute
the stock much -- and Sassine is still going to get his usual $225 M in annual sales
to NVIDIA. Plus it's free money for Jensen because NVIDIA gets $2 Billion in assets
and it makes all SNPS+ANSS customers into NVIDIA AI chips buyers!)
---- ---- ---- ---- ---- ---- ----
---- ---- ---- ---- ---- ---- ----
---- ---- ---- ---- ---- ---- ----
JENSEN IS PLAYING 3-D CHESS -- even though he currently owns 80% to 90% marketshare in
AI chip sales today, Jensen knows that in high tech things can easily change overnight
if NVIDIA isn't careful. (For example, how Intel went from clearly owning the chip
world for 5 decades -- over into desperately struggling to make a comeback in 2025.)
Looking over his shoulder, Jensen can clearly see his AL chip rivals chasing him.
As some quick and lazy research, I googled "Nvidia AI chip competitors" and using what
Google AI spit out as a base, I did another 25 other follow-up google searches to suss
out exactly who Jensen is worried about -- and why (or why not) they're a threat.
Here's what I found:
JENSEN'S 2026 THREAT ASSESSMENT REPORT:
NVIDIA's main competitors in AI chips include AMD and Intel, which offer rival
GPUs, as well as major cloud providers like Google, Amazon (AWS), and Microsoft
that design their own custom AI chips. Other significant competitors are
Cerebras Systems, Huawei, Qualcomm, and Groq.
Chip manufacturers
- Advanced Micro Devices (AMD): Offers the Instinct MI400 line of GPUs as a
direct competitor to NVIDIA's offerings. AMD did a 10 percent of its stock deal
with OpenAI ChatGDP expecting $100 B in revenue over the next 4 years from it.
The AMD Instinct AI chips are absolutely the #1 2026 THREAT to NVIDIA.
- Cerebras Systems: is known for creating very large AI chips, such as
its Wafer Scale Engine. It's WSE-3 has 900,000 AI cores with 125 petaflops
of compute power on one chip. Cerebras CS-3 is 21x faster, 1/3 lower cost, and
1/3 lower power than Nvidia's DGX B200 Blackwell GPU. #2 2026 threat to NVIDIA.
- Huawei: has the Ascend 910, 950, 960, 970 series which is aimed to directly
compete against NVIDIA in the Chinese market to power China's DeepSeek AI.
Ascend series is #1 threat to NVIDIA in China -- but not a threat anywhere else.
- Intel: Competes again NVIDIA its Gaudi 3 AI accelerators and other offerings.
Problem is Gaudi is in 5nm and Intel itself is struggling. Marketed as cheaper
rivial to Nivia H200, but users leary. Intel AI is NOT a 2026 threat to NVIDIA.
- Qualcomm: A smartphone chip giant expanding into the data center market with its
own AI200 and AI250 AI processors. Qualcomm sells its AI chips to datacenters,
PCs, and the Android smartphone market. This is a possible #3 threat to NVIDIA
for 2026 in Qualcomm's respective niches.
- Apple: has its owwn GPU-based M4, A19 Pro, and recently announced M5 AI chips.
Is NOT a threat to NVIDIA in any way because all Apple AI chips are 100% used
internally on Apple products.
Cloud service providers developing custom AI chips
- Google: Develops its Tensor Processing Units (TPUs), such as the Ironwood TPU,
for its own AI systems. No threat to NVIDIA other than loss of sales to Google.
- Amazon (AWS): Develops its own custom AI chips, including the Trainium
and Inferentia series, for use in its cloud services. Not a threat to
NVIDIA other than loss of sales to AWS.
- Microsoft: Is also developing custom AI MAIA 100 silicon to reduce reliance on
external suppliers. No threat to NVIDIA other than loss of sales to Microsoft.
- Meta (Facebook): Like other hyperscalers, Meta produces its own MTIA v2.
Not a threat to NVIDIA other than loss of sales to Meta (Facebook).
Other specialized competitors
- Groq: Offers its Language Processing Unit (LPU) specifically for AI inference.
It enables speeds up to 10x faster than GPUs for certain tasks with lower power
use, serving both cloud and on-prem needs. Unknown threat status to NVIDIA.
- SambaNova Systems: Develops its own SN40L processors for AI workloads.
Unknown (but probably small) threat status to NVIDIA.
- IBM: has its own NorthPole AI chips. No threat to NVIDIA.
---- ---- ---- ---- ---- ---- ----
DENY AMD THE HIGH GROUND -- out of the 13 AI chip rivals listed above, it's obvious that
Jensen's immediate concern is the AMD Instinct MI400 line of GPUs. AMD is the clear
and present danger for now.
So Jensen's best strategy to counter AMD is have his NVIDIA chips lock in the seriously
compute intensive users as possible before the AMD AI chips can sneak in.
And name one thing more compute intensive than multi-physics simulation & design!
(click pic to enlarge image)
Which is why Jensen is being so generous & lovey dovey with Sassine's Ansys + Synopsys.
By using free money ($2 B for SNPS shares that'll most likely grow in value) -- plus
committing some of his NVIDIA AI engineering geniuses to the project -- Jensen gets
to lock in the users of the largest multiphysics tool vendor, Ansys, into having to
buy his NVIDIA AI chips if they want those kickass accelerated runtimes! Sweet!
---- ---- ---- ---- ---- ---- ----
"THIS IS A NON-EXCLUSIVE DEAL!" -- this is the Get Out Of Jail card for both Jensen's
and Sassine's benefit.
That way, Jensen can work with Siemens/Altair (multiphysics vendor #2 in 2026), and with
Cadence/Hexagon (multiphysics vendor #5 in 2026) to deeply adopt his NVIDIA AI solution.
(EDITOR'S NOTE: Click to enlarge the bar graph image above the Get Out Of Jail
Free card to see what I'm talking about. The bar graph two images up!)
And Sassine can work with AMD, Cerebras, Intel, Qualcomm, Google, AWS, Groq, etc., to
use Synopsys/Ansys tools to design chips that directly compete against NVIDIA.
All this deal does is gives Sassine $2 billion to spread NVIDIA AI all throughout his
entire SNPS/ANSS offering -- but I won't be surprised if Cadence and Siemens later also
get some sort of simular incentive to spread NVIDIA AI in their offerings, too.
So in summary, the NVIDIA-SNPS deal was Jensen's clever way to lock NVIDIA AI chips into
the seriously compute intensive applications; while, through intertia, also effectively
blocked AMD's Instinct MI400 GPUs from entering this compute intensive high ground, too!
"Sorry, AMD, but we only have so many R&D guys to support AI chips; and they're all
focused on Nvidia CUDA Blackwell-Vera-Rubin OmniVerse in our EDA tools right now..."
And Jensen did this using effectively free money! Effing sweeeeeeet!
And Sassine won because he got a customer to fork out $2 Billion to pay for the upgrade!
- John Cooley
DeepChip.com Holliston, MA
---- ---- ---- ---- ---- ---- ----
"A man always has two reasons for doing anything; a good reason and the real reason."
- J.P. Morgan, American financier & investment banker (1837 - 1913)
Join
Index
Next->Item
|
|
