Nvidia, which has won the AI wave, now has a market value of 3.1 trillion, far exceeding the other two chip giants AMD and Intel.
But an insider revelation caused heated discussion:
Nvidia was almost sold to AMD that year.
The whistleblower is an old AMD employee named Hemant Mohapatra.
He revealed that AMD almost wanted to buy Nvidia at one point, but Huang's condition at the time was that he had to become the CEO of the newly merged company, otherwise Nvidia would not be sold.
This led to the failure of the transaction.
Hemant also shared many inside stories about AMD, which attracted more than 2.4 million people to watch. Netizens sat around on the sofa and listened with great interest.
As the heat continued to rise, a current AMD engineer named Phil Park also spoke up:
I have never met Hemant, and although I have different opinions on some issues, it is true that Huang wants to be the CEO.
Full story
Hemant worked at AMD in the mid-to-late 00s, involved in designing CPU/APU/GPU.
As mentioned above, he also shared more inside information about why AMD saw the GPU opportunity, missed it, and then regained the market when Nvidia developed more frantically.
Below is the full version.
After getting my master's degree, I received offers from both Intel and AMD. I finally chose AMD, even though the salary was 20% lower than Intel.
Growing up in India, I always thought AMD was the hacker choice, overclockable, cheaper, noisier, and more rugged. Somehow, it felt like rooting for the little David against the mighty Intel Goliath!
AMD had been playing catch-up to Intel throughout the 90s, but in 2003 we were the first to market with 64-bit chips and for the first time had a core architecture that was far superior to our competition.
Oh, those were exciting times! I haven’t seen such a strong embrace of hardcore engineering anywhere outside of Silicon Valley, or maybe NASA.
I joined AMD when its stock price was around $40 and was involved in the development of the first generation of dual-core architecture (single chip, dual core) AMD-X2.
Our first mistake, and AMD insiders probably won't like me saying this, was made here.
We are always engineering-driven and have excessive confidence in “building a pure dual-core” –
A single chip is split into two independent CPU cores with independent instruction and data pipelines, but shared caches, page tables, etc.
Even the wafer fabs are not yet ready for the relevant processes.
While we continued to struggle with “pure dual-core”, Intel, taking advantage of its x64 competitive defeat, moved quickly to simply combine two single cores together, do some clever interconnect processing, and market it as “dual-core”.
We joke inside AMD that Intel's marketing budget is higher than our R&D budget, and that's true. Consumers have been completely taken advantage of by Intel.
We did launch a “real” dual-core processor, but no one took it, and by then Intel’s “pseudo” dual-core was already gaining wide traction in AR/PR.
Then we started working on “real” quad cores, but again Intel simply combined two dual cores together and called it a quad core processor.
How did we miss this strategy?!
AMD always releases better CPUs, but they are always late to market. Consumers often cannot tell the difference between so-called “pseudo” dual-core/quad-core and “real” dual-core/quad-core, and are happy to see cpu{0-3} when executing the command `cat /proc/cpu`.
Until then, I was still a proud engineer, but then I saw the value of being first and getting to market quickly. I gradually understood that market recognition is a moat.
In the process of moving from dual-core to quad-core, AMD acquired Canadian GPU company ATI. In 2006, as an engineer, I didn't think it made much sense to acquire a GPU company.
At that time, the market was mainly focused on server and client CPUs, and GPUs were still a niche market. We didn't really need a GPU company, so much so that there was an internal joke: AMD+ATI=DAMIT.
But clearly someone at AMD saw the future. And we only saw part of it.
We should have bought Nvidia, and we tried.
Nvidia was mainly providing “niche” CPUs for hardcore gamers at the time, and they vigorously developed CUDA, while AMD firmly believed in OpenGL. Developers were more inclined to choose OpenGL due to the closed nature of CUDA.
Huang Renxun is clearly thinking long-term as he builds his “Apple” strategy of dual lock-in of hardware and software.
Huang refused to sell unless he was named CEO of the combined company. AMD backed off, and our future trajectories diverged forever.
ATI is a tricky thing, we don't really “understand” them, they see the world very differently than we do.
We wanted to merge the GPU and CPU into one APU, but that would take years of trust and process building to make them work together. Maybe if we had Slack that would be better, but we only had MSFT Sharepoint at the time.
While all this APU work was going on, AMD completely missed the iPhone wave.
When we made chips for PCs, the world moved to laptops; when we moved to laptops, the world moved to tablets; and when we moved to tablets, the world moved to phones.
We also missed the GPU wave when we tried to introduce the APU, a fundamentally better but completely new architecture. Creating a new category is extremely difficult, but “if you're going through hell, keep going”.
We hesitated, and then…
…The 2008 financial crisis hit, and we were in deep trouble. Since then, AMD has basically lost the market to everyone: Intel, ARM, Nvidia.
I learned the hard way how a quality product can lose out to a superior distribution lock-in and go-to-market strategy.
A few years later, in 2011, I left AMD and joined Google after a one-year short MBA program.
At the time, AMD's stock price was hovering between $2 and $4 – if we sold it for scrap, its stock price would even be below its book value.
I should have bought some stocks, but my salary came out about 30% lower than when I went in, and with the loans from my MBA, I was basically broke.
However, I still have great respect for Nvidia. They were a small player at that time, and we did lose some GPU sales to them, but we never regarded them as a competitor on the same level as ARM and Intel.
They stayed the course, and when the AI wave came, the market eventually turned to them. Believing in your vision and relentlessly pursuing your goals is a highly underrated skill.
Most people would have given up, but Huang persisted.
At the same time, I admire Lisa Su, who now leads AMD. Many of my engineering friends are still there developing the best technology in the world, and I can't imagine what they went through throughout the 2010s: multiple large-scale layoffs, salary cuts, leadership changes (CEOs were replaced twice between 2008 and 2011 alone), and low morale.
And Lisa Su is definitely a world-class leader. I sincerely hope that NVIDIA and AMD can merge now and leverage each other's manufacturing capabilities for technology cross-licensing, which will greatly help further reduce GPU cycle costs!
AMD is like a “clan”. For many people, it was their first job right out of college in the late 80s, 90s and early 2000s, and they are still there 30 or 40 years later!
My first manager at AMD retired after working there for more than 30 years, my second manager is still working there, and my super-manager has been working there for 40 years now.
I guess that’s how AMD has survived multiple near-death situations, and why I still say “we” when talking about AMD, like I’m still that big fanboy from college!
Now AMD engineers have something to say
The above is all that Hemant shared. I wonder what his family members felt after hearing what he said. The discussion among netizens was very heated.
Some netizens felt like reading a book, and thought these stories were very interesting and leaders could learn a lot from them:
The current AMD engineer mentioned at the beginning also shared his views on some issues.
For example, regarding what the brother said earlier, “…our first mistake, and AMD insiders may not like me saying this, was made here”:
I totally agree with this point of view.
I think K7/Athlon had a very competitive and superior core architecture, and K8/Opteron added better interconnect technology and moved to a NUMA architecture. Opteron's huge success also coincided with the rapid growth of Google's data centers.
It’s worth noting that for a small company like AMD, developing a very expensive processor core and then having to cancel it, coupled with the TLB bug in Barcelona, these issues set us back several years. We began a long effort to change our execution strategy, which we finally achieved in the Zen architecture.
One explanation for why Huang almost became CEO is that ATI was a better cultural fit. I've heard that Nvidia's corporate culture changed sometime in the last decade (if anyone has more accurate information, please correct me).
I have mentioned before that after the ATI merger, the company focused on graphics processing and APUs, but in my opinion, a strategic mistake the company made at the time was to deprioritize CPU core improvements because benchmarking and pricing power were determined by the CPU.
In addition, before this, regarding AMD's acquisition of Nvidia, there was a saying in the industry that the market value of AMD and Nvidia was actually not much different at the time (Nvidia was about 7.5 billion, AMD was about 8.6 billion), and if one was not careful, AMD might be swallowed up by Nvidia (doge).
What do you think about the story told by this old man?
Reference Links:
[1]https://x.com/MohapatraHemant/status/1809135345683841050
[2]https://x.com/philparkbot/status/1809326295609930135























