Introduction
In the fast-evolving semiconductor and AI industry, timing and strategic partnerships can make or break a company’s position. One such pivotal moment took place in 2018 when Nvidia CEO Jensen Huang visited Korea with a vision to deepen collaboration with Samsung. However, internal turmoil within Samsung led to a rejection of Nvidia’s proposals, dramatically altering the course of the AI memory and GPU landscape.
This blog dives into how Samsung’s missed opportunity back then has shaped the current industry dynamics and what lessons it holds for semiconductor leaders today.
Overview: 5 Key Takeaways
Nvidia approached Samsung in 2018 with a strategic partnership proposal across memory, foundry, and software domains.
Samsung rejected all three proposals due to leadership instability and legal challenges at the time.
SK Hynix, another Korean memory giant, became Nvidia’s preferred partner for high-bandwidth memory (HBM), gaining significant market advantages.
TSMC, Taiwan’s leading foundry, solidified its position as Nvidia’s go-to chip manufacturer for advanced GPU nodes beyond 8nm.
Samsung, despite being an early leader, now finds itself playing catch-up in the high-performance memory and GPU foundry sectors.
The 2018 Meeting That Could Have Changed the Game
In 2018, Jensen Huang made a somewhat under-the-radar trip to South Korea, meeting with Samsung executives with a comprehensive three-part partnership proposal aimed at accelerating Nvidia’s AI hardware capabilities:
Deepen Co-Development of High Bandwidth Memory (HBM):
HBM is critical for AI and GPU performance, providing ultra-fast memory bandwidth with low power consumption.
Nvidia wanted to work closely with Samsung to co-develop next-generation HBM technology.
Collaborate on Advanced Foundry Technology Beyond 8nm:
Nvidia’s GPUs require cutting-edge semiconductor manufacturing processes. Huang sought to leverage Samsung’s foundry capabilities to produce advanced chips at nodes smaller than 8nm.
techovedas.com/18-billion-usd-south-korea-unveils-semiconductor-ecosystem-support-package
Joint Expansion of CUDA Software Ecosystem:
CUDA, Nvidia’s proprietary parallel computing platform and API, is essential for AI and compute workloads. The proposal included scaling CUDA support and integration in Samsung’s systems.
However, Samsung was going through a turbulent period in 2018. Its leadership was distracted by legal troubles involving the company’s top executives, including the high-profile arrest of former Chairwoman Lee Jae-yong over corruption charges.
These issues reportedly made Samsung’s leadership unable or unwilling to commit to long-term strategic discussions.
In Huang’s own words, “There was no one at Samsung to discuss long-term strategy with me.”
techovedas.com/how-intel-samsungs-struggles-affect-asmls-future
The Fallout: Who Benefited and Who Lost?
Samsung’s decision not to engage with Nvidia during this crucial moment had lasting consequences:
SK Hynix: Rising as Nvidia’s Memory Partner
SK Hynix stepped in as Nvidia’s primary partner for HBM supply. The company has since reaped enormous benefits, especially with the production of HBM3E memory modules used in Nvidia’s latest Blackwell GPUs.
- Market Gains: SK Hynix’s involvement in Nvidia’s HBM ecosystem has bolstered its revenues and solidified its position as a top memory supplier in AI and HPC (High-Performance Computing).
- Technological Leadership: Working closely with Nvidia has accelerated SK Hynix’s innovation trajectory in advanced memory technologies.
TSMC: The Foundry Champion
With Samsung unavailable for collaboration on advanced process nodes beyond 8nm, Nvidia turned to Taiwan Semiconductor Manufacturing Company (TSMC).
- Advanced Nodes: TSMC now manufactures Nvidia’s most advanced GPUs, utilizing cutting-edge 5nm and 3nm process technologies.
- Dominance in GPU Manufacturing: TSMC’s foundry services have become synonymous with high-performance GPU manufacturing, giving it a substantial competitive advantage.
techovedas.com/tsmc-taiwan-expansion-15-new-chip-fabs-amid-global-push-and-safety-setbacks
Samsung: A Leader Playing Catch-Up
Samsung’s absence in this partnership left it behind in two critical segments:
- Memory: Despite being an early HBM pioneer, Samsung now faces stiff competition from SK Hynix and others.
- Foundry: Samsung’s foundry business, although strong, has not secured Nvidia as a major client for advanced GPU manufacturing, limiting its growth potential in this sector.
Why Leadership Clarity and Vision Matter in Semiconductors
The 2018 episode highlights a broader lesson for semiconductor companies and tech executives:
- Crisis Management: Internal leadership crises and distractions can derail strategic opportunities, especially in industries that require long-term collaboration.
- Strategic Alignment: Partnerships in semiconductor and AI technology demand a shared vision and clear decision-making from the top.
- Timing is Everything: Delays or missed engagements can allow competitors to gain significant advantages that are difficult to overcome later.
This moment is a powerful reminder that semiconductor companies must cultivate strong, visionary leadership to navigate complex technology landscapes and global competition.
/techovedas.com/21-semiconductor-companies-fuelling-the-ai-revolution
The Broader Impact on the AI Memory and GPU Race
AI workloads continue to grow exponentially, driving demand for faster memory and more powerful GPUs. The stakes in this race are high:
- HBM Technology: High Bandwidth Memory is a key enabler of AI performance, reducing bottlenecks between processors and memory.
- Advanced Foundry Processes: Smaller, more efficient chip manufacturing nodes improve performance and energy efficiency — essential for large-scale AI models.
- Software Ecosystem: Software platforms like CUDA allow developers to fully exploit hardware capabilities, accelerating AI innovation.
Samsung’s 2018 missed moment with Nvidia arguably slowed its ability to capitalize on these trends. Meanwhile, SK Hynix and TSMC leveraged their partnerships to capture leading roles in the AI hardware supply chain.
What’s Next for Samsung?
Samsung is actively investing in both memory technology and foundry services to regain competitiveness:
- Memory Innovation: Samsung continues to develop HBM variants and other advanced memory technologies.
- Foundry Expansion: Samsung is pushing aggressively into advanced nodes like 3nm and beyond, aiming to attract more GPU and AI chip customers.
- Software and AI Ecosystem: Samsung is exploring software tools and AI ecosystem partnerships but still faces challenges catching up to Nvidia’s entrenched CUDA dominance.
The company’s success in these areas depends heavily on stable leadership and the ability to form strategic, forward-looking partnerships.
Follow us on Linkedin for everything around Semiconductors & AI
Conclusion
The 2018 meeting between Jensen Huang and Samsung executives serves as a powerful case study in semiconductor industry strategy. When leadership is distracted or fragmented, companies risk losing pivotal partnerships that shape technology leadership and market dominance.
For founders, executives, and industry watchers alike, Samsung’s missed moment reminds us that vision, clarity, and timely decision-making are as crucial as technological innovation. In the AI memory and GPU race, these factors often decide winners and losers.
