Posted on Semiconductor Engineering: Click here to view original article
By: John Kibarian
After more than three decades navigating the complexities of chip manufacturing and witnessing the transformation, I would like to share my perspective on how the industry changed and what it takes to succeed today. In my view, the future belongs to companies that can learn faster from smaller datasets while effectively collaborating with partners across the industry.
Let me explain. The semiconductor industry has always been rooted in collaboration; it draws on the best skills and capacities from around the world. Perhaps more striking than collaboration is the industry’s relentless cycle of creative destruction.
Consider this: In the 1990s, companies like NEC, Motorola and Digital Equipment Corporation were at the cutting edge of semiconductor innovation. Today, none of them are involved in semiconductors. It’s a sober reminder that even industry giants aren’t immune to disruption.
Now, with the AI revolution in full swing, we’re witnessing another “changing of the guard.” Government involvement has intensified as semiconductors have become critical to everything from national security to artificial intelligence.
History rhymes, leaders evolve, but the story stays the same and the stakes have never been higher.
Modern semiconductor manufacturing isn’t just complex, it’s mind-bogglingly sophisticated. A single chip takes about three months to manufacture and involves more than 1,000 steps. Companies routinely make $5 billion R&D bets and build $15-20 billion factories. In this environment, the gap between being top-notch and slightly off can be ruinous.
Just being second or third to market can be the difference between profit and loss. When making these kinds of investments, operational inefficiency is not an option. The manufacturing process is so intricate and long that if something goes wrong early in production, it might not be caught until the end, potentially wasting hundreds of millions of dollars.
Companies like PDF Solutions provide software that helps detect issues early by collecting and interpreting process data in real time, flagging unusual patterns before they cause widespread waste.
Our recently launched Sapience Manufacturing Hub Enterprise represents a new approach to managing semiconductor manufacturing data; one designed to remove human bottlenecks from AI-enabled production decisions. It also helps solve deceptively simple problems but potentially costly. Decisions such as where to send a chip next based on its test data or assessing the true manufacturing cost of a semiconductor lot.
With Sapience, engineering connects with enterprise resource planning and manufacturing execution systems, allowing AI to orchestrate workflows with limited manual input. This low-code integration layer enables autonomous decisions, routing chips based on demand, factory readiness, or end-market requirements.
Perhaps more important, a fundamental disconnect in semiconductor manufacturing companies is addressed. Engineering and manufacturing form the backbone of performance, yet critical business decisions often occur in isolation from the rich operational data these functions generate. By integrating granular engineering and manufacturing insights directly into enterprise-level decision-making, Sapience gives executives unprecedented visibility into production realities.
The power of global collaboration in semiconductor manufacturing cannot be overstated and is highlighted by our acquisition of secureWise. It now links 200 fabrication facilities globally to their equipment suppliers through a secure, private network. The value is clear when individual manufacturing machines can cost up to $300 million and generate enormous amounts of data.
The semiconductor industry is thriving and PDF Solutions’ revenue is growing. Our growth is happening in a context marked by three trends. The first is advanced packaging through the move to 3D chip packaging for higher density and faster signal transfer, requiring new tools to predict and solve yield issues. The next is geopolitical decentralization. The US, European, Japanese and Indian Chips Acts are decentralizing manufacturing, making remote collaboration crucial. The final trend is AI augmentation caused by the talent shortages across the industry. AI must augment human decision-making to maintain competitive advantage.
We sit at the intersection of these trends, built on a common platform and database to help customers advance, collaborate and scale AI without friction. And now, we’re working to educate the next generation.
I recently asked a room full of chip industry executives how many would recommend the chip industry to their children. No one raised their hand. That’s when I knew Moore’s Law was dead, not for technical reasons, but due to waning interest.
This realization sparked an educational push within PDF, a Carnegie Mellon University spinout, to embrace its academic roots rather than feeling “nerdy” about them. CMU partnered with PDF and Intel to teach AI in manufacturing, offer workshops and start publishing educational content.
Looking ahead, I see a fundamental shift coming to semiconductor manufacturing that I call learning from less. The traditional paradigm that “volume equals dominance” is being challenged by AI’s demand for the most advanced nodes without necessarily requiring high volumes. This vision represents more than just a business strategy; it’s a reimagining of how semiconductor manufacturing can adapt to serve the AI-driven future while building resilience through geographic diversification.
As the semiconductor industry continues its relentless evolution, let’s embrace collaboration. In complex manufacturing, working hand-in-hand with the ecosystem isn’t just helpful, it’s crucial for survival. Let’s also invest in real-time intelligence because the gap between success and failure often comes down to detecting problems before they compound. As much as possible, let’s remove human bottlenecks. As complexity increases, AI-enabled automation becomes essential for maintaining competitive timelines. And let’s think globally. Modern manufacturing requires seamless collaboration across continents and time zones.
Finally, let’s commit to educating the future workforce. Industry vitality depends on attracting and developing the next generation of talent.
The semiconductor industry’s story is far from over. As AI continues to drive demand for ever-more sophisticated chips, and as geopolitical forces reshape global supply chains, companies that can learn faster, collaborate more effectively, and manufacture more efficiently will write the next chapter on innovation and disruption.
As an aside: If you are attending SEMICON West 2025 and want to hear more of my perspective on the changing semiconductor industry, I’m giving a keynote presentation titled “Revolutionizing Semiconductor Collaboration: The Emergence of AI-Driven Industry Platforms.” It will be held Wednesday, October 8, at 10:20 a.m. in SEMICON West’s new location—the Phoenix Convention Center, Phoenix, AZ.