Exclusive Interview with GlobalFoundries: Transforming Delivery Relationships into Co-creation Relationships

"The automotive industry is in a healthy but accelerated cycle, with chip content per vehicle still rising, and China's iteration speed has become the benchmark for the entire industry, with cost pressures transmitting upstream along the chain." September 22, GlobalFoundries (GlobalFoundries) Global Executive Vice President Faisal Saleem laid out his judgment on the automotive semiconductor business—from capacity responsibility to localization paths, from the landing point of Physical AI to the boundaries of customization. All these judgments ultimately point to one thing: GlobalFoundries does not want to be regarded merely as a wafer foundry.

Industry is healthy, but pace is accelerated

Faisal's judgment on the current situation is clear: the automotive industry overall is healthy.

After the pandemic, the industry went through a period of inventory clearance, and the pace slowed down at one point—he used a vivid metaphor to describe those days, as if there was a "speed bump" on the road. But slowing down does not mean stopping; the overall market is still moving forward steadily, and the more critical structural change is that the semiconductor value and chip quantity per vehicle have been rising. In his view, this momentum will continue in the foreseeable future.

The example he gave comes from the US market: autonomous ride-hailing taxis are gradually being deployed on a large scale; the chip quantity in these vehicles is significantly higher than in ordinary models—the smarter the car, the more chips it uses.

In the past five years, the number of vehicles equipped with GlobalFoundries chips has not changed much each year, staying at the level of 90 million, but the number of chips equipped on these vehicles continues to increase. Quantity hasn't changed, content has—this is probably the most straightforward footnote to the current way of automotive semiconductor growth.

"What really makes this business tense is not scale, but speed. And that speed is set by China." Faisal stated.

In Faisal's narrative, China has already been regarded as a benchmark market by the entire automotive market, the reason is simple—iteration speed here is very fast, and design cycles are being compressed constantly. The transmission path brought by this speed is clear: car manufacturers need to pack more functions in a shorter time, carrying more functional demand per unit cost; from the perspective of unit cost, costs are rising, and pressure moves upstream along the chain, landing on chip factories as higher price pressure.

In addition to internal industry pressure, changes in the external environment are also bringing new challenges to automotive semiconductors. In the past two years, the most prominent narrative in the chip industry was how AI data centers absorb capacity, with automotive seen as one of the bearers.

What Faisal repeatedly emphasized was another thing: responsibility.

"We absolutely do not want this situation to occur—because GlobalFoundries cannot fulfill commitments, leading to customer production lines shutting down."

The weight of this sentence lies in transforming the problem from capacity arithmetic to a credit issue. His approach is early communication, early investment, keeping visibility and transparency between supply and demand sides, ensuring supply continuity.

"We cannot adopt a passive approach, waiting until supply shortages or production line stoppages occur before taking action; instead, we must take a step forward and act proactively." Faisal pointed out.

Supporting this practice is a substantial shift in cooperation methods over the past five years. In the past, once capacity tightened, the common action of car manufacturers and tier-one suppliers was to place more orders and maintain larger inventory; now information between both sides is more transparent—car manufacturers communicate the pace of launching new cars in advance, and GlobalFoundries tells them when capacity comes online and how much can be delivered when supply tightens.

"At least among the clients we have contacted, no one solves problems by increasing inventory; the solution has been changed to mutual honesty and increased visibility." Faisal introduced.

Automotive is the first large-scale sample of Physical AI

"Looking from both capacity and technology, the real variable is the moment AI steps out of the cloud."

Faisal's definition of Physical AI is clear: AI moving from the cloud to machines in the real world. And in GlobalFoundries' framework, it is composed of four closed loops—sensing, thinking, acting, communicating.

Using these four to fit, automotive is almost a perfect sample. It is a machine in the real world that you can touch with your hand, capable of sensing, thinking, acting, and communicating. Communication in the car is divided into three layers: many components inside the car communicate with each other, the car communicates with the driver and passengers, and the car also communicates with the cloud to complete remote upgrades.

This is also why he judges that automotive will become the first large-scale deployment scenario for Physical AI. The capabilities being integrated in the car are highly similar to humanoid robots—as long as time passes, the same functions will also migrate to robots. He does not hide that humanoid robots still have uncertainty and are currently at the initial stage; but the logic is inferable: if a product combination can serve the automotive market well, it is highly likely to also serve robots.

When it comes to specific technical landing points, the list is quite long: sensing side is radar, lidar, camera, ultrasonic, and cockpit sensors, plus ISP platforms responsible for image processing; data needs to move, requiring Ethernet, SerDes, and power management chips; thinking side is domain controllers, zone controllers, and MCU; acting side is motor control—motor control itself is not new, using AI to drive it is the new thing.

There is a premise easily ignored here: no single process can handle all requirements. Chips needed for mmWave radar transceivers are completely different from chips responsible for battery monitoring and lifecycle management. GlobalFoundries' response is a combination of dedicated technology platforms serving their respective scenarios.

Electrification is another main line. China already has more than half of the cars being some type of electric vehicle, and behind the eye-catching indicators of battery, range, and fast charging, what really happens is a fundamental change in the overall vehicle power architecture.

GlobalFoundries believes that the evolution of automotive to 48-volt is inevitable and must happen, but it is currently at the early stage, the entire conversion needs time, needs to happen step by step."

The non-propulsion power usage on the car is expanding rapidly—ADAS systems, central CPUs, and zone controllers are all consuming more power; if 12-volt solutions continue to be used, they must carry very large current, and large current inevitably brings high loss. But 12-volt architecture has been used by car manufacturers for decades, conversion paces for different manufacturers and models are not consistent, migration can only happen step by step.

There is also something often harder to replace than technology. When customers buy automotive chips, they never just buy technology, but also quality assurance and consistency. A car may run for 10, 15, or even 20 years, and must work under extreme temperatures. GlobalFoundries' answer is the AutoPro platform, embedding the highest quality standards into the entire process from design, process, manufacturing, PDK, IP to manufacturing operations.

Localization is not building factories, but a partner network

In 2025, GlobalFoundries reached an automotive-grade chip manufacturing cooperation with Guangzhou Zengxin Technology. There are various interpretations of this cooperation by the outside world, but its business model is actually quite intuitive: GlobalFoundries provides process technology and professional knowledge in the automotive manufacturing field—this is behind the accumulation of global manufacturing for over ten years; Zengxin Technology provides capacity and local infrastructure.

Why choose cooperation instead of building a factory in China yourself?

"Our ultimate goal is not to own all assets, as long as we can serve customers well." Faisal thinks.

This answer makes GlobalFoundries' true thoughts very clear. In its view, if there is a sufficiently strong partner network in the ecosystem that can help it serve customers well, the goal has been achieved. Zengxin Technology's value is not just capacity, but also local experience, local infrastructure, and understanding of the surrounding environment and talents.

When it comes to customers, the freedom brought by this arrangement is specific: customers design a product, it can be produced with European raw materials at GlobalFoundries' European factory, or brought to China for production by GlobalFoundries' Chinese manufacturing partners, both bases can place orders and ship.

GlobalFoundries has entered China for 12 years, established an office in Beijing last year, and established a Guangzhou office in early September this year, plus places like Shanghai, Wuhan, etc., resource networks are being spread. The Guangzhou office is regarded as a key step to support China's production ecosystem—it is closer to customers and can provide support directly from the production link.

And in the matter of going global, what GlobalFoundries can provide is supply chain resilience. Faisal splits it into different levels.

First is the breadth of manufacturing layout. GlobalFoundries has production bases in China, Singapore, Europe, and the US. But he points out immediately that having a global layout is not enough, peers also possess this point.

Second is the consistency of technology and products. The same product can be produced in Singapore or Dresden, Germany—when customers have demand, they do not need to start designing and verifying from scratch. If a product is verified only in Singapore, US, Germany, not in China, then when Chinese customers find GlobalFoundries, they still cannot get it.

Faisal also mentioned: "Geopolitical risks cannot be completely eliminated by setting up factories everywhere. What GlobalFoundries can provide is diversity—more choices, helping customers reduce risks as much as possible."

And this year's strong growth in China's automotive exports gives this global layout a new place to be used. Faisal judges this trend will continue, car manufacturers and tier-one suppliers want to get the same quality products in different places, use local products to serve local customers, and get equivalent local technical team support.

Boundaries of Customization

Looking back at the timeline, what GlobalFoundries is actually rewriting is its own identity. From doing pure wafer foundry, turning into becoming a customer's technical partner—not just manufacturing wafers, but providing comprehensive technical solutions.

The boost for this transformation is acquisition. The joining of MIPS and the merging of ARC IP, let GlobalFoundries' capability boundaries be expanded, more importantly, it largely moved the contact point with customers forward earlier: participate in innovation process earlier, conduct technical evaluation earlier. In Faisal's words, their communication with customers was advanced—previously did not have this condition.

And the place this new identity is most easily misread, is exactly customization. Semiconductors are a business about economies of scale, customization once too fine-grained, input-output ratio will be out of balance.

Faisal corrects this saying: "Customization doesn't mean we need to customize every chip, but in the whole process, we need to reduce waste."

Examples are extremely specific: customers may not need that many interfaces, reduce interfaces in design; may not need that much memory, reduce memory; may not use that many CPU cores, cut them off. In the end, customization must comply with economic benefits to do.

When is customization needed? The conditions he gives are only two: one is it can reduce or eliminate system-level waste, two is it can achieve differentiation competition. Satisfy one of them is enough, ideal state is both.

These two conditions exactly correspond to two realities of the current China auto market: players entering the car making track are much more than ten years ago, competition is fiercer, differentiation is more important; price wars are getting fiercer, no car company can still afford system-level waste. One side to stand out, one side to compress waste—in this crack, customization is established.

Earlier intervention also means undertaking risk earlier, and GlobalFoundries' algorithm for this thing is not quite different. In Faisal's view, early communication itself is a way to build customer relationships, spending time and energy finally not winning business, can be fully accepted—this process builds trust, and trust is very valuable in business. Maintaining early contact with hundreds or thousands of customers can also let GlobalFoundries see product trends, market trends earlier, and what customers really care about.

Design, process, manufacturing must finally be integrated together, only by disclosing to each other from the beginning and lowering thresholds, can find suitable solutions for both sides. His conclusion is, early intervention amplified business, not the opposite—it turns GlobalFoundries and customer relationships from transactional cooperation to strategic cooperation.

As for the biggest variable in the next few years, Faisal thinks AI defines cars, energy technology route changes, geopolitical risks, each trend will continue to evolve, just speed different.

What he cares more about is GlobalFoundries' position in this picture. In the next five years, this company hopes to be regarded as a trustworthy technical partner, rather than a wafer foundry manufacturing company; hopes people think of GlobalFoundries when talking about automotive industry, think of automotive industry when talking about GlobalFoundries.

What this company is doing—acquiring IP, promoting customization, laying out global capacity, intervening customer communication earlier—looks like four different actions, landing on the automotive this market that emphasizes long cycle, emphasizes trust, actually point to one thing: transform delivery relationships into co-creation relationships.