Tesla Expands Samsung AI Chip Production to 40,000 Wafers Per Month: What It Means for FSD | Taha Abbasi

Tesla has reportedly requested a massive expansion of its AI chip production at Samsung’s facilities, seeking an additional 24,000 wafers per month that would bring total production capacity to roughly 40,000 wafers. Taha Abbasi examines what this expansion means for Tesla’s autonomous driving ambitions, its AI training infrastructure, and the broader semiconductor supply chain.

The Scale of Tesla’s AI Chip Appetite

The numbers are staggering. If Tesla secures the additional 24,000 wafers per month it has reportedly requested from Samsung, the total monthly production of 40,000 wafers would make Tesla one of the largest consumers of advanced semiconductor manufacturing capacity in the world. Each wafer contains multiple chips, meaning the actual chip output could number in the hundreds of thousands per month depending on die size and yield rates.

Tesla uses its custom-designed AI chips, currently known as Hardware 5 (HW5), in both its vehicles and its Dojo training supercomputer. The vehicle-side chips power the FSD computer that processes camera feeds and makes driving decisions in real time. The training-side chips power the massive neural network training operations that continuously improve the FSD system’s capabilities.

The requested expansion suggests that Tesla is preparing for a significant increase in both vehicle production with the latest AI hardware and a major scaling of its training infrastructure. Both of these are critical for Tesla’s autonomous driving timeline, and the chip supply has been one of the key bottlenecks limiting the pace of deployment.

Why Samsung and Not TSMC

Tesla’s choice of Samsung as its AI chip manufacturer is noteworthy because the majority of the world’s most advanced AI chips, including those from NVIDIA, AMD, and Apple, are manufactured by TSMC (Taiwan Semiconductor Manufacturing Company). TSMC is widely considered the industry leader in cutting-edge chip fabrication, with its most advanced process nodes offering superior performance and energy efficiency.

Tesla’s decision to work with Samsung likely reflects a combination of factors including manufacturing capacity availability, pricing, and strategic diversification. TSMC’s manufacturing capacity is heavily booked by its largest customers, and securing additional capacity at TSMC may have been more difficult or expensive than working with Samsung.

Taha Abbasi sees strategic logic in the Samsung partnership. “Tesla’s AI chips do not need to be on the absolute bleeding edge of semiconductor fabrication,” Abbasi explains. “They need to be good enough for the specific workload, available in volume, and competitively priced. Samsung offers all three of those things, and working with Samsung gives Tesla supply chain diversification that reduces risk.”

The AI Training Infrastructure Race

Tesla’s chip expansion cannot be understood in isolation. It is part of a broader arms race in AI computing infrastructure that involves every major technology company. Google has its TPU (Tensor Processing Unit) chips, Amazon has its Trainium and Inferentia chips, and Microsoft relies primarily on NVIDIA’s GPUs. Each company is trying to secure enough computing capacity to train and deploy increasingly large AI models.

What makes Tesla’s approach unique is that its AI chips serve a dual purpose. The same fundamental architecture that trains the FSD neural network also runs inference in the vehicle. This vertical integration, where the company designs the chip, builds the training infrastructure, collects the training data from its fleet, and deploys the resulting model back into the fleet, gives Tesla a feedback loop that no other company in the autonomous driving space can replicate.

The 40,000 wafers per month target suggests that Tesla is thinking about AI chip supply not just for today’s needs but for a future that includes millions of vehicles running advanced AI, a fleet of robotaxis requiring the highest-performance inference hardware, and potentially a humanoid robot (Optimus) that would need its own AI chips for real-time decision making.

What This Means for FSD Deployment

One of the persistent questions about Tesla’s FSD timeline is whether the company can produce enough HW5 computers to equip its entire fleet. Currently, Tesla vehicles ship with HW4, and a retrofit program exists for owners who want to upgrade. The expansion of chip production at Samsung could accelerate the timeline for making HW5 widely available, both in new vehicles and as retrofits.

More computing power per vehicle means the FSD system can run more complex neural network models in real time, which directly translates to better driving performance. Each generation of Tesla’s AI hardware has enabled more sophisticated driving behavior, and HW5 is expected to provide the computational headroom necessary for fully autonomous driving without human supervision.

Taha Abbasi connects the chip production expansion to Tesla’s robotaxi ambitions. “You cannot launch a robotaxi service at scale if you cannot produce enough AI computers to equip the vehicles,” Abbasi notes. “Tesla’s push for 40,000 wafers per month at Samsung is a supply chain move that directly supports the Austin robotaxi expansion and future launches in other cities.”

Samsung’s Position in the AI Chip Market

For Samsung, the Tesla contract expansion represents a significant win in the competitive foundry market. Samsung has been investing heavily in advanced semiconductor manufacturing, including its Gate-All-Around (GAA) transistor technology and sub-3nm process nodes. Landing and expanding a high-volume customer like Tesla validates Samsung’s position as a credible alternative to TSMC for advanced AI chips.

The foundry business has become one of Samsung’s strategic priorities as the company seeks to reduce its dependence on memory chips, which are subject to cyclical boom-and-bust pricing. A stable, high-volume customer like Tesla provides predictable revenue and helps Samsung achieve the manufacturing volume needed to improve yields and reduce per-chip costs.

However, Samsung’s foundry operations have faced challenges in recent years, including lower yields on advanced process nodes compared to TSMC. If Tesla’s chip expansion pushes Samsung to improve its manufacturing processes, it could have beneficial knock-on effects for the entire semiconductor industry by strengthening the competitive alternative to TSMC.

The Geopolitical Dimension

Tesla’s chip production expansion also has geopolitical implications. Samsung’s foundries are located in South Korea, providing geographic diversification away from Taiwan, where TSMC’s primary facilities are located. Given ongoing tensions in the Taiwan Strait, diversifying semiconductor supply chains away from a single geographic risk point is a prudent strategy.

The US government has been encouraging domestic semiconductor manufacturing through the CHIPS Act, and both Samsung and TSMC are building new fabrication facilities in the United States. Tesla’s chip demand could eventually be served by Samsung’s planned facility in Taylor, Texas, which would keep the entire AI chip supply chain within the United States.

Taha Abbasi frames the expansion in strategic terms. “Tesla is building the AI infrastructure needed for autonomous driving, robotaxis, and humanoid robots,” Abbasi concludes. “Securing 40,000 wafers per month at Samsung is not just a procurement decision. It is a bet on vertical integration as the key competitive advantage in the AI era. Companies that control their own chip supply will move faster than those who depend on merchant chip vendors.”

For the autonomous driving industry, Tesla’s chip expansion sends a clear message: the company is investing billions of dollars in the hardware infrastructure needed to make fully autonomous driving a reality. Whether FSD achieves that goal in 2026, 2027, or later, Tesla is ensuring that chip supply will not be the bottleneck.

---

About the Author: Taha Abbasi is a technology executive, CTO, and applied frontier tech builder. Read more on Grokpedia | YouTube: The Brown Cowboy | tahaabbasi.com