Chiplets

Chiplets are the semiconductor world's answer to LEGO blocks. Instead of designing and manufacturing a single, massive, complex silicon chip (known as a monolithic design), companies can now create smaller, specialized functional blocks called “chiplets” or “tiles.” These individual chiplets are then pieced together and interconnected on a shared base, or substrate, to form a single, powerful processor. Imagine building a high-end computer processor: instead of carving the entire intricate design out of one giant piece of marble and hoping it doesn't crack, you craft smaller, perfect marble tiles—one for processing, one for memory, one for graphics—and assemble them seamlessly. This approach dramatically improves manufacturing success rates (yields), lowers costs, and gives designers unprecedented flexibility to mix and match the best technologies for a specific task. For investors, this isn't just a technical detail; it's a fundamental shift in how the most valuable components of our digital world are built, creating new winners and losers across the entire tech industry.

For decades, progress in computing was defined by Moore's Law, the observation that the number of transistors on a chip doubled about every two years. To keep this pace, chips had to get bigger and more complex. However, this “monolithic” approach is hitting a wall, creating an opportunity for a new paradigm—the chiplet—to take over.

Think of a traditional, monolithic chip as a master chef trying to cook a complex seven-course meal all in one giant pan.

• High Risk of Failure: If even one part of a large, single chip has a defect, the entire chip is often useless. The larger the chip, the higher the probability of a defect, leading to lower yields and wasted silicon. This is like burning one part of the meal and having to throw the whole pan away.
• Inflexible Design: Every component on the chip must be manufactured using the same, most advanced (and most expensive) production process, even if some parts don't need it. This is inefficient and drives up costs enormously.
• Slowing Innovation: Designing these massive, all-in-one chips is incredibly expensive and time-consuming, slowing down the pace of new product introductions.

The chiplet model breaks up the problem. Instead of one giant pan, the chef uses several specialized pots and pans. This has profound economic and strategic advantages.

By breaking a large design into smaller chiplets, the manufacturing yield for each small piece is much higher. If one chiplet is defective, only that small piece is discarded, not the entire processor. This drastically lowers the overall cost of production, directly boosting the gross margins of semiconductor companies. It's a classic value investing principle in action: finding a more efficient, lower-cost way to produce a high-value product.

Chiplets allow for a “mix-and-match” strategy. A company can combine a high-performance CPU chiplet made with cutting-edge technology with an I/O (input/output) chiplet made on an older, more mature, and cheaper manufacturing process. This has two key benefits:

• Faster Time-to-Market: Companies can update specific chiplets (e.g., the CPU core) without having to redesign the entire processor, allowing them to bring new innovations to market much faster.
• Customization: It becomes easier to create specialized processors for different markets (e.g., data centers, gaming, mobile) by combining different chiplets, much like using different LEGO pieces to build a car versus a spaceship.

The shift to chiplets is redrawing the map of the semiconductor industry. For the savvy investor, understanding who benefits from this change is crucial.

The rise of chiplets creates value across the supply chain, but some players are better positioned than others.

• Chip Designers (Fabless & IDMs): Companies that design processors are the most obvious beneficiaries. AMD was a pioneer, using chiplets to power its Ryzen and EPYC processors and gain significant market share from Intel. Intel is now aggressively adopting a chiplet strategy to compete. Other giants like Nvidia and Apple are also leveraging the technology for their advanced chips.
• Foundries: The factories that manufacture the chips, like TSMC and Samsung, are indispensable. They not only produce the individual chiplets but also provide the critical “advanced packaging” technologies needed to connect them with high speed and low power. This creates a strong, sticky demand for their most advanced services.
• EDA and IP Providers: The “picks and shovels” of the industry. Electronic Design Automation (EDA) companies like Synopsys and Cadence Design Systems provide the complex software tools required to design and verify these multi-chiplet systems. Intellectual Property (IP) providers like Arm Holdings license out pre-made designs that can be used as individual chiplets, accelerating the design process for everyone.
• Packaging & Test (OSATs): Once the chiplets are made, they need to be assembled and tested. Specialized Outsourced Semiconductor Assembly and Test (OSAT) companies like Amkor Technology play a vital role in this final, crucial step.

While the trend is powerful, investors should be aware of the challenges.

• Standardization: For chiplets from different companies to work together, they need a common language. The industry is converging around standards like the Universal Chiplet Interconnect Express (UCIe), but broad adoption will take time. Companies with proprietary interconnects may create walled gardens, but industry-wide standards will unlock the most value.
• Technical Complexity: Connecting multiple chips to act as one is an immense engineering challenge. Issues like heat management, signal integrity, and software complexity can create new bottlenecks. The companies that master this integration will build a formidable competitive moat.