Preface:

Between 2025 and early 2026, the semiconductor IP industry experienced three remarkable transformations: Rambus' stock price nearly doubled, Synopsys divested its processor IP business ARC, and high-speed interconnect IP company Alphawave Semi was acquired by Qualcomm.

Amid the explosive demand for AI computing power, the value of the semiconductor IP industry is shifting from processor cores to interface and connectivity technologies.

Three Major Shifts Redefine the IP Market Landscape

The three major changes in the semiconductor IP industry over the past year have fully exposed the transformative impact of generative AI on underlying chip architectures.

Once controversial, Rambus has made a stunning comeback, with its stock price nearly doubling in 2025, becoming an indispensable key player in the AI server supply chain.

Meanwhile, EDA and IP giant Synopsys made a strategic adjustment by selling its star processor IP business, ARC, to foundry GlobalFoundries.

Alphawave Semi, which once vowed to disrupt the high-speed interconnect landscape, was ultimately fully acquired by Qualcomm by the end of 2025.

Three companies, three fates: one revalued by the market, one divested, and one wholly acquired.

These changes reflect the harshest truth of the semiconductor industry in the AI era: the throne of computing power is shifting.

According to IPnest, the total semiconductor design IP market reached $8.5 billion in 2024, with a year-on-year growth rate of 20%, the highest in history.

Over the past year, the top four companies have seen overall growth of around 25%, further driving industry consolidation.

However, behind the overall growth, companies in different segments are enjoying vastly different growth dividends.

From a business model perspective, the semiconductor IP industry primarily generates revenue through two methods: upfront licensing fees and royalty income. However, this traditional business model is now facing challenges amid the AI wave.

Rambus' Resurgence: The Rise of Interface IP

Rambus once had a controversial reputation in the semiconductor industry, labeled as a "patent troll" for collecting fees from major memory manufacturers through intensive patent litigation.

However, the advent of the AI era has enabled this company, deeply rooted in high-speed interface technologies, to make a stunning transformation.

By early 2026, Rambus' stock price had reached the $115-125 range, with a cumulative increase of nearly 100% over the past year.

Investment bank analysts have repeatedly raised their target prices, believing that Rambus has moved beyond a business model solely reliant on patent licensing, securing a more central position in AI and data center infrastructure.

The market has begun to reprice Rambus primarily because the biggest bottleneck in AI computing power is no longer the speed of GPUs but the inability to transmit data fast enough.

Rambus has seized three critical pain points: DDR5 RCD interface chips, HBM4 controller IP, and MRDIMM technology.

Against the backdrop of AI servers fully transitioning to DDR5 standards, Rambus' RCD chips have become indispensable components, with a market share exceeding 40%.

Meanwhile, as HBM becomes a key component of AI accelerator chips, Rambus' early introduction of HBM4 physical layer and controller IP makes it difficult for any company aiming to develop AI accelerators to bypass its licensing.

Rambus' 2025 financial report shows a staggering 42% growth in product revenue, driven primarily by demand for DDR5 memory interface chips in AI servers.

Its HBM3E/4 interface IP can provide throughput exceeding 1.2 TB/s, becoming a standard feature in AI accelerators.

The Decline of ARC: The Shrinking Market for General-Purpose Processor IP

In stark contrast to Synopsys' divestiture of its ARC processor business, ARC's exit represents the trend of democratization of general-purpose computing in the AI era.

ARC processors were once leaders in the embedded space, consistently ranking among the top global IP shipments.

In the era of microcontrollers and early SoCs, their low power consumption and customizability were their killer features.

However, under the pressure of generative AI, the efficiency of this "jack-of-all-trades" architecture appears inadequate.

In AI chips, 90% of transistors are allocated to tensor processing units and caches.

Traditional CPUs have degraded from decision-making cores to "foremen" responsible for task scheduling and I/O processing.

When the "brain" is no longer the performance bottleneck, customers' willingness to pay for general-purpose CPU IP begins to decline.

What truly suffocates closed-source processor IP like ARC is the architectural democratization brought by RISC-V.

AI vendors need to customize instruction sets based on their own operators, and RISC-V's open-source nature allows developers to freely add matrix extensions without paying hefty architecture licensing fees.

Synopsys realized that instead of struggling in a red ocean, it would be better to "make way for newcomers" and tap into the new infrastructure of AI.

It has redirected its capital and R&D efforts toward AI-enhanced EDA and its mega-merger with simulation giant Ansys.

Synopsys plans to enable chip companies to simulate heat dissipation, stress, and signal integrity across tens of thousands of compute nodes in the digital world before chips are even taped out.

The value in the IP market is shifting from cores to peripherals. In the past, everyone bought CPU cores; now, everyone is vying for protocols that connect these cores, interfaces that solve data throughput, and simulation tools that optimize overall power consumption.

The Sale of Alphawave: The Dilemma of Mid-Sized IP Vendors

Alphawave Semi, initially known as Alphawave IP, was a pure IP supplier whose core product was SerDes technology.

This technology enables ultra-high-speed data transmission within and outside chips, earning the company the reputation of being the "Lamborghini of interface IP."

In 2022, the company rebranded as Alphawave Semi, dropping the "IP" from its name. This marked its ambition to move beyond just selling blueprints.

By acquiring SiFive's custom chip business, OpenFive, it gained the capability to design complete SoCs. However, this strategy failed to keep it independent.

Qualcomm completed its acquisition of Alphawave Semi by the end of 2025, making it a core connectivity technology department for Qualcomm's foray into data centers and AI infrastructure.

This acquisition holds significant value for Qualcomm in two main areas: UCIe chiplet interconnect technology and 224G SerDes capabilities.

AI vendors assemble different functional units using Chiplet technology, and the core "glue" is Alphawave's expertise in UCIe technology.

Meanwhile, as AI model sizes explode, switch speeds are evolving from 800G to 1.6T, making 224G SerDes the ticket for AI switches to enter the 1.6T era.

Alphawave's disappearance proves the high barriers to entry in high-speed interconnect IP. In the arms race of the AI era, mid-sized, technologically singular IP companies are gradually losing their ability to survive independently and must become the sharpest spears in the hands of vertical giants.

Industry Value Restructuring: From Computing to Connectivity

Over the past few decades of computer architecture evolution, from CPU clock speed competition to multi-core architectures and then to GPU parallel computing, the industry has long followed a "computing-centric" paradigm.

Whoever possesses a stronger computing core holds the ceiling on system performance. However, the emergence of generative AI is shaking this premise.

Under large model workloads, the system's performance function is undergoing structural changes.

In the past, system performance was approximately equal to computing power; now, system performance is determined by the slowest link among [computing power, memory bandwidth, interconnect bandwidth, system latency, and energy efficiency constraints].

As process technologies, EDA toolchains, compiler technologies, and software stacks mature, more vendors are capable of designing usable AI accelerators, and the technological barriers to single computing units are declining.

This does not mean computing is unimportant; rather, it means computing power is transforming from a "decisive competitive advantage" to an "infrastructure-type capability."

As computing power becomes infrastructure-like, the new center of power is shifting to data pathways.

In AI systems, data flow paths include memory pathways between processors and HBM/DDR, interconnects between dies, communication between chiplets, board-level high-speed SerDes, and network switching between nodes.

Once these pathways become bottlenecks, even the strongest computing cores can only idle.

Therefore, companies that master key technologies in data pathways are beginning to occupy high positions in the system power structure.

According to IPnest's monitoring, the semiconductor IP market is undergoing structural reorganization. The market share of processor IP has continuously declined from 57.6% in 2017 to less than 45% in 2025.

Meanwhile, the share of interface IP has against the trend grown from 18%, and it is expected to account for more than a quarter of the entire IP market by 2026.

While the traditional IP market maintains conventional growth of 8%-10%, interface IP segments represented by high-speed SerDes, PCIe 6.0/7.0, and HBM controllers are surging at a compound annual growth rate exceeding 20%.

In the AI era, the value of general-purpose CPU IP is declining, while specialized IP will become the mainstream.

Single IP products can no longer meet customer needs; customers require complete solutions, including IP, EDA tools, and software support.

Therefore, vendors with strong ecosystems will occupy a dominant position.

IP Giants Hold the Shield, Newcomers Break Through

The semiconductor IP industry has strong technological and ecological barriers. Leading vendors like Synopsys, with decades of technological accumulation, comprehensive product matrices, and powerful ecological layouts, will continue to dominate the industry and are unlikely to be disrupted in the short term.

Established giants like Synopsys still possess broad layouts and deep service capabilities across the overall IP ecosystem.

However, in new segments such as high-speed interconnects, AI on-chip collaboration, and new security systems, newcomers like Alphawave have significant explosive potential.

IP R&D in the AI era requires deep technological accumulation and continuous capital investment, making it difficult for emerging vendors to achieve comprehensive transcend in the short term.

Moreover, the synergistic ecosystem of EDA+IP and the deep integration of IP with foundries require long-term accumulation and are difficult to replicate quickly.

IP is no longer just a module seller but a provider of systemic collaboration solutions. AI chip demand reflects collaborative bottlenecks across the entire system rather than the performance of a single interface or core.

Future IP suppliers will need not only leading module performance, power and area optimization but also cross-module performance collaboration, system-level verification and support, and upfront customization capabilities integrated into chip architecture design.

Focusing on niche segments can quickly accumulate leading advantages, as demonstrated by Alphawave's explosion in the high-speed interconnect segment, reflecting the value of a focused strategy.

In future multi-core and Chiplet architectures, high bandwidth and efficient interconnects will become core performance bottlenecks.

Emerging IP companies that can establish leading barriers in single high-growth segments will gain faster growth space.

However, this model also has limitations in lacking horizontal extensibility, similar to AI's own development—specialized accelerators can explode, but systemic coordination remains indispensable.

In the AI era, there are no "eternal IP giants," only niche kings who ride the wave correctly.

The paradigm shift in semiconductor IP is happening, but it's not an overall disruption; rather, it's a restructuring along memory and interconnects.

Over the next five years, the real winners will be those who can deeply bind with leading AI customers while mastering advanced node interface and memory technologies.

Conclusion:

The greatest transformation in the AI era is not singular computing power growth but the increasing complexity of invisible data movement and collaboration.

The role of IP is shifting from peripheral functional modules to core collaborative capabilities, and this transformation process is accelerating.

How to balance full-category coverage with breakthroughs in focused segments will become a long-term proposition for IP companies.

Partial References: Semiconductor Industry Review: "AI Thrives, Interface IP Market Grows 23.5%", Sina Tech: "2026 AI Chip IP Supplier Authority Recommendations: Top Ten Core Choices for the Edge Intelligence Era", Semiconductor Industry Observations: "The Semiconductor IP Market Has Changed!", Chip World: "Top 10 Semiconductor Market Focus Points in 2026", "EDA/IP and Foundry Industry in the AI Era"