On the evening of September 16, the semiconductor industry announced a major acquisition. Winbond Electronics, a memory manufacturer from China's Taiwan region, held a press conference for significant information. The board of directors of Winbond Electronics passed a resolution to sign a share purchase agreement with Infineon Technologies LLC, a subsidiary of Infineon, to acquire 100% equity of its NOR Flash and F-RAM (Ferroelectric Random Access Memory) business in an all-cash transaction. The base transaction price is USD 1.12 billion, approximately TWD 35.392 billion and CNY 8 billion. The transaction is calculated on a cash-free and debt-free basis and is expected to be completed in the second half of 2027, subject to approvals from relevant regulatory authorities.
What is the strategy behind the "global number one" acquiring the "global number four"? Shifting the focus away from market share reveals that what this transaction truly acquires is a long-dormant brand, a fully established overseas team, and a ticket to the high-barrier automotive and industrial markets.
The Full Picture of the Deal: Cash, Equity, and the Revived Spansion
The transaction structure is not complex. Winbond is acquiring 100% equity of Infineon's restructured NOR Flash and F-RAM business, settled entirely in cash with a base consideration of USD 1.12 billion, which will be adjusted according to the relevant provisions of the share purchase agreement prior to closing. The transaction has been approved by Winbond's board of directors and is expected to be completed in the second half of 2027, subject to necessary regulatory approvals.
Looking at the target's composition, the two types of products have clear divisions of labor. NOR Flash primarily handles system boot code and firmware storage, featuring fast read speeds and high reliability; F-RAM (Ferroelectric Random Access Memory) excels in fast, frequent writing and retains data after power loss, making it widely used in scenarios such as critical state logging. The customers for both are highly concentrated in the automotive, industrial, and infrastructure sectors. This is a typical high-barrier, long-cycle market: product validation is slow, and once integrated into a customer's design, the supply relationship often lasts for many years.
Even more noteworthy are the geographical and brand arrangements of the assets. The target company is headquartered in San Jose, California (Silicon Valley), USA, and has been deeply engaged in the memory solutions field for over three decades. Upon completion of the transaction, it will operate as an independent entity with its headquarters unchanged, and the "Spansion" brand will be revived. In a statement, Winbond's Chairman and CEO Jiao Youjun stated that Spansion was once a pioneer in the global memory industry, achieving multiple technological and innovative breakthroughs, "We are deeply honored to carry forward this mission once again under the name of Spansion."
Infineon is not "exiting the memory business." The company explicitly retains its SRAM, HYPERRAM, nvSRAM, and radiation-hardened memory product portfolio based on SONOS technology. These solutions will continue to serve the automotive, industrial, infrastructure, aerospace, and defense markets, which also include high-reliability memory products provided for space missions. In other words, what Infineon is selling are the highly standardized NOR Flash and F-RAM products, while what it keeps are the niche memories more tightly coupled with its power and automotive mainlines.
At the funding level, Winbond's board of directors on the same day also resolved to issue the fourth and fifth unsecured convertible corporate bonds, with maximum total face values of TWD 14 billion and TWD 4 billion respectively, totaling up to TWD 18 billion, for capacity expansion, advanced process R&D, and production equipment procurement. Coupled with the established capital expenditure plans, this acquisition is not a "lightly loaded" endeavor.
The Ledgers of Both Buyers and Sellers: A "Position Swap" for Mutual Benefit
For Infineon, this is a proactive subtraction. The company's revenue for fiscal year 2025 (ending in September) is approximately EUR 14.7 billion, with a global workforce of about 57,000, and its business focus is continuously shifting towards power semiconductors, automotive electronics, and energy-related fields. The demand for NOR Flash is stable, but the products are highly standardized with limited price elasticity. For a company that needs to tilt its resources towards core mainlines, maintaining an entire NOR Flash and F-RAM business system is not highly capital-efficient.
The statement from Peter Schiefer, President of Infineon Technologies' Automotive Electronics division, was quite direct: this transaction is "a highly attractive strategic outcome for all stakeholders," allowing Infineon to further focus its product portfolio and capital allocation on core growth drivers, while enabling the original NOR Flash and F-RAM business to "fully unlock its potential." Translated plainly, it means: it is not that this business is bad, but keeping it within Infineon might not be the most suitable combination.
For Winbond, this is a typical addition. There is an easily overlooked contrast here—Winbond itself is the global number one in NOR Flash market share. Since it is already the top player, why spend USD 1.12 billion? The answer is that it is not buying capacity or market share, but three things: position, time, and trust.
First is customers and certifications. Automotive and industrial customers have stringent requirements for memory reliability, product validation, and supply cycles; it often takes over a year or two for a product to go from introduction to mass production. Infineon's original NOR Flash and F-RAM business has already secured an established position in the supply chains of top-tier customers in Europe and the Americas. After the handover, Winbond essentially gains direct access to a proven business path, bypassing the lengthy process of building overseas channels from scratch and spending years on customer certifications.
Second is products and teams. With memory as its core business, Winbond owns its 12-inch wafer fab and has mature memory manufacturing experience; the target's inherent R&D system and differentiated product portfolio perfectly complement this. Combining the two, Winbond can offer a more comprehensive product line externally and absorb talent and supply chain partners from over a dozen countries through the acquisition.
Third is brand and technological lineage. Reviving Spansion means taking over not just a trademark, but a legacy of customer trust and technological reputation spanning three decades. For the community of engineers in Europe and the Americas who still have a historical recognition of this brand, this layer of trust is itself a channel.
Industry Impact: The "Oligopoly Rebalancing" of NOR Flash
The most direct consequence of this transaction is the reshuffling of the global NOR Flash rankings. According to 2024 market data, Winbond, GigaDevice, and Macronix have collectively captured about two-thirds of the global market share, with Winbond at approximately 27%, GigaDevice at about 23%, and Macronix at around 16%, while Infineon holds about 10%. Institutional estimates suggest that if the transaction is successfully completed, Winbond's NOR Flash market share will approach 40%, firmly securing the global number one spot, and by leveraging Infineon's original automotive and industrial control customer base, it will leap to become the leader in automotive NOR Flash. This also means that memory manufacturers from China's Taiwan region are further widening the gap with latecomers in the niche track.
But market share is just the surface. What is more noteworthy is that the industrial status of NOR Flash is being redefined by AI. It has long lived in the shadow of DRAM and NAND—unable to scale up capacity or raise unit prices, and constantly squeezed for survival space by NAND. However, the construction of AI infrastructure is pushing this "supporting actor" to the forefront. According to TrendForce's calculations, the demand for NOR Flash in a single AI server is about 3 to 5 times that of a traditional server: the densely populated Retimer chips inside the server each require an independent NOR Flash to load microcode; GPUs and communication equipment also need it to host boot and operational basic programs. The denser the AI computing power, the higher the usage of such chips, with a clear preference for high-capacity, Octal SPI, and other mid-to-high-end specifications.
The transmission to the terminal side is equally rapid. As inference capabilities sink into AI smartphones, AI PCs, and robots, the size of built-in models and firmware in devices multiplies, directly driving up the demand for high-density NOR Flash. However, the supply side is almost "a pool of stagnant water": after multiple rounds of reshuffling, there are only a handful of players in the track, and the addition of new bit shipments is extremely restrained. The result is an out-of-control surge in prices—in the first half of 2026, the average cumulative increase in NOR Flash contract prices reached 100% to 120%; TrendForce estimates that in the second half of the year, supported by AI server and automotive demand, high-capacity products of 256Mb and above still have an average price upside of 90% to 110%.
At such a juncture, trading cash for time is a rational choice. Rather than building an overseas system from scratch and spending years on customer certifications, it is better to directly acquire a mature team already embedded in the supply chains of European and American customers. This path also circumvents some geopolitical sensitivities—the target is headquartered in the US and operates as an independent entity, nominally remaining a memory company with "American lineage."
Of course, the execution risks of the transaction cannot be ignored. The closing will not happen until the second half of 2027 and requires regulatory approval; the target's business spans over a dozen countries, and the integration of brands, customers, and R&D tests Winbond's organizational capabilities; more critically, whether Spansion can truly be "revived" depends on whether Winbond can transform this asset into sustainable pricing power before the AI demand cycle peaks. At least from the current perspective, this transaction has already thrown a weighty stone into the global NOR Flash market.
What Remains for Mainland Manufacturers is a Time-Bound Window
Shifting the focus back to the mainland region, the reason this acquisition has attracted widespread attention is that it marks the current competitive dimensions of this track.
The echelon of mainland NOR Flash manufacturers has become clearly defined. GigaDevice firmly holds the second place globally and the first on the mainland, with a global market share of about 20% for automotive-grade NOR Flash, and its products cover most domestic independent automakers; Puya Semiconductors, Dosilicon, and Zbit Semiconductor are ramping up shipments in the low-to-medium capacity market, where products of 128Mb and below face particularly fierce price competition due to the concentrated release of domestic output; Ingenic Semiconductor secured the global number one spot in automotive-grade SRAM through the acquisition of ISSI; CXMT and YMTC have each opened up space in DRAM and 3D NAND. Yet, during the "chip shortage" in 2021, the domestic self-sufficiency rate for automotive-grade NOR Flash was less than 5%.
The USD 1.12 billion spent by Winbond is not for capacity, but for a system already validated by European and American customers. Automotive-grade NOR Flash must pass AEC-Q100 certification, undergoing thousands of hours of continuous testing under extreme temperature, humidity, and vibration conditions; it typically takes a year and a half to two years from introduction to mass production. During this cycle, what chip companies truly hand over is a reliability track record, not a quotation sheet. Mainland manufacturers have pushed capacity and cost to the extreme in recent years, but in such high-threshold scenarios, price has never been the customer's priority.
Another variable is time. Samsung, SK Hynix, and Micron have shifted 70% to 80% of their advanced capacity towards HBM and AI servers, continuously compressing the supply of automotive-grade and industrial control products on mature nodes, stretching lead times to over 20 weeks, with shortages expected to last at least until 2027. This gap is a golden period for mainland manufacturers to complete certifications and bind downstream clients, and it is also a time-bound window—once overseas giants free up their hands to adjust capacity, the difficulty of substitution will rise again. It is also important to see clearly that what they are giving up are the mature node categories with thin margins, while high-value-added automotive-grade and high-end products remain firmly in their hands.
The market itself is also stratifying. Low-to-medium capacity products of 128Mb and below face white-hot competition due to the concentrated release of output from mainland manufacturers, while high-capacity products above 256Mb are driven by AI servers and vehicle digital cockpits, remaining in long-term short supply. Beyond holding the basic market for low-to-medium capacities, pushing resources towards high capacity, Octal/xSPI, automotive-grade industrial control, and communication base stations, and building barriers with process and certifications rather than price, might be more meaningful than fighting another round of price wars. Winbond's transaction actually illustrates only one thing: in this market, the name written into the customer's bill of materials is more valuable than the shipment volume ranking.
This article does not constitute any investment advice. If you have any questions, please contact us via WeChat: xsychief.