"Mergers Are Bound to Happen"

In July 2015, a small Chinese delegation flew into Boise, Idaho, and asked for a meeting with Mark Durcan. Durcan, the chief executive of Micron Technology, ran the last large American maker of memory chips, a company that had survived the Japanese DRAM siege of the 1980s and emerged as the only domestic producer of dynamic random-access memory at scale. The delegation came from Tsinghua Unigroup, the holding company controlled by China’s most prestigious science university, with a number on the table. Twenty-one dollars a share. About twenty-three billion dollars in cash. A nineteen percent premium to Micron’s previous Friday close. It would have been the largest takeover of an American company by a Chinese acquirer in history.

Durcan listened. He did not say yes. Within a week, word of the approach leaked to Reuters, then to Bloomberg, then to every wire service that covered chips. By the time Senator John McCain had read the headlines and tweeted that the deal would be a national security disaster, Micron’s board had already concluded what its lawyers were telling it: a Chinese state-backed bid for the only American DRAM maker would not survive a review by the Committee on Foreign Investment in the United States, the inter-agency body chaired by Treasury that vetted foreign acquisitions of strategic American firms. Micron declined before the talks turned serious. The bid never reached a formal filing. By autumn it was dead.

The man who had sent the delegation was Zhao Weiguo, the Tsinghua Unigroup chairman who had walked Xi Jinping through Yangtze Memory’s clean room three years later. Zhao had taken over Unigroup in 2009, when it was a sleepy university spinoff that once made medical equipment and now did almost nothing of note, and reorganized it as a private-equity vehicle for Chinese semiconductor consolidation. Tsinghua Holdings retained a controlling stake; Zhao’s own Beijing Jiankun Investment owned about forty-nine percent. The structure was unusual. The company looked private from one angle and state-owned from another, and Zhao spent the next decade exploiting the ambiguity. By 2015 he was the most aggressive deal-maker in the global chip industry.

His timing was excellent. In June 2014, the State Council had published a guideline document for the development of the Chinese integrated-circuit industry, and three months later Beijing had launched the National Integrated Circuit Industry Investment Fund, a vehicle styled as private but capitalized initially with about twenty-two billion dollars from the Ministry of Finance, China Development Bank Capital, and a handful of state-owned enterprises. Industry shorthand called it the Big Fund. Its mandate, plainly stated, was to invest in chip manufacturing and “promote mergers and acquisitions” until China had a domestic semiconductor industry that could rival the foreign incumbents. Zhao’s Unigroup was not the Big Fund, but it moved in lockstep with the policy that had produced it, and Zhao understood his role. He intended to be the consolidator.

The first acquisitions had already been done. In July 2013 Unigroup had agreed to buy Spreadtrum Communications, a Shanghai-incorporated baseband-chip designer listed on NASDAQ, for about $1.78 billion. In November it filed to take over RDA Microelectronics, another fabless Chinese firm with an American listing, for around $907 million. The two companies, once merged and rebranded as UNISOC, would become China’s principal challenger to Qualcomm and MediaTek in cellular modem chips. In September 2014, Intel had bought a twenty percent stake in Unigroup for $1.5 billion, ostensibly to access the Chinese mobile market through a UNISOC-Intel joint design effort. The deal stunned the industry. The largest American chipmaker had taken a minority position in a Chinese state-affiliated holding company, less than a year before that same holding company would try to buy the largest American memory maker. Zhao made no secret of what came next. In a Bloomberg interview in mid-2015, with the Micron approach already in motion, he gave the chip industry the sentence it would remember. Mergers between big American and Chinese companies, he said, were bound to happen. They should be viewed from a business perspective rather than treated under nationalist or political context. The line read, depending on the reader, as commercial common sense or as a brazenly ahistorical denial of where the world was heading.

Where it was heading became visible within months. In May 2015, Hewlett-Packard, in the middle of splitting itself into two firms, sold a fifty-one percent stake in its Chinese networking subsidiary, H3C Technologies, to Tsinghua Holdings for $2.3 billion. The deal was structured as a joint venture rather than as a foreign acquisition, and cleared without CFIUS friction. Then, in late September, Unisplendour announced a more direct move: a $3.78 billion equity investment in Western Digital, the California maker of hard drives, that would have given Unigroup a fifteen percent stake and made it the company’s largest shareholder. The strategic logic ran through Western Digital’s pending acquisition of SanDisk, the Milpitas NAND-flash maker. By becoming Western Digital’s largest shareholder, Unigroup would acquire de facto influence over a leading American producer of flash memory. The deal was filed with CFIUS in late 2015, and in February 2016, Western Digital announced that Unisplendour would not move forward. CFIUS had signaled its intent to investigate. Both sides walked away rather than enter a review they expected to lose.

The deals that did not happen accumulated faster than the deals that did. In November 2015 Zhao went to Taipei, told reporters he was interested in buying a stake in TSMC, and floated the idea of merging MediaTek with Unigroup’s chip-design business. Morris Chang dryly observed that any twenty-five percent stake in TSMC would cost on the order of thirty billion dollars and that the price might be a problem; Foxconn’s Terry Gou, less politely, called Zhao a stock-market speculator with no real technology. Taiwan’s investment regulations, which barred Chinese capital from controlling Taiwanese chip-design firms, made the MediaTek idea moot anyway. Zhao retreated, but only partly. Unigroup had already announced a $600 million purchase of a twenty-five percent stake in Powertech Technology, a Taiwanese assembly-and-test firm outside the protected design domain, and in December it filed bids for parts of two more Taiwanese back-end firms, SPIL and ChipMOS. Each would eventually be reduced or unwound after pushback in Taipei, where lawmakers had begun to read the Unigroup spree as a strategic threat rather than a series of commercial transactions.

By early 2016, a list had begun to circulate in Washington, kept loosely by the Commerce Department and the Senate Intelligence Committee staff. It tracked every Chinese-led bid for a foreign semiconductor firm of consequence. OmniVision, the Santa Clara CMOS image-sensor maker, taken private by a consortium including Hua Capital and CITIC for about $1.9 billion. Integrated Silicon Solution, the San Jose memory-controller firm, bought by a Chinese-led group for about $750 million. Mattson Technology, the Fremont equipment maker, sold to Beijing E-Town for about $300 million. The failed approaches were more numerous than the completed ones, and included Fairchild, GlobalFoundries, Atmel, and Micron itself. Rhodium Group later estimated that Chinese firms managed to close roughly fourteen billion dollars in foreign chip-sector acquisitions between 2015 and 2018, against tens of billions more withdrawn, blocked, or never formally announced. Tsinghua headlined the aggressive ones, but the structure behind them was uniformly the same: state-directed funds, often layered through nominally private vehicles, going after foreign capability China had not yet been able to build at home.

The American response began inside the Pentagon and the White House science office before it surfaced as policy. In the spring of 2016, Commerce Secretary Penny Pritzker asked the President’s Council of Advisors on Science and Technology to convene a working group on the long-term competitiveness of the U.S. semiconductor industry. It was co-chaired by John Holdren, the president’s science adviser, and Paul Otellini, the recently retired chief executive of Intel. They worked through the summer and fall, interviewing executives and reading the Chinese policy documents. The report they delivered to Obama in early January 2017, titled “Ensuring Long-Term U.S. Leadership in Semiconductors,” opened with a sentence that read like a warning shot. A concerted push by China to reshape the market in its favor, the report said, threatened the competitiveness of the United States industry and the national and global benefits it brought. The threat was not the existence of a Chinese chip industry. The threat was the form China had chosen, in which the state directed roughly a hundred and fifty billion dollars across a decade into subsidized capacity and into outbound acquisitions whose purpose was to absorb capability rather than to compete on the market. Pritzker, at CSIS in November 2016, summarized the conclusion plainly. The United States would not allow any nation to dominate the semiconductor industry through unfair trade practices and massive non-market intervention. The signal was that the next CFIUS review of a Chinese chip acquisition was going to look different.

The next review came almost immediately. In May 2016, a Chinese-led investment group called Fujian Grand Chip Investment had filed a tender offer to buy Aixtron, a German maker of metal-organic chemical-vapor-deposition equipment, the kind of tool used to grow the thin compound-semiconductor layers that go into power and radio-frequency chips. The deal was modestly sized, around 670 million euros, and it was for a German firm, but Aixtron had a small American subsidiary, and that subsidiary was enough to bring the transaction within CFIUS’s reach. American intelligence agencies concluded that the tool set Aixtron made was more militarily relevant than its civilian customer base suggested. Fujian Grand Chip had no plausible mitigation. CFIUS recommended that the President block the transaction, and on December 2, 2016, Obama signed an executive order doing so. It was the first time a sitting president had used the CFIUS statute to prohibit a foreign acquisition before it closed; the three previous presidential blocks, dating back to 1990, had unwound deals already consummated. Obama’s order opened the prohibition tool. The next administration would pick it up.

The first Trump-era test arrived in the form of Lattice Semiconductor, a Portland, Oregon designer of low-power field-programmable gate arrays whose chips, while no longer a large-volume Pentagon product, still ended up in enough sensitive applications to make the company a recognized supplier to the U.S. government. In November 2016, Canyon Bridge Capital Partners, a newly formed private-equity fund founded by Benjamin Chow with former Cadence chief executive Ray Bingham as advisory partner, agreed to buy Lattice for $1.3 billion. Canyon Bridge described itself as a private American fund. Its sole capital backer was China Reform Management, the investment arm of a corporation owned by China’s central government and seeded with several billion dollars of state capital. It was the cleanest available example of how Chinese state money was now reaching American chip firms through privately styled wrappers. Canyon Bridge filed for CFIUS review in December 2016. Over the next nine months, the parties refiled twice, in March and June 2017, each time agreeing to a fresh thirty- or forty-five-day clock, attempting to negotiate mitigation terms that would let the deal close. CFIUS was not persuaded. On September 13, 2017, eight months into his presidency, Trump signed an executive order prohibiting the acquisition. The White House statement cited the potential transfer of IP, the role of the Chinese government in backing the transaction, supply-chain integrity, and the use of Lattice products by the U.S. government. It was only the fourth time in the statute’s history that a president had blocked a deal, and the first by Trump. Within months Chow was arrested by federal prosecutors in Manhattan on insider-trading charges related to the Lattice transaction. He was convicted in April 2018.

The Lattice precedent reframed the calculus for every subsequent deal. CFIUS, working under the same Exon-Florio authority it had held since 1988, began to operate with a wider definition of what constituted a national-security risk. Senator John Cornyn of Texas, who had spent the year quietly drafting reform legislation with Senator Dianne Feinstein and the Pentagon’s policy office, used the Lattice case as Exhibit A. The CFIUS statute, Cornyn said, was written for a different China. It needed to handle joint ventures, minority investments, and shell-fund structures of the kind Canyon Bridge had used. The bill he and Feinstein introduced in November 2017, the Foreign Investment Risk Review Modernization Act, expanded CFIUS jurisdiction to cover non-controlling investments in critical-technology firms, real-estate transactions near sensitive sites, and changes in foreign-investor rights even when no new equity changed hands. It also gave CFIUS its own funding, its own staff, and an explicit mandate to track emerging technologies, of which semiconductors were the first item on the list. FIRRMA was attached to the National Defense Authorization Act for fiscal 2019 and signed by Trump on August 13, 2018. From that day, no Chinese investment in an American chip firm of any size was likely to escape review.

The year FIRRMA passed was also the year of the most dramatic CFIUS intervention to date, and the firm being protected was not American at all. In late 2017, Hock Tan, the chief executive of Singapore-headquartered Broadcom, had stood in the Oval Office beside Trump and announced that his company would redomicile from Singapore to Delaware. The president, smiling, had called Broadcom one of the really great companies. Days later, Tan launched a hostile $117 billion bid for Qualcomm, the San Diego designer of mobile-phone modems and the leading American holder of cellular standards-essential patents. Qualcomm’s board rejected the offer; Broadcom proposed to circumvent the board through a proxy contest at Qualcomm’s annual meeting. On January 29, 2018, Qualcomm took the unusual step of filing its own CFIUS notice, asking the committee to review what would happen to American 5G leadership if Broadcom, with its short-term financial discipline, replaced Qualcomm’s long-term R&D commitment. CFIUS agreed to look. On March 4, before Qualcomm’s shareholder meeting could vote on Broadcom’s slate, the committee issued an interim order delaying the meeting. On March 12, Trump signed an executive order permanently prohibiting the takeover. The official rationale was that any weakening of Qualcomm’s position on 5G would create an opening for Chinese firms, principally Huawei, to take the lead on the next-generation wireless standard, and that the United States would not accept that outcome. It was the first time CFIUS had blocked a transaction not because the buyer was Chinese but because the deal would, in the committee’s view, indirectly benefit China. National-security review now extended to transactions whose effect on competitive position, rather than whose direct ownership change, was deemed strategically intolerable.

Beijing was watching, and drawing its own conclusions. Four months after the Broadcom-Qualcomm block, Qualcomm reached the deadline on a different deal, a $44 billion acquisition of NXP Semiconductors of the Netherlands announced in October 2016 that needed antitrust clearance from nine jurisdictions. By July 2018 it had clearances from eight, including the European Commission and the U.S. Federal Trade Commission. The ninth was China’s State Administration for Market Regulation, which had taken Qualcomm’s filing under review and then, as the trade war between Washington and Beijing escalated through the spring and early summer, had stopped responding to Qualcomm’s questions. The official reason was unresolved competition concerns. Qualcomm’s executives, the American business press, and most of the antitrust bar in Beijing assumed something else. SAMR’s silence was the Chinese government’s first formal use of merger review as retaliation against American export controls. On July 26, 2018, with the deadline expiring at midnight Singapore time and SAMR still silent, Qualcomm announced the termination of the NXP deal. It paid NXP a two-billion-dollar break fee. It was the first time a global chip transaction approved by every other major regulator had been killed by Beijing.

The asymmetry of the year was striking. The United States had moved, deal by deal, to close a window through which Chinese capital had flowed into the global semiconductor industry. The Lattice block, the Qualcomm-Broadcom block, FIRRMA’s expanded jurisdiction: each was a discrete decision about a discrete transaction, and together they amounted to a doctrine. China’s response had been more limited in shape but no less consequential. By using SAMR to kill Qualcomm-NXP, Beijing had signaled that the chokepoints in the system worked in both directions, and that any future American chip merger requiring Chinese antitrust approval would need to pass a political review whose criteria would not be published. The era of Chinese acquisitions of foreign chip capability had ended. The era of Chinese leverage over foreign chip mergers had begun.

Zhao never bought Micron. He never bought Western Digital. He never bought a stake in TSMC or MediaTek or SanDisk. By the late 2010s, with the global door closed, he turned the Unigroup machine inward, pouring borrowed billions into Yangtze Memory Technologies, the 3D NAND fab in Wuhan, and into a planned DRAM operation in Chongqing run by Charles Kau. The domestic spending was on a scale that even the Big Fund could not sustain, and by November 2020 Unigroup had defaulted on a $200 million bond, triggering cross-defaults across the conglomerate’s stack of debt. By July 2021 a creditor had filed for bankruptcy. By April 2022 Zhao had been displaced as chairman by a new owner, JAC Capital, in a state-orchestrated restructuring he denounced as criminal. In July of that year he was detained by the Central Commission for Discipline Inspection. In May 2025 he was convicted of corruption and embezzlement and sentenced to death with reprieve.

The line he had given Bloomberg ten years earlier, that mergers between big American and Chinese companies were bound to happen, would by then read as a fragment from a different era. The era in which it had seemed plausible had lasted, by the calendar, fewer than three years. What replaced it, on both sides of the Pacific, was a new vocabulary of decoupling, of strategic technologies, of supply-chain integrity, in which a check from a Chinese fund and a board seat at an American chip firm had become not merely difficult but politically inconceivable. Zhao had read the moment exactly wrong. The mergers had not been bound to happen. The blocks had.