The Highway

And the best-documented stretch of that highway runs from Berkeley, California, to Shenzhen, China. It has a name, a charter, a campus in each country, and a paper trail thick enough to fill a congressional file. Let’s walk it.

The Institute

The Tsinghua-Berkeley Shenzhen Institute — abbreviated everywhere as TBSI — was established in 2014 as a three-way joint venture among the University of California, Berkeley, Tsinghua University, and the municipal government of Shenzhen.¹ Its declared research areas were the unobjectionable surface vocabulary of any modern technology institute: information technology, data science, precision medicine, energy, the environment. Its operational reality, as a House investigation would later document, ran deeper into the technologies the United States was actively trying to keep out of Chinese military hands.

The structure is worth pausing on. TBSI is not a contract or a research grant. It is an institution — with its own non-profit entity in Shenzhen, its own faculty, its own dual-degree master’s program that rotates students between Shenzhen, Berkeley, and back to Tsinghua, an industrial advisory board, and a residency requirement obliging Berkeley faculty to spend eight weeks each year on Chinese soil.² When the marriage was formalized, the Shenzhen government reportedly committed two hundred and twenty million United States dollars to build the campus.³ Tsinghua’s American non-profit arm — the Tsinghua Education Foundation North America, a 501(c)(3) registered in the United States — agreed to provide Berkeley with millions more to stand up the venture.³

And, the House Select Committee on the Chinese Communist Party would later find, Berkeley appears never to have disclosed any of it under United States Department of Education rules.

Section One Hundred Seventeen

That phrase — “never disclosed” — turns on a specific statute, and it is worth understanding because it is the legal mechanism the Berkeley story rides on. Section 117 of the Higher Education Act requires American universities receiving federal funding to report foreign gifts and contracts above a threshold of two hundred and fifty thousand dollars. The reporting is filed semiannually with the Department of Education. It is the public record by which the country knows where foreign money is flowing into its research enterprise.

When the Daily Beast reported in May 2023 that Berkeley had received what would eventually be identified as roughly eighty-seven million dollars in undisclosed Chinese contracts — the two largest being a nineteen-million-dollar and a fifteen-million-dollar payment, both tied to TBSI — the institutional defense was procedural. A Berkeley spokesperson told the Daily Beast that the campus in China was not yet finished, so disclosure had not yet been triggered.⁴ A secondary defense surfaced as the investigation deepened: Berkeley argued that some of the money had not flowed from China directly but through the Tsinghua Education Foundation of North America, an American-based non-profit, and that this technicality removed it from the foreign-disclosure requirement.⁵ The Department of Education opened its own inquiry.⁶

Take a moment with that. The university accepted at least eighty-seven million dollars from entities tied to a foreign government and a state-controlled university, ran a joint institute on the strength of that money for the better part of a decade, and did not file the disclosure forms designed expressly to make such arrangements visible to the American public. Whether any law was ultimately broken is for the Department of Education to determine. The fact of the nondisclosure is undisputed.

The Letter

On July 13, 2023, the chairman of the House Select Committee on the Chinese Communist Party, Mike Gallagher of Wisconsin, and the chairwoman of the House Committee on Education and the Workforce, Virginia Foxx of North Carolina, sent a joint letter to Berkeley’s president, Michael Drake, and its chancellor, Carol Christ. The letter was the formal opening of a congressional investigation into TBSI.¹

The letter laid out a sequence of findings that, taken together, amounted to an indictment in everything but name. I will paraphrase the substance closely, because the language of the document itself does the work.

TBSI’s research priorities, the committee found, tracked the People’s Republic of China’s national science and technology priorities so closely that some of them appeared to mirror the Chinese government’s 13th Five-Year Plan almost word-for-word.¹ The institute had engaged in research on dual-use technologies — semiconductor chips and imaging technology among them — that the committee considered capable of contributing to Chinese intelligence and military capabilities. TBSI alumni had gone on to work at entities tied to the People’s Liberation Army, including the China National Space Administration. Representatives of Chinese companies on the United States Commerce Department’s Entity List — the trade-restriction blacklist — sat on TBSI’s Industrial Advisory Board, with Huawei among the named members; the institute’s founding 2015 memorandum to the University of California Board of Regents had specifically identified Tencent as an example of a potential sponsor.⁷ And Berkeley faculty serving at the institute had received funding from the United States Defense Advanced Research Projects Agency, the Pentagon’s blue-sky research arm.¹

Read that last item twice. The very same Berkeley researchers operating inside a Chinese state-government-funded institute were also drawing money from the agency that exists to give the United States military a technological edge over its strategic competitors. The committee’s letter did not need to draw the conclusion explicitly. It only needed to lay the facts side by side.

A Lab With a Name

It is one thing to read “dual-use technologies” in a committee letter and another to put a name on what actually moved down the highway. Here is a name. In June 2019, at the RISC-V Workshop in Zurich, Dr. David Patterson — the Pardee Professor of Computer Science Emeritus at the University of California, Berkeley, and the winner of the 2017 A.M. Turing Award — announced the creation of a new entity housed inside the Tsinghua-Berkeley Shenzhen Institute: the RISC-V International Open Source Laboratory, known as RIOS Lab.⁸

RIOS Lab’s mission, in its own words, is to bring “the research effort of RISC-V CPU with its software and hardware ecosystems from UC Berkeley to the rest of the world.”⁹ RISC-V is the open-source processor architecture born at Berkeley in 2010 — the candidate technology that the Chinese semiconductor industry, locked out of advanced Western chips by the October 2022 export controls, has embraced as its strategic ladder out of dependence. China’s 14th Five-Year Plan names semiconductor self-sufficiency as a national objective; the French Institut Montaigne notes that Chinese players, as of 2022, hold half the seats on the board of directors of RISC-V International.¹⁰

RIOS Lab does not write white papers. It builds chips. Its PicoRio series of RISC-V processors scales up to 128-core configurations with coherent-mesh support and has been taped out for commercial designs at multiple foundries. Its GreenRio series uses a complete open-source manufacturing flow.⁹ And in July 2023 — the same month the House Select Committee’s letter arrived on the chancellor’s desk — a three-person team from TBSI took second place in the Efabless AI-Generated Design Contest with a RISC-V CPU whose Verilog code had been generated through prompts to ChatGPT-4.¹¹

The Berkeley defense, throughout, has been that this is fundamental, openly published research, posing no national-security concern. That defense is not absurd; RISC-V is open by design. But the broader fact of the matter is harder to wave away: a Berkeley-anchored laboratory, named after Berkeley’s own legacy, operating inside a defense-co-administered Chinese university, producing the chip designs that fill the precise capability gap the United States has been trying, for the better part of a decade, to keep China from closing.

The Seven-Nanometer Question

And among the questions the committee directed at Berkeley, one stood out. I am going to quote the question verbatim, because its specificity is the entire point:

A TBSI team won a contest in April 2023 for 7 nanometer chip technology. Does Berkeley consider this activity to be in violation of the October 7th, 2022, restrictions on advanced chip manufacturing and research? Letter from the House Select Committee on the Chinese Communist Party to UC Berkeley, July 13, 2023

The dates matter. On October 7, 2022, the United States Commerce Department imposed sweeping export controls designed to cut China off from the most advanced semiconductor technology — the chips at the cutting edge of artificial intelligence and military computing. Seven-nanometer process technology fell squarely inside the zone of restriction. Six months later, a team operating inside a joint institute partly owned by an American public university won a contest in precisely that technology in China.¹

The committee asked Berkeley whether that activity had violated the export controls. Berkeley, in its public response, said it engaged only in fundamental research the results of which are openly disseminated, and that it was not aware of any TBSI research conducted for any other purpose.⁶ Both statements may be perfectly accurate at the level of Berkeley’s own laboratories. The committee’s point was that Berkeley did not, and possibly could not, vouch for what was happening at the other end of the highway.

What the Universities Said

Worth noting, here, what was not said. Neither Berkeley nor Tsinghua nor the city of Shenzhen has publicly disputed any of the specific factual findings in the committee’s letter — not the eighty-seven million dollars, not the Huawei advisory-board seat, not the 13th Five-Year Plan alignment, not the seven-nanometer win, not the DARPA-funded faculty serving at the joint institute, not the TBSI alumni working for the China National Space Administration. Berkeley’s assistant vice chancellor Dan Mogulof told the New York Times that the campus “takes the concerns very seriously” and would “fully and transparently cooperate with any federal inquiries.”¹² Vice Chancellor for Research Katherine Yelick said Berkeley researchers engage only in research whose results are openly disseminated.⁶ Both statements address the framing of the findings. Neither contradicts them.

That silence, in journalism, is itself a piece of evidence.

The Timeline

The ten-year arc, in seven dates:

• September 2014Tsinghua-Berkeley Shenzhen Institute formally established by Berkeley, Tsinghua University, and the Shenzhen municipal government.
• June 2019David Patterson announces RIOS Lab, the RISC-V International Open Source Laboratory, housed inside TBSI.
• October 7, 2022United States Commerce Department imposes export controls cutting China off from advanced semiconductor manufacturing and research, including at the seven-nanometer node.
• April 2023A TBSI team wins a contest in seven-nanometer chip technology — six months after the export controls take effect.
• May 2023The Daily Beast reports that Berkeley failed to disclose approximately eighty-seven million dollars in Chinese funding under Section 117 of the Higher Education Act.
• July 13, 2023House Select Committee on the Chinese Communist Party opens its formal investigation with a joint letter from Chairman Gallagher and Chairwoman Foxx.
• September 2024 – early 2025The Select Committee’s broader report describes joint institutes as “conduits for transferring critical United States technology.” Berkeley announces it is “in the process of relinquishing all ownership in the TBSI non-profit entity in Shenzhen.”

The Pattern, Not the Outlier

And it would be easy to read this as a one-campus aberration. It was not. In September 2024, the House Select Committee on the Chinese Communist Party published a broader report describing joint education institutes like Berkeley’s as “conduits for transferring critical United States technology” to China.¹³ The investigators named other arrangements: the Georgia Institute of Technology’s partnership with Tianjin University, which Georgia Tech ultimately wound down after Tianjin University was placed on the United States Commerce Department Entity List in 2020.¹⁴ The University of Michigan, on January 10, 2025, announced it would terminate its two-decade partnership with Shanghai Jiao Tong University.¹⁵

The American higher-education response, that is to say, did eventually arrive. Several major institutional joint ventures were unwound. But the broader picture, the one Tsinghua itself publishes openly on its own website, is bigger than the United States.

Read those numbers slowly. The retreat from TBSI does not close the corridor. It removes one lane. The highway remains open in dozens of others, and Tsinghua itself maintains and publicizes that fact on the front page of its global-engagement website. The most prestigious universities in the United Kingdom, Germany, Japan, France, and the United States are listed as ongoing partners. Some of those partnerships are limited in scope and pose no plausible defense-research risk. Others are structurally similar to TBSI. We do not yet know, from the outside, which is which.

The Retreat — and What It Doesn't Undo

What happened next, at Berkeley, is worth recording with precision because it is the one corner of this story where the American institutional response actually moved.

In September 2024, the House Select Committee released its broader report on joint institutes — the document that used the word “conduits.” Around the same time, Katherine Yelick, Berkeley’s vice chancellor for research, told the higher-education press that because Berkeley had no oversight of the research TBSI conducted with outside partners, the university had decided to relinquish ownership of the institute.⁶ In early 2025, Berkeley’s spokesman Dan Mogulof confirmed in writing that the campus was “in the process of relinquishing all ownership in the Tsinghua Berkeley Shenzhen Institute non-profit entity in Shenzhen.”¹⁹

That is a real outcome, and the people who pressed for it — the House Select Committee, the journalists at the Daily Beast who first reported the undisclosed funding, the watchdog group Open the Books, the New York Times reporters who broke the story of the committee’s letter — deserve credit for delivering it. After a decade of operation, the Tsinghua-Berkeley Shenzhen Institute, on the American side, is being unwound.

But two facts qualify the victory.

The first is timing. TBSI ran for ten years before its American partner began the process of exit. The seven-nanometer contest had been won. The Defense Advanced Research Projects Agency funding had flowed through faculty also serving at the joint institute. RIOS Lab had produced its PicoRio and GreenRio designs. The students had cycled between Shenzhen and Berkeley and back to Tsinghua. The eighty-seven million dollars had moved without being declared. Whatever transferred, transferred. The American withdrawal does not unwind the years of operation that preceded it; it only ends the years that would otherwise have followed.

The second is scope, as of this writing in May 2026. Berkeley’s own global-engagement website still hosts the TBSI page, listing the institute as a “central hub of global research and education at UC Berkeley.”² Tsinghua’s Shenzhen International Graduate School still lists Berkeley among its partner institutions.¹⁸ The TBSI website itself continues to operate, still publishing research output through 2025 in Nature Index-tracked journals.²⁰ The relinquishment of the non-profit ownership is one administrative action. The broader pattern of joint programs, dual-degree pipelines, and faculty co-affiliations is not a building you can sell. It is a network of relationships, and networks do not unwind on the timeline of press releases.

The talent crosses on foot. The intellectual property crosses by wire. The highway carries them both.

What the Highway Carried

And so the geometry of the corridor, after four installments, is now visible in three dimensions.

Part One showed Tsinghua as the leadership incubator of the Chinese Communist Party — the institution from which the politburo, the party committees, and the technical-administrative class are disproportionately drawn. Part Two showed it as the prestige hub of the defense system, co-administered by the defense-industry ministry, housing dedicated weapons laboratories, integrated with the military-civil fusion priorities that have been the heart of Chinese strategic policy since 2015. Part Three showed the recruitment apparatus: hundreds of programs, billions in budgets, the largest single share of awardees in the country, drawing scientists out of the most prestigious laboratories in the free world. This installment shows the institutional rails the recruitment ran on — the joint institutes, the dual-degree programs, the eight-week residency requirements, the unfiled disclosure forms, the Shenzhen-government-funded campus, the seven-nanometer contest won under American export controls, the RISC-V laboratory taping out 128-core chips.

It is one machine. The talent crosses on foot. The intellectual property crosses by wire. The highway carries them both, and it was built — not by foreign intelligence officers, not by stolen blueprints, but by signed agreements between universities, with budgets and dedications and ribbon cuttings.

The next question, the one Part Five takes up, is what happens when the same Tsinghua infrastructure is used not to build a chip but to attack a network. The corridor has a fourth lane, and it is the one the public most associates with the word espionage. It is also the lane where, in 2018, the cybersecurity firm Recorded Future identified network infrastructure at Tsinghua University as the apparent staging ground for sustained reconnaissance against Tibetan diaspora organizations, Kenyan government targets, and Brazilian government targets — a pattern of activity Western researchers have traced back to the same campus across multiple subsequent reports. That is where we go next.