Tech Giants: Starting from Bankrupt Companies

98: Chapter 98 The Silent Currents of the Chip BattlefieldSeptember 10th, 9:00 AM, the circular conference room on the top floor of the Xingchen Academy of Sciences in Shenzhen.

The sweltering heat of August had not yet completely receded, but the air conditioning in the conference room was running full blast. Eleven people sat around the oval conference table, each with an expression as solemn as if they were discussing an impending battle.

In fact, this was indeed a battle.

Li Weimin stood in front of the large screen, the sleeves of his white lab coat rolled up to his elbows, holding a laser pointer. This seventy-two-year-old Academician had specially shaved today, but the dark circles under his eyes still betrayed his exhaustion.

”As of 4:00 AM today, the overall progress of the EDA toolchain is 85% complete.” The red dot of the laser pointer landed on the progress bar. ”All key algorithm verifications have passed, including the timing-driven co-design algorithm we broke through last week. Measured data shows the chip design iteration cycle has been compressed from an average of fifteen rounds to five, a three-fold increase in efficiency.”

Su Chen sat in the head seat with a notebook open in front of him, but he hadn’t written a single word. His gaze was fixed on every number on the screen.

”For the 16nm process library, the library building work is 70% complete.” Li Weimin switched to the next chart. ”I want to give a special mention to the colleagues in the process group—we adopted the FD-SOI route, successfully bypassing the patent walls of TSMC and Samsung regarding the FinFET structure. Although FD-SOI is slightly inferior to FinFET in terms of peak performance, it has a natural advantage in power consumption control.”

The chart displayed comparison data: FD-SOI transistor performance reached 90% of TSMC’s FinFET, but power consumption was reduced by 25%.

”What does this mean?” Li Weimin looked at the young engineers present. ”It means our chips might not reach the most extreme frequencies, but they will be more power-efficient and generate less heat in daily use. For smartphones and IoT devices, this is a huge advantage.”

The process group leader, a female PhD in her early thirties, raised her hand. ”Academician Li, the wafer cost of FD-SOI is 30% higher than bulk silicon. This part...”

”Spark has already invested in and taken a stake in NSIG.” Su Chen spoke, his voice calm but carrying an unquestionable power. ”They are currently expanding a 12-inch dedicated FD-SOI production line, which will go into production next June. Our wafer costs can be controlled within a reasonable range.”

The female PhD nodded and quickly took notes in her notebook.

”Third item: equipment adaptation.” Li Weimin pulled up a photo—a prototype of a 28nm lithography machine from Shanghai Micro Electronics Equipment, its massive yellow body occupying the entire cleanroom. ”Last week, we completed the first joint debugging with Shanghai Micro Electronics Equipment. Although it was only an exposure test, it marks the first time that domestic EDA, domestic process libraries, and domestic lithography machines have run through the complete process.”

The photo was enlarged, showing the patterns exposed on the wafer—a feature size of 28nm, with clear lines and tidy edges.

”How is the uniformity?” Zhang Wei asked. This former Intel senior director now served as the chief architect for Spark chips. In his forties, he wore a simple polo shirt, but his eyes were as sharp as a scalpel.

”Currently, it’s only 80% of ASML’s equivalent equipment,” Li Weimin admitted frankly. ”But remember—this is a breakthrough from zero to one. Three months ago, this machine didn’t even have a stable light source.”

The conference room door was gently pushed open, and a white-haired foreigner poked his head in. André Dupont, an Academician of the French Academy of Sciences and former chief optical scientist at ASML, was now the chief consultant for the Spark equipment group.

”Sorry, I’m late.” Dupont’s French-accented English was heavy. He held a phone in his hand, his face showing uncontrollable excitement. ”But I’ve brought something good.”

He walked to the conference table and connected his phone to the projector. The screen flickered a few times before stabilizing—it was a real-time surveillance video from the Shanghai Micro Electronics Equipment factory.

In the center of the frame, that 28nm lithography machine was running. The robotic arm moved the wafer smoothly, and the laser light source emitted a faint purple glow. The engineers at the console stared tensely at the screen, no one speaking.

Then, a green ”PASS” sign popped up on the main control screen.

A brief silence followed, and then cheers erupted in the workshop. Young engineers hugged each other; some jumped up, while others took off their cleanroom caps and waved them vigorously.

The video ended, and the conference room fell silent.

Dupont looked at everyone and said word for word, ”This was recorded at 6:00 AM this morning. We successfully exposed a test pattern with a feature size of 28 nanometers. Although the uniformity is only 80%, ladies and gentlemen, we did it. Without EUV, we will push Deep Ultraviolet (DUV) to the limit.”

He held up three fingers. ”Give me another three months. I can increase the uniformity to 85% and the production capacity to 30,000 wafers per month. By then, Spark will truly have an independent and controllable 28nm production line.”

Applause rang out, sparse at first, then enthusiastic.

Su Chen stood up, walked over to Dupont, and extended his hand. ”André, thank you for your hard work.”

Dupont shook his hand, his eyes a bit moist. ”Su, I believe we can definitely break America’s technological blockade. Technological progress should serve all of humanity, not be used as a political weapon.”

The meeting continued.

Zhang Wei took over from Li Weimin at the front and pulled up the design diagrams for the ”Flint Architecture.” It was an extremely complex block diagram, with various modules marked in different colors.

”A completely independent instruction set, designed from scratch, bypassing all of ARM’s patents.” Zhang Wei circled the core part with his laser pointer. ”The modular design is the biggest feature—from four cores to sixty-four cores, it’s like building with blocks. The quad-core version is for IoT and automotive ECUs; the eight-core version is for phones and tablets; the sixteen-core version is for servers and autonomous driving; and the sixty-four-core version is for supercomputing and AI training.”

He switched to a performance comparison chart. ”The power efficiency ratio is 15% higher than ARM’s latest A76. Why can we do this? Because we have no historical baggage and can redesign the pipeline, cache structure, and power management from the ground up.”

Su Chen asked, ”How will the ecosystem be solved? No matter how advanced the instruction set is, it’s just scrap metal without software.”

This question hit the most critical point. Everyone in the conference room looked at Zhang Wei.

Zhang Wei smiled; it was his first smile of the day. ”James, you answer that.”

James Wilson stood up. This Fellow of the Royal Academy of Engineering and former chief architect of ARM was 62 years old this year. Wearing rimless glasses, he spoke slowly but with clear logic.

”Over the past month, we’ve done three things.” Wilson pulled up another set of slides. ”First, we’ve deeply integrated with the HarmonyOS team led by Huawei. The Flint Architecture will become one of the ’officially recommended architectures’ for HarmonyOS. HarmonyOS version 3.0 will natively support the Flint instruction set.”

A gasp sounded in the conference room. Although HarmonyOS was still in its early stages, it had huge potential backed by Huawei’s ecosystem.

”Second, we’ve developed a complete compilation toolchain,” Wilson continued. ”Software for the ARM architecture can run on Flint chips through binary translation, with performance loss kept within 15%. While not perfect, it’s enough to solve the ecosystem issues during the transition period.”

”Third—” He paused, his smile deepening. ”We have already signed strategic cooperation agreements with the top ten software companies in the country. These mainstream applications will all launch native versions for the Flint Architecture before the end of the year.”

Su Chen leaned back against his chair and let out a long breath.

Only he knew the pressure of this past month. chip R&D is like digging a tunnel in the dark; you don’t know if there’s rock or a cliff ahead, you just have to keep chipping away. And now, he could finally see the light.

”Timeline?” he asked.

Li Weimin stood at the front again. ”On September 30th, we will complete the first design version of ’flint no. 2’ using the 16nm process. By December 31st, we will have a successful tape-out. In March of next year, small-batch mass production will begin. If everything goes smoothly, we can release the first smartphone prototype powered by the flint no. 2 chip at the Shenzhen Electronics Fair next June.”

”Have you talked to the phone manufacturers?”

”We’ve talked to four,” Li Weimin nodded. ”Huawei, OPPO, vivo, and Transsion. The most enthusiastic is Transsion—they urgently need cost-effective chips for the African market. Mr. Ren from Huawei even called me yesterday, saying, ’Elder Li, as soon as the chip is out, Huawei will order at least five million units.’”

The meeting ended at noon. After the meeting adjourned, Su Chen specifically asked Li Weimin to stay.

”Elder Li, you don’t look well.” Su Chen poured a cup of hot water for the old Academician. ”Staying up late again?”

Li Weimin smiled bitterly. ”I can’t sleep. As soon as I close my eyes, I see transistors and circuit diagrams. Su Chen, to be honest with you—the schedule looks beautiful, but the risks are still very high. There are still bugs in the EDA tools, the SPICE models in the process library aren’t accurate enough, and the stability of the lithography machine...”

”I know,” Su Chen interrupted him. ”But we have no choice. TSMC has officially notified us that due to pressure from America, they cannot manufacture any advanced process chips for Spark. SMIC’s 14nm capacity is prioritized for Huawei. We can only rely on ourselves.”

Li Weimin was silent. Outside the window, the Shenzhen sky was as blue as if it had been washed; this young city was always full of vitality. But in this building, a group of people was fighting for the technological destiny of this country.

”I have a request,” Li Weimin said suddenly.

”Go ahead.”

”Expand the scope of the mandatory rest system.” The old Academician’s expression was serious. ”Not just the EDA group, but the entire chip team must implement it. I heard yesterday that two more young engineers fainted. Su Chen, technology can be caught up with, but if the people break down, then everything is truly gone.”

Su Chen looked at this seventy-two-year-old man—his own health wasn’t good, with high blood pressure and heart disease, yet he was still worrying about the young people.

”Okay,” Su Chen made a solemn promise. ”Starting today, the entire Xingchen Academy of Sciences will strictly implement working hour management. I will personally oversee it.”

Li Weimin nodded, picked up his thermos, and hobbled out of the conference room.

Su Chen stood by the window, watching the engineers walking in the garden below. Most of them were very young, in their twenties or early thirties, an age when they should be enjoying life, yet they had chosen this most difficult path.

His phone vibrated. It was Zhou Qingyu.

”President Su, something has happened,” the voice on the other end of the line carried a suppressed anxiety. ”The competitors have started a price war.”