259: Chapter 259 The Rise of Domestic CPUs

But Li Wanqing was not finished yet.

"Next is the 14nm Flint 3 process."

She pulled up the new parameters:

Xinghuo Flint 3 14nm FD-SOI

Transistor density: 44 MTr/mm² (approaching 40% of TSMC's 10nm)

Typical power consumption: 30% lower than industry 14nm processes

Yield: 72% (industry average 68%)

Tape-out successful, small-batch production underway

"Based on Flint 3, we have partnered with Loongson to launch the 'Loongson 3C6000' series of processors."

Li Wanqing displayed the CPU performance comparison:

Loongson 3C6000 (4 cores, 8 threads) vs Intel i5-8400 (6 cores, 6 threads)

Single-core performance: 15% behind

Multi-core performance: On par

Power consumption: 65W vs 95W (31% reduction)

Price: Estimated 1500 rmb vs 1800 rmb

"This means that in the desktop and server fields, we have a usable domestic high-performance CPU for the first time."

Li Wanqing's voice remained calm, but the weight of this statement silenced the entire venue.

Domestic CPUs have been the pain of the Huaxia semiconductor industry for decades.

From the 'Hanxin' scandal to Loongson's difficult breakthroughs, from relying on Intel to being choked... now, there is finally a product that can actually be used.

Zhang Rujing, the founder of SMIC in the audience, had slightly moist eyes.

He had fought for the Huaxia chip industry his whole life, and today he finally saw the dawn.

But today's climax had only just begun.

Li Wanqing switched the PPT.

A extremely minimalist concept diagram appeared on the big screen—no complex process details, just three data points and a timeline:

Xinghuo Flint 4 7nm FD-SOI

Transistor density: 102 MTr/mm²

(Comparison: TSMC 7nm first generation / Intel 10nm 100.8 MTr/mm²)

Power consumption: 20% lower than industry 7nm

Mass production time: Q1 2019

The audience was first dead silent, then completely erupted.

"Q1 2019?!"

"TSMC's 7nm just went into mass production this April, and the yield is still climbing..."

"How many times has Intel's 10nm been delayed? They are announcing mass production of 7nm next year?"

Industry professionals all knew what this meant:

TSMC's 7nm process had just entered risk production in April 2018, with a yield of only about 30%, and it was expected to reach a mature level (yield > 70%) by the end of 2019.

But Xinghuo directly announced: You just started mass production, we will be able to mass produce by the first quarter of next year, and with better performance and lower power consumption.

This was a blatant declaration of a generational gap.

Li Wanqing gave everyone a minute to digest this, then added nonchalantly:

"In addition, our 5nm FD-SOI process has completed its first tape-out in the laboratory. Test data shows that the transistor density has reached 180 MTr/mm², and the power consumption is 15% lower than the industry 5nm."

This last sentence hit everyone's heart like a heavy hammer.

5nm... laboratory tape-out...

TSMC's 5nm is expected to be mass-produced in 2020, Intel is still struggling with 10nm, and Samsung's 5nm roadmap is blurry.

And Xinghuo has already reached the laboratory stage.

After the shock came the cruel commercial reality.

Li Wanqing displayed the most critical PPT slide of the day:

Xinghuo Chip Capacity and Price List (2018-2019)

Process | Current Monthly Capacity | Unit Price (12-inch wafer) | Open Order Ratio

28nm | 400,000 wafers | $2800 | 70%

16nm | 80,000 wafers | $4700 | 40%

14nm | 30,000 wafers | $5200 | 30%

Note: 16nm/14nm capacity is prioritized for companies within the xinghuo system (Hongqi Automobile, Xingchen Defense, Huaxing Mobile Phone, etc.)

70% of 28nm capacity is open to the public

All prices exclude tax; bulk orders are negotiable

7nm chips will accept pre-orders in Q1 2019, price TBD

When the number '$2800' appeared, the TSMC delegation collectively turned pale.

TSMC's quote for the 28nm process is $3200-$3500, with a cost of about $2600. Xinghuo is selling it directly for $2800 and still making a profit—this means their cost is likely below $2200.

This is a price butcher, and a price butcher backed by technology.

"We are accepting pre-orders starting today." Li Wanqing checked her watch. "28nm chips will begin delivery next month, and 16nm and 14nm will begin delivery in October 2018. The online ordering system opens promptly at 11:00 AM. VIP customers on-site can proceed to the signing area."

She said one last thing: "The press conference ends here. I wish you all... good luck with your orders."

Then she turned and walked off the stage, without another word.

After three seconds of stunned silence, the crowd surged toward the exit.

Half the people rushed to the signing area—there were eight signing tables there, and the Xinghuo business team was already in place.

The other half took out their phones or tablets, logged into the Xinghuo B2B platform, and prepared to grab orders online.

The chip war had moved from a technical manifesto to a stage of commercial slaughter.

11:00 AM, VIP Signing Area

The signing area was located in the conference center next to Hall 1, with eight signing tables arranged in a circle, each with three Xinghuo business representatives and lawyers.

The first to rush to the signing table was Yu Dong, CEO of Huawei terminal.

"Ms. Li." Yu Dong laid his cards on the table directly, skipping the pleasantries. "Our entire mobile phone line next year will use Flint 3 (14nm), and we need at least 1 million wafers of capacity. In addition, we also need 16nm for base station chips, also 1 million."

Li Wanqing personally received a client of Huawei's level. She shook her head: "Mr. Yu, the total capacity for 14nm next year is only 1.2 million wafers (100,000 wafers/month x 12 months). If we give you 1 million... won't the other manufacturers have to starve?"

"Then how much can you give at most?"

"300,000 wafers for the first year." Li Wanqing provided the number. "But we can sign a three-year framework agreement: 300,000 in 2019, 600,000 in 2020, and 1 million in 2021. The price will be adjusted annually based on cost, but the increase will not exceed 5%."

Yu Dong calculated quickly: 300,000 wafers of 14nm at a unit price of $5200 is $1.56 billion. For Huawei, this was an acceptable number—they spent over $8 billion purchasing chips from Qualcomm and TSMC last year.

"Agreed." Yu Dong nodded. "But what about 16nm? We need it for base station chips."

"At most 500,000 wafers for 16nm." Li Wanqing said. "We can also sign a three-year agreement."

"Price?"

"$4700/wafer, discounted to $4500 for bulk orders."

Yu Dong calculated again: 500,000 wafers x $4500 = $2.25 billion.

14nm + 16nm, totaling $3.81 billion. Compared to continuing to use Qualcomm and TSMC, he could save at least $1.5 billion, and the performance was better.

"Sign!" Yu Dong said decisively.

The assistant immediately handed over the prepared framework agreement.

Key terms:

14nm chip: 300,000 in 2019, 600,000 in 2020, 1 million in 2021, unit price $5200 (negotiable annually)

16nm chip: 500,000 in 2019, 800,000 in 2020, 1.2 million in 2021, unit price $4500

Additional terms: Huawei needs to help Xinghuo Chip establish an R&D center in Europe, providing local talent resources and policy alignment

Payment method: 30% advance payment, 40% before delivery, 30% after acceptance

Settlement currency: rmb or Euro, USD not accepted

When Yu Dong saw the additional terms, he raised an eyebrow: "European R&D center?"

"Yes." Li Wanqing smiled. "Huawei has a well-established local team and political and business connections in Europe. We want to leverage that."

This was effectively trading market access for overseas expansion—I give you chips, and you help me open up the European market.

Yu Dong thought for a moment and nodded: "Agreed. We have offices in Munich, Paris, and London, and we can assist."

He signed his name on the agreement.

Huawei, contract signed. Total amount: $3.81 billion (first year).