The China Mail - China's chip challenge: the race to match US tech

USD -
AED 3.672504
AFN 65.000368
ALL 81.652501
AMD 376.168126
ANG 1.79008
AOA 917.000367
ARS 1431.790402
AUD 1.425923
AWG 1.8025
AZN 1.70397
BAM 1.654023
BBD 2.008288
BDT 121.941731
BGN 1.67937
BHD 0.375914
BIF 2954.881813
BMD 1
BND 1.269737
BOB 6.889932
BRL 5.217404
BSD 0.997082
BTN 90.316715
BWP 13.200558
BYN 2.864561
BYR 19600
BZD 2.005328
CAD 1.36855
CDF 2200.000362
CHF 0.77566
CLF 0.021803
CLP 860.890396
CNY 6.93895
CNH 6.929815
COP 3699.522179
CRC 494.312656
CUC 1
CUP 26.5
CVE 93.2513
CZK 20.504104
DJF 177.555076
DKK 6.322204
DOP 62.928665
DZD 129.553047
EGP 46.73094
ERN 15
ETB 155.0074
EUR 0.846204
FJD 2.209504
FKP 0.735067
GBP 0.734457
GEL 2.69504
GGP 0.735067
GHS 10.957757
GIP 0.735067
GMD 73.000355
GNF 8752.167111
GTQ 7.647681
GYD 208.609244
HKD 7.81385
HNL 26.338534
HRK 6.376104
HTG 130.618631
HUF 319.703831
IDR 16855.5
ILS 3.110675
IMP 0.735067
INR 90.596504
IQD 1306.186308
IRR 42125.000158
ISK 122.710386
JEP 0.735067
JMD 156.057339
JOD 0.70904
JPY 157.200504
KES 128.622775
KGS 87.450384
KHR 4023.848789
KMF 419.00035
KPW 900.021111
KRW 1463.560383
KWD 0.30721
KYD 0.830902
KZT 493.331642
LAK 21426.698803
LBP 89293.839063
LKR 308.47816
LRD 187.449786
LSL 16.086092
LTL 2.95274
LVL 0.60489
LYD 6.314009
MAD 9.153622
MDL 17.000296
MGA 4426.402808
MKD 52.129054
MMK 2100.115486
MNT 3570.277081
MOP 8.023933
MRU 39.425769
MUR 46.060378
MVR 15.450378
MWK 1728.952598
MXN 17.263604
MYR 3.947504
MZN 63.750377
NAD 16.086092
NGN 1366.980377
NIO 36.694998
NOK 9.690604
NPR 144.506744
NZD 1.674621
OMR 0.383441
PAB 0.997082
PEN 3.354899
PGK 4.275868
PHP 58.511038
PKR 278.812127
PLN 3.56949
PYG 6588.016407
QAR 3.634319
RON 4.310404
RSD 99.268468
RUB 76.789716
RWF 1455.283522
SAR 3.748738
SBD 8.058149
SCR 13.84955
SDG 601.503676
SEK 9.023204
SGD 1.272904
SHP 0.750259
SLE 24.450371
SLL 20969.499267
SOS 568.818978
SRD 37.818038
STD 20697.981008
STN 20.719692
SVC 8.724259
SYP 11059.574895
SZL 16.08271
THB 31.535038
TJS 9.342721
TMT 3.505
TND 2.891792
TOP 2.40776
TRY 43.612504
TTD 6.752083
TWD 31.590367
TZS 2577.445135
UAH 42.828111
UGX 3547.71872
UYU 38.538627
UZS 12244.069517
VES 377.985125
VND 25950
VUV 119.620171
WST 2.730723
XAF 554.743964
XAG 0.012866
XAU 0.000202
XCD 2.70255
XCG 1.797032
XDR 0.689923
XOF 554.743964
XPF 100.858387
YER 238.403589
ZAR 16.04457
ZMK 9001.203584
ZMW 18.570764
ZWL 321.999592
  • SCS

    0.0200

    16.14

    +0.12%

  • RBGPF

    0.1000

    82.5

    +0.12%

  • GSK

    1.0600

    60.23

    +1.76%

  • CMSD

    0.0600

    23.95

    +0.25%

  • BCE

    -0.4900

    25.08

    -1.95%

  • NGG

    1.1700

    88.06

    +1.33%

  • BTI

    0.8400

    62.8

    +1.34%

  • AZN

    5.8700

    193.03

    +3.04%

  • CMSC

    -0.0400

    23.51

    -0.17%

  • RIO

    2.2900

    93.41

    +2.45%

  • RELX

    -0.7100

    29.38

    -2.42%

  • JRI

    0.0900

    12.97

    +0.69%

  • VOD

    0.4900

    15.11

    +3.24%

  • BCC

    1.8700

    91.03

    +2.05%

  • RYCEF

    0.2600

    16.88

    +1.54%

  • BP

    0.8400

    39.01

    +2.15%

China's chip challenge: the race to match US tech
China's chip challenge: the race to match US tech / Photo: © AFP/File

China's chip challenge: the race to match US tech

China's push to develop top-end artificial intelligence microchips is gaining momentum, but analysts say it will struggle to match the technical might of US powerhouse Nvidia within the current decade.

Text size:

Ramping up its chip industry is a way for Beijing to beat restrictions imposed by Washington on exports of the most advanced chips -- used to power AI systems -- to China.

The United States cites national security concerns, such as the risk of giving China a military advantage, for the block, a geopolitical bind that shows no sign of easing.

"China wants chips that policy cannot take away," said Stephen Wu, a former AI software engineer and founder of the Carthage Capital investment fund.

However, "full end-to-end parity with Nvidia's best chips, memory packaging, networking and software is not guaranteed" by 2030 or even beyond, Wu told AFP.

Announcements of computing upgrades by Chinese companies and reports of plans to dramatically increase output of advanced semiconductors have driven up chip-related shares in the country.

But to catch up with Nvidia, China needs to make fast progress on high-bandwidth memory and packaging -- "the hardest and most complex parts of the chip", Wu said.

Other challenges include building the right software to harness the chips' power, and upgrading manufacturing tools.

"These chips are extremely advanced and tiny, so imagine carving a stone sculpture with a hammer instead of a chisel," Wu said.

- 'Only way' to succeed -

"The industry consensus is China at least needs five to ten years to catch up," said George Chen of The Asia Group, a view reflected by Dilin Wu, research strategist at Pepperstone.

"The future is bright, but not yet," she told AFP.

"It's maybe a 2030 story", as "significant gaps remain in terms of performance, and also in terms of energy efficiency and ecosystem maturity".

Public demand for AI services is booming in China, and while government support for new chips is "substantial", the investment required is "immense", she added.

Shares in Alibaba, the e-commerce titan ploughing billions of dollars into AI tech, have more than doubled since January.

And Chinese chip industry leader Huawei will reportedly double output of its top Ascend 910C chip in the next year.

The hype has also sharply driven up stocks in the smaller chipmaker Cambricon, sometimes dubbed "China's Nvidia".

"I think this rally can be sustained", partly because it is driven by Chinese government policy, Pepperstone's Wu said.

Even Xiaomi, whose 2014 venture into chip design was a self-confessed flop, is turning back to semiconductors.

"Chips are the only way for Xiaomi to succeed," the company's CEO Lei Jun said in Beijing last month, referring to the production of high-end smartphone chips.

- 'Best in China' -

China, the world's biggest consumer of semiconductors, is a huge market for California-based Nvidia.

Nvidia chips are still "the best... to train large language models", the systems behind generative AI, said Chen Cheng, general manager for AI translation software at tech firm iFLYTEK.

Faced with US restrictions, "we overcame that difficulty" by shifting to Chinese-made tech, she said in a group interview.

"Now our model is trained on Huawei chips" -- currently the best in China, Cheng said.

Meanwhile Nvidia, the world's largest company by market capitalisation, is under pressure from both sides.

The Financial Times reported last month that Beijing had barred major Chinese firms from buying a state-of-the-art Nvidia processor made especially for the country.

And the company must now pay the US government 15 percent of revenue from certain AI chip sales in China.

Nvidia boss Jensen Huang has warned that restrictions on exporting his most cutting-edge semiconductors to China will only fuel the country's rise.

"They're nanoseconds behind us," the leather jacket-clad Huang said on a tech business podcast.

"So we've got to go compete."

M.Chau--ThChM