bodyType

customContents

entityId

printArticlePageQuery

filter

article

body

urlAlias

articleSectionsContent

articleSeoArticle

articleSourceHeaderArticle

contentSchemaRenderContent

contentSubHeadlineContent

contentHeadlineContent

mainImagePrintArticleArticle

authorBlockPrintArticleArticle

contentSponsorTagContent

articlePianoPaywallContent

scmpYoutubeContent

hooksTrackPageViewArticle

query printArticlePageQuery(
  $entityId: String!
  $customContents: [CustomContentInput]
  $bodyType: BodyFormatTypeEnum
) {
  article(filter: {entityId: $entityId}) {
    entityId
    body(customContents: $customContents, type: $bodyType) {
      json
    }
    urlAlias
    ...articleSectionsContent
    ...articleSeoArticle_3o8td4
    ...articleSourceHeaderArticle
    ...contentSchemaRenderContent
    ...contentSubHeadlineContent
    ...contentHeadlineContent
    ...mainImagePrintArticleArticle
    ...authorBlockPrintArticleArticle
    ...contentSponsorTagContent
    ...articlePianoPaywallContent
    ...scmpYoutubeContent
    ...hooksTrackPageViewArticle
    id
  }
}

fragment articlePianoPaywallContent on Content {
  __isContent: __typename
  ...hooksPaywallPromiseContent
}

fragment articleSectionsContent on Content {
  __isContent: __typename
  sectionsV2 {
    value {
      __typename
      ... on Section {
        name
        published
      }
      ... on VideoSection {
        name
        published
      }
      ...helpersCheckIfNewsSectionInFirstIndexSection
      ...entityLink
      ... on Node {
        __isNode: __typename
        id
      }
    }
  }
}

fragment articleSeoArticle_3o8td4 on Article {
  ...hooksArticleSeoArticle
  ...helpersGetArticleTypeArticle
  entityId
  publishedDate
  updatedDate
  summary {
    text
  }
  subHeadline {
    text
  }
  body(customContents: $customContents, type: $bodyType) {
    text
  }
  keywords
  sectionsV2 {
    value {
      __typename
      ... on Section {
        name
      }
      ... on Node {
        __isNode: __typename
        id
      }
    }
  }
  topics {
    name
    entityId
    id
  }
  urlAlias
  authors {
    name
    id
  }
  images {
    style(filter: {style: "768x768"}) {
      url
    }
  }
  readingTime
}

fragment articleSourceHeaderArticle on Article {
  urlAlias
  ...entityLink
}

fragment authorBlockPrintArticleArticle on Article {
  ...defaultArticleDateContent
  ...authorNamesArticle
}

fragment authorNameAuthor on Author {
  ...entityLink
  name
  entityUuid
}

fragment authorNamesArticle on Article {
  authorLocations
  authors {
    entityId
    location
    ...authorNameAuthor
    id
  }
}

fragment blockquoteQuoteContent on Content {
  __isContent: __typename
  ...contentShareWidgetContent
}

fragment checkersCheckContentContainsStarlingImageContent on Content {
  __isContent: __typename
  topics {
    entityId
    id
  }
}

fragment contentHeadlineContent on Content {
  __isContent: __typename
  headline
  socialHeadline
  ...contentHeadlineTagContent
  ...helpersUseContentSponsorContent
}

fragment contentHeadlineTagContent on Content {
  __isContent: __typename
  ...hooksContentHeadlineTagContent
}

fragment contentNewsletterWidgetContent on Content {
  __isContent: __typename
  topics {
    relatedNewsletters {
      ...helpersMassageToNewsletterItemNewsletter
      id
    }
    id
  }
  sectionsV2 {
    value {
      __typename
      ... on Section {
        relatedNewsletters {
          ...helpersMassageToNewsletterItemNewsletter
          id
        }
      }
      ... on Node {
        __isNode: __typename
        id
      }
    }
  }
  ... on Article {
    relatedNewsletters {
      ...helpersMassageToNewsletterItemNewsletter
      id
    }
  }
}

fragment contentSchemaRenderContent on Content {
  __isContent: __typename
  ...blockquoteQuoteContent
  ...hooksInlinePlusWidgetContent
  ...imageContent
  ...inlineWidgetContent
  ...photoGalleryWidgetContent
  ...scmpYoutubeContent
}

fragment contentShareWidgetContent on Content {
  __isContent: __typename
  __typename
  entityId
  urlAlias
  shortUrl
  socialHeadline
  headline
  ... on Video {
    youtubeId
  }
}

fragment contentSponsorTagContent on Content {
  __isContent: __typename
  ...helpersUseContentSponsorContent
}

fragment contentSubHeadlineContent on Content {
  __isContent: __typename
  subHeadline {
    json
  }
  summary {
    json
  }
  ...contentSchemaRenderContent
}

fragment defaultArticleDateContent on Content {
  __isContent: __typename
  ...hooksArticleDateUtilsContent
}

fragment entityLink on BaseWithApplicationAndUrlAlias {
  __isBaseWithApplicationAndUrlAlias: __typename
  __typename
  entityId
  entityUuid
  urlAlias
  urlRedirect {
    toUrl
    id
  }
  application {
    entityId
    id
  }
  ... on Section {
    fullSectionPath {
      entityId
      id
    }
    parentSection
  }
  ...helpersAppBarVariantBase
  ...helpersApplicationBase
  ...helpersCheckIsPlusArticle
  ...helpersStyleSectionContent
}

fragment helpersAppBarVariantBase on BaseWithApplicationAndUrlAlias {
  __isBaseWithApplicationAndUrlAlias: __typename
  ...helpersCheckIsLiveArticle
  ...helpersCheckIsPlusPage
  ...helpersCheckIsPostMagazineArticle
  ...helpersCheckIsStyleArticle
  ...helpersSectionContent
  ...helpersStyleSectionContent
}

fragment helpersApplicationBase on BaseWithApplicationAndUrlAlias {
  __isBaseWithApplicationAndUrlAlias: __typename
  ...helpersCheckIsPlusPage
}

fragment helpersCheckIfNewsSectionInFirstIndexSection on SectionUnion {
  __isSectionUnion: __typename
  ... on Section {
    name
  }
  ... on VideoSection {
    name
  }
}

fragment helpersCheckIsAmpEnabledArticle on Article {
  entityId
  types {
    value {
      entityId
      id
    }
  }
  sectionsV2 {
    value {
      __typename
      ... on Section {
        entityId
      }
      ... on Node {
        __isNode: __typename
        id
      }
    }
  }
  disableOffPlatformContent
  sponsorType
  published
  ...helpersCheckIsPlusArticle
}

fragment helpersCheckIsLiveArticle on Article {
  flattenTypeIds
  liveArticle {
    status
    id
  }
}

fragment helpersCheckIsMainTopicContent on Content {
  __isContent: __typename
  topics {
    entityId
    id
  }
}

fragment helpersCheckIsPlusArticle on Content {
  __isContent: __typename
  ... on Article {
    paywallTypes {
      entityUuid
      id
    }
  }
}

fragment helpersCheckIsPlusPage on Page {
  urlAlias
}

fragment helpersCheckIsPostMagazineArticle on Content {
  __isContent: __typename
  sectionsV2 {
    value {
      __typename
      ... on Section {
        entityId
      }
      ... on Node {
        __isNode: __typename
        id
      }
    }
  }
}

fragment helpersCheckIsStyleArticle on Content {
  __isContent: __typename
  sectionsV2 {
    value {
      __typename
      ... on Section {
        entityId
      }
      ... on VideoSection {
        entityId
      }
      ... on Node {
        __isNode: __typename
        id
      }
    }
  }
}

fragment helpersGetArticleTypeArticle on Content {
  __isContent: __typename
  ... on Article {
    types {
      value {
        entityId
        id
      }
    }
  }
}

fragment helpersMassageToNewsletterItemNewsletter on Newsletter {
  entityId
  name
  alternativeName
  description {
    text
  }
  summary {
    text
  }
  homepageDescription {
    text
  }
}

fragment helpersSanitizeArticleTypeEntityIdsArticle on Article {
  types {
    value {
      entityId
      id
    }
  }
}

fragment helpersSanitizeAuthorTypesArticle on Article {
  authors {
    types
    id
  }
}

fragment helpersSanitizedAuthorEntityIdsArticle on Article {
  authors {
    entityId
    id
  }
}

fragment helpersSectionContent on Section {
  fullSectionPath {
    entityId
    id
  }
}

fragment helpersStyleSectionContent on Section {
  fullSectionPath {
    entityId
    id
  }
}

fragment helpersUseContentSponsorContent on Content {
  __isContent: __typename
  topics {
    entityId
    sponsor {
      ...sponsorTagSponsor
      name
      type
      id
    }
    id
  }
  ... on Article {
    sponsorType
  }
  ...helpersCheckIsMainTopicContent
}

fragment hooksArticleDateUtilsContent on Content {
  __isContent: __typename
  createdDate
  publishedDate
  updatedDate
}

fragment hooksArticleSeoArticle on Article {
  ...helpersCheckIsPostMagazineArticle
  ...helpersCheckIsStyleArticle
  ...hooksContentHeadlineTagContent
  ...helpersCheckIsAmpEnabledArticle
  entityId
  headline
  socialHeadline
  types {
    value {
      entityId
      id
    }
  }
  authors {
    entityId
    id
  }
  images {
    generic_og: style(filter: {style: "og_image_scmp_generic"}) {
      url
    }
    styles(filter: {styles: ["og_postmag", "og_image_style", "og_image_scmp_analysis", "og_image_scmp_explainer", "og_image_scmp_editorial", "og_image_scmp_fact_check", "og_image_scmp_debate", "og_image_scmp_live", "og_image_scmp_obituary", "og_image_scmp_opinion", "og_image_scmp_review", "og_image_scmp_series", "og_image_scmp_generic"]}) {
      url
      style
    }
    type
  }
  urlAlias
}

fragment hooksContentHeadlineTagContent on Content {
  __isContent: __typename
  __typename
  ... on Article {
    types {
      value {
        name
        entityId
        description {
          text
        }
        id
      }
    }
    ...helpersCheckIsStyleArticle
  }
  flag
  sectionsV2 {
    value {
      __typename
      ... on Section {
        name
      }
      ... on VideoSection {
        name
      }
      ... on Node {
        __isNode: __typename
        id
      }
    }
  }
  ...hooksContentHelperContent
}

fragment hooksContentHelperContent on Content {
  __isContent: __typename
  ...helpersSanitizeArticleTypeEntityIdsArticle
  ...helpersSanitizeAuthorTypesArticle
  ...helpersSanitizedAuthorEntityIdsArticle
  flag
}

fragment hooksInlinePlusWidgetContent on Content {
  __isContent: __typename
  topics {
    entityId
    id
  }
}

fragment hooksPaywallPromiseContent on Content {
  __isContent: __typename
  entityId
}

fragment hooksTrackPageViewArticle on Content {
  __isContent: __typename
  entityId
  publishedDate
  headline
  sectionsV2 {
    value {
      __typename
      ... on Section {
        name
      }
      ... on Node {
        __isNode: __typename
        id
      }
    }
  }
  urlAlias
  authors {
    name
    id
  }
  entityUuid
}

fragment imageContent on Content {
  __isContent: __typename
  headline
  ...checkersCheckContentContainsStarlingImageContent
}

fragment inlineWidgetContent on Content {
  __isContent: __typename
  ...contentNewsletterWidgetContent
}

fragment mainImagePrintArticleArticle on Article {
  images {
    title
    url
    isSlideshow
    size_652x446: style(filter: {style: "652x446"}) {
      url
    }
  }
}

fragment photoGalleryDialogContent on Content {
  __isContent: __typename
  __typename
  ...contentShareWidgetContent
}

fragment photoGalleryWidgetContent on Content {
  __isContent: __typename
  ...photoGalleryDialogContent
}

fragment scmpYoutubeContent on Content {
  __isContent: __typename
  entityId
  urlAlias
  publishedDate
}

fragment sponsorTagSponsor on Sponsor {
  name
  type
  images {
    url
  }
}


types

sectionsV2

disableOffPlatformContent

sponsorType

published

headline

socialHeadline

authors

images

publishedDate

updatedDate

summary

subHeadline

keywords

topics

readingTime

authorLocations

Lucy Quaggin

At a New York City ballroom late last month, a crowd of hi-tech luminaries gathered at an honours ceremony. The biggest draw for the assembled tech geniuses was Nvidia founder Jensen Huang, the superstar head of one of the world’s most valuable companies, set to receive the night’s top award.
In some ways, engineers have been at the centre of every industrial revolution, Huang told the crowd, from steam to electricity to information technology.
But this one, artificial intelligence (AI), will be different, he pronounced with evangelical flair and a veiled warning. “This revolution will be deeply profound, and we have to engage it with enthusiasm and optimism on the one hand, but surely with responsibility at its core.”
What largely went unspoken at the Institute of Electrical and Electronics Engineers ceremony that night was the surety that this revolution will play out as a high-stakes race between the United States and China, with those in the room potentially on the losing side.


“China is going to win the AI race,” Huang told the Financial Times pointedly on the sidelines of the Future of AI summit in London last year.
Not everyone agrees, however, with others convinced that China will remain close behind the US but not pass it any time soon.
Huang himself later hedged on social media, saying China was “nanoseconds behind” the US in AI and that “it’s vital that America wins by racing ahead and winning developers worldwide”.
What is clear is that the two countries will remain in an intense wrestling match for global AI supremacy for the foreseeable future, with their rivalry almost certainly a point of negotiation when US President Donald Trump sits down with Chinese counterpart Xi Jinping on May 14 and May 15.
Two countries tackle AI advancement in different ways
The two countries’ approach to AI has historically differed. The US has been more market-driven, funded by aggressive venture firms and focused on creating the highest-performing models.
It has also taken a turn lately, with the Trump administration announcing a policy shift in May that would allow the government access to tech giants’ AI models before they are released.
In contrast, China’s ecosystem – heavily subsidised and otherwise incentivised by Beijing and local governments – has focused on open-source and open-weight models, where parameters are publicly released, as well as large-scale industrial deployment.
Ultimately, winning could come down to who has the most innovative or the most useful tools.
While America had an early lead with OpenAI’s ChatGPT, advancements, the emergence of models from Chinese firms such as DeepSeek in 2025, and Beijing’s push to become a global leader in the sector by 2030, have begun to close the gap between the world’s two largest economies.
That gap may be closer than many think, according to a report released by Stanford in late April, even as experts envision a protracted contest between the two giants involving opaque benchmarks.
The most recent AI Index, released yearly since 2017 by Stanford University’s Human-Centred Artificial Intelligence (HAI) department, aims to give a “deeper understanding of the complex field of AI” by providing “unbiased, rigorously vetted, and globally sourced data”. The highly regarded report is often cited by policymakers, researchers, executives and journalists.
Among the key findings of the 2026 Stanford report was the conclusion that the US-China AI model performance gap has “effectively closed”.
“The US still produces more top-tier AI models and higher-impact patents, while China leads in publication volume, citations, patent output, and industrial robot installations,” Stanford wrote.
China’s Tsinghua University outperforms US Ivy Leagues such as Harvard and MIT when it comes to the volume of top-cited research papers and patents related to AI, according to data analysed by Bloomberg.
DeepSeek’s new model may help drive demand
A report by the Centre for AI Standards and Innovation (CAISI), a US government entity, put the gap between DeepSeek’s latest model, released in April, and leading US models at around eight months.
Martin Chorzempa, a senior fellow at the Peterson Institute for International Economics, said he “tends to agree” with the CAISI figures, noting the US lead over China is “very challenging to measure”.
“Anthropic does not allow anyone in China to use its models, and Chinese law doesn’t allow foreign AI models to just diffuse their non-censored models in the country,” he said, adding, “what happens is people in China are using VPNs and paying intermediaries to get access to US models”.


This is an important indication that US models have an advantage that is not necessarily reflected in the benchmarks, Chorzempa explained, especially since Chinese models are often substantially cheaper than their US counterparts.
If the model is cheaper and as good, “you probably wouldn’t be bothering to go through all this rigmarole to access the model”, he added.
When people refer to “China ‘catching up’, they are typically referring to the gap between US and Chinese models on benchmark scores”, Kyle Miller, a research analyst at Georgetown’s Centre for Security and Emerging Technology, said.
Stanford described AI benchmarks as standardised tests that are used to measure and compare how well AI systems perform on specific tasks.
Just three years ago, America’s lead over China was substantial, according to the Stanford report.
Since China’s DeepSeek came onto the scene, the top performance rankings have shown there’s tight competition, with the leading US model Claude Opus 4.6 by Anthropic, as of March 2026, sitting just 2.7 per cent higher than China’s top model Dola‑Seed‑2.0 Preview, owned by ByteDance.
In terms of model performance, Stanford found “over the past year, the gap has fluctuated between near parity and low single digits”. Stanford HAI Department declined to comment for this story.
The gap has remained narrow for some time, said Kyle Chan, a fellow of the John L Thornton China Centre of the Brookings Institution, and is “really significant”, especially given the large compute gap between US labs and Chinese labs. Compute is the processing power used to train and run AI models.
“That to me is quite stunning that even despite that, the Chinese models can basically keep up,” said Chan, who testified in Congress recently about China’s AI capabilities.
The US’ lead in model capabilities over China has shrunk from years to months, said Scott Singer, a fellow at the Technology and International Affairs Programme at the Carnegie Endowment for International Peace, but whether China will be able to close the gap fully remains unclear.
“China is home to the world’s leading open-weight model providers. Developers in places ranging from Shenzhen to San Francisco are building on top of Chinese models like Alibaba’s Qwen to create new products,” Singer said.
“But the US models remain the world’s most capable and appear to be the choice of most users internationally.” Alibaba is the owner of the South China Morning Post.

“My read of the data is that the benchmark performance of Chinese models has trailed frontier US models by about five to six months,” said Miller.
But Anthropic’s powerful new AI model, Claude Mythos Preview, could widen that gap again, underscoring the rivalry’s dynamic nature.
Mythos is stoking global cybersecurity fears, and China lacks access to it. Announced in early April, Mythos was not included in Stanford’s AI index.
“One could argue that Mythos is expanding the US-China AI gap, at least in terms of capabilities,” said Miller.
The model has drawn attention for its ability to autonomously identify and exploit cybersecurity vulnerabilities, seemingly exceeding prior AI capabilities. Mythos has not been made publicly available.
Yet, the standardised test used to compare AI systems is not the only indicator. “That is one of many metrics you can use to gauge that gap, but it doesn’t paint the whole picture,” Miller said.
The Stanford report showed that while the US is investing more in AI, its ability to attract talent from overseas is declining.
This comes amid stricter immigration policies under Trump, which have contributed to a slowdown in the number of highly skilled foreigners heading to America.
While the US was found to be a clear leader in entrepreneurial activity in 2025 with 1,953 newly funded AI companies, over 10 times the next closest country, “the number of AI researchers and developers moving to the US has dropped 89 per cent since 2017, with an 80 per cent decline in the last year alone”, according to the Stanford report.
Spending is another benchmark, although figures here are a little murky. US private AI investment reached US$285.9 billion in 2025, more than 23 times the US$12.4 billion invested in China, according to Stanford, which wrote that “looking at just private investment figures likely understates China’s total AI spending, given its government guidance funds”.
Future of AI in US, China face domestic constraints
Singer pointed to adoption and public trust as another area of importance, noting that both the US and China face domestic constraints. “Both countries will need to convince their publics that building better AI will improve the lives of everyday people in the long run.”
A November poll by Edelman found a trust gap. In China, 87 per cent said they trusted AI, compared to 32 per cent in the US. Meanwhile, 54 per cent in China said they embraced greater use of AI, compared with 17 per cent in the US.
Despite this, China’s AI supremacy is far from assured. Miller said China is one of only two countries that can credibly claim to be competing at the frontier of AI development, but how that translates into real-world influence could depend on what happens domestically.
“Chinese companies will need to find a way to remain profitable, navigate potential novel risks that come from misuse of open-weight models, and secure compute amid export controls,” he added.
Without this, China will remain an important player globally, but Miller said there will be limits to what it can offer.
Chan thinks the Chinese side is keen to find areas of AI cooperation with the US, but what that means exactly is less clear.
“Mythos has really captured attention in Washington over the potential AI risks around cybersecurity. The problem is that there is very low trust and very high competition between the US and China on AI,” he said.
“But there could be some symbolic gestures towards AI cooperation coming out of this summit.”


China and US in AI revolution race as tech rivals battle for global supremacy

China and the US are set to fight out an intense battle for global AI supremacy in a high-stakes race to the top. Photo: Shutterstock 

Qwen was created by Alibaba as a personal AI assistant for professional and personal use. Photo: Shutterstock

News

China

Diplomacy

Two countries will lock horns for foreseeable future, with key industry certain to be a point of negotiation at next week’s Trump-Xi summit


Two countries will lock horns for foreseeable future, with key industry certain to be a point of negotiation at next week’s Trump-Xi summit.


China and US in AI revolution race as tech rivals battle for global supremacy

.css-1c6uqr6{color:inherit;font-weight:inherit;font-size:inherit;font-family:inherit;line-height:inherit;overflow-wrap:break-word;}Two countries will lock horns for foreseeable future, with key industry certain to be a point of negotiation at next week’s Trump-Xi summit

Two countries will lock horns for foreseeable future, with key industry certain to be a point of negotiation at next week’s Trump-Xi summit