China would love a domestic Nvidia rival — but that’s proving quite the challenge
The flags of China and the USA are being displayed on a smartphone, with an NVIDIA chip visible in the background.
Raa | Nurphoto | Getty Images
Chinese companies are ramping up efforts to produce a viable alternative to Nvidia’s chips that power artificial intelligence as Beijing continues its efforts to wean itself off American technology.
U.S. sanctions slapped on China over the past few years, along with Nvidia‘s dominance in the space, have provided big challenges for Bejing’s efforts, at least in the short term, analysts told CNBC.
Nvidia’s well-documented boom has been driven by large cloud computing players buying its server products which contain its graphics processing units, or GPUs. These chips are enabling companies, such as ChatGPT maker OpenAI, to train their huge AI models on massive amounts of data.
These AI models are fundamental to applications like chatbots and other emerging AI applications.
The U.S. government has restricted the export of Nvidia’s most advanced chips to China since 2022, with restrictions tightening last year.
Such semiconductors are key to China’s ambitions to become a leading AI player.
CNBC spoke to analysts who identified some of China’s leading contenders that are looking to challenge Nvidia, including technology giants Huawei, Alibaba and Baidu and startups such as Biren Technology and Enflame.
The overarching view is that they are lagging behind Nvidia at this point.
“These companies have made notable progress in developing AI chips tailored to specific applications (ASICs),” Wei Sun, a senior analyst at Counterpoint Research, told CNBC.
“However, competing with Nvidia still presents substantial challenges in technological gaps, especially in general-purpose GPU. Matching Nvidia in short-term is unlikely.”
China’s key challenges
Chinese firms have a “lack of technology expertise”, according to Sun, highlighting one of the challenges.
However, it’s the U.S. sanctions and their knock-on effects that pose the biggest roadblocks to China’s ambitions.
Some of China’s leading Nvidia challengers have been placed on the U.S. Entity List, a blacklist which restricts their access to American technology. Meanwhile, a number of U.S. curbs have restricted key AI-related semiconductors and machinery from being exported to China.
China’s GPU players all design chips and rely on a manufacturing company to produce their chips. For a while, this would have been Taiwan Semiconductor Manufacturing Co., or TSMC. But U.S. restrictions mean many of these firms cannot access the chips made by TSMC.
They therefore have to turn to SMIC, China’s biggest chipmaker, whose technology remains generations behind TSMC. Part of the reason why it’s lagging behind, is because Washington has restricted SMIC’s access to a key piece of machinery from Dutch firm ASML, which is required to manufacture the most advance chips.
Meanwhile, Huawei has been pushing development of more advanced chips for its smartphones and AI chips, which is taking up capacity at SMIC, according to Paul Triolo, a partner at consulting firm Albright Stonebridge.
“The key bottleneck will be domestic foundry leader SMIC, which will have a complex problem of dividing limited resources for its advanced node production between Huawei, which is taking up the lion’s share currently, the GPU startups, and many other Chinese design firms which have been or may be cutoff from using global foundry leader TSMC to manufacture their advanced designs,” Triolo told CNBC.
Nvidia is more than just GPUs
Nvidia has found success due to its advanced semiconductors, but also with its CUDA software platform that allows developers to create applications to run on the U.S. chipmaker’s hardware. This has led to the development of a so-called ecosystem around Nvidia’s products that others might find hard to replicate.
“This is the key, it is not just about the hardware, but about the overall ecosystem, tools for developers, and the ability to continue to evolve this ecosystem going forward as the technology advances,” Triolo said.
Huawei leading the pack
Triolo identified Huawei as one of the leaders in China with its Ascend series of data center processors.
The firm’s current generation of chip is called the Ascend 910B, and the company is gearing up to launch the Ascend 910C, which could be on par with Nvidia’s H100 product, according to a Wall Street Journal report in August.
In its annual report earlier this year, Nvidia explicitly identified Huawei, among other companies, as a competitor in areas such as chips, software for AI and networking products.
In the area of software and building a developer community, Huawei “holds lots of advantages,” Triolo said. But it faces similar challenges to the rest of the industry in trying to compete with Nvidia.
“The GPU software support ecosystem is much more entrenched around Nvidia and to a lesser degree AMD, and Huawei faces major challenges, both in producing sufficient quantities of advanced GPUs such as part of the Ascend 910C, and continuing to innovate and improve the performance of the hardware, given U.S. export controls that are limiting the ability of SMIC to produce advanced semiconductors,” Triolo said.
Chip IPOs ahead?
The challenges facing China’s Nvidia competitors have been evident over the past two years. In 2022, Biren Technology carried out a round of layoffs, followed by Moore Threads the year after, with both companies blaming U.S. sanctions.
But startups are still holding out hope, looking to raise money to fund their goals. Bloomberg reported last week that Enflame and Biren are both looking to go public to raise money.
“Biren and the other GPU startups are staffed with experienced industry personnel from Nvidia, AMD, and other leading western semiconductor companies, but they have the additional challenge of lacking the financial depth that Huawei has,” Triolo said.
“Hence both Biren and Enflame are seeking IPOs in Hong Kong, to raise funding for additional hiring and expansion.”
Beta Technologies shares surged more than 9% after air taxi maker Eve Air Mobility announced an up to $1 billion deal to buy motors from the Vermont-based company.
Eve, which was started by Brazilian airplane maker Embraer and is now under Eve Holding, said the manufacturing deal could equal as much as $1 billion over 10 years. The Florida-based company said it has a backlog of 2,800 vehicles.
Shares of Eve Holding gained 14%.
Eve CEO Johann Bordais called the deal a “pivotal milestone” in the advancement of the company’s electric vertical takeoff and landing, or eVTOL, technology.
“Their electric motor technology will play a critical role in powering our aircraft during cruise, supporting the maturity of our propulsion architecture as we progress toward entry into service,” he said in a release.
Mateusz Slodkowski | SOPA Images | Lightrocket | Getty Images
Amazon’s cloud unit on Tuesday announced AI-enabled software designed to help clients better understand and recover from outages.
DevOps Agent, as the artificial intelligence tool from Amazon Web Services is called, predicts the cause of technical hiccups using input from third-party tools such as Datadog and Dynatrace. AWS said customers can sign up to use the tool Tuesday in a preview, before Amazon starts charging for the service.
The AI outage tool from AWS is intended to help companies more quickly figure out what caused an outage and implement fixes, Swami Sivasubramanian, vice president of agentic AI at AWS, told CNBC. It’s what site reliability engineers, or SREs, do at many companies that provide online services.
SREs try to prevent downtime and jump into action during live incidents. Startups such as Resolve and Traversal have started marketing AI assistants for these experts. Microsoft’s Azure cloud group introduced an SRE Agent in May.
Rather than waiting for on-call staff members to figure out what happened, the AWS DevOps Agent automatically assigns work to agents that look into different hypotheses, Sivasubramanian said.
“By the time the on-call ops team member dials in, they have an incident report with preliminary investigation of what could be the likely outcome, and then suggest what could be the remediation as well,” Sivasubramanian told CNBC ahead of AWS’ Reinvent conference in Las Vegas this week.
Commonwealth Bank of Australia has tested the AWS DevOps Agent. In under 15 minutes, the software found the root cause of an issue that would have taken a veteran engineer hours, AWS said in a statement.
The tool relies on Amazon’s in-house AI models and those from other providers, a spokesperson said.
AWS has been selling software in addition to raw infrastructure for many years. Amazon was early to start renting out server space and storage to developers since the mid-2000s, and technology companies such as Google, Microsoft and Oracle have followed.
Since the launch of ChatGPT in 2022, these cloud infrastructure providers have been trying to demonstrate how generative AI models, which are often training in large cloud computing data centers, can speed up work for software developers.
Over the summer, Amazon announced Kiro, a so-called vibe coding tool that produces and modifies source code based on user text prompts. In November, Google debuted similar software for individual software developers called Antigravity, and Microsoft sells subscriptions to GitHub Copilot.
WATCH: Amazon rolls out AI-powered tools to help big AWS customers update old software
Technology
Amazon to let cloud clients customize AI models midway through training for $100,000 a year
Attendees pass an Amazon Web Services logo during AWS re:Invent 2024, a conference hosted by Amazon Web Services, at The Venetian hotel in Las Vegas on Dec. 3, 2024.
Noah Berger | Getty Images
Amazon has found a way to let cloud clients extensively customize generative AI models. The catch is that the system costs $100,000 per year.
The Nova Forge offering from Amazon Web Services gives organizations access to Amazon’s AI models in various stages of training so they can incorporate their own data earlier in the process.
Already, companies can fine-tune large language models after they’ve been trained. The results with Nova Forge will lean more heavily on the data that customers supply. Nova Forge customers will also have the option to refine open-weight models, but training data and computing infrastructure are not included.
Organizations that assemble their own models might end up spending hundreds of millions or billions of dollars, which means using Nova Forge is more affordable, Amazon said.
AWS released its own models under the Nova brand in 2024, but they aren’t the first choice for most software developers. A July survey from Menlo Ventures said that by the middle of this year, Amazon-backed Anthropic controlled 32% of the market for enterprise LLMs, followed by OpenAI with 25%, Google with 20% and Meta with 9% — Amazon Nova had a less than 5% share, a Menlo spokesperson said.
The Nova models are available through AWS’ Bedrock service for running models on Amazon cloud infrastructure, as are Anthropic’s Claude 4.5 models.
“We are a frontier lab that has focused on customers,” Rohit Prasad, Amazon head scientist for artificial general intelligence, told CNBC in an interview. “Our customers wanted it. We have invented on their behalf to make this happen.”
Nova Forge is also in use by internal Amazon customers, including teams that work on the company’s stores and the Alexa AI assistant, Prasad said.
Reddit needed an AI model for moderating content that would be sophisticated about the many subjects people discuss on the social network. Engineers found that a Nova model enhanced with Reddit data through Forge performed better than commercially available large-scale models, Prasad said. Booking.com, Nimbus Therapeutics, the Nomura Research Institute and Sony are also building models with Forge, Amazon said.
Organizations can request that Amazon engineers help them build their Forge models, but that assistance is not included in the new service’s $100,000 annual fee.
AWS is also introducing new models for developers at its Reinvent conference in Las Vegas this week.
Nova 2 Pro is a reasoning model whose tests show it performs at least as well as Anthropic’s Claude Sonnet 4.5, OpenAI’s GPT-5 and GPT-5.1, and Google’s Gemini 3.0 Pro Preview, Amazon said. Reasoning involves running a series of computations that might take extra time in response to requests to produce better answers. Nova 2 Pro will be available in early access to AWS customers with Forge subscriptions, Prasad said. That means Forge customers and Amazon engineers will be able to try Nova 2 Pro at the same time.
Nova 2 Omni is another reasoning model that can process incoming images, speech, text and videos, and it generates images and text. It’s the first reasoning model with that range of capability, Amazon said. Amazon hopes that, by delivering a multifaceted model, it can lower the cost and complexity of incorporating AI models into applications.
Tens of thousands of organizations are using Nova models each week, Prasad said. AWS has said it has millions of customers. Nova is the second-most popular family of models in Bedrock, Prasad said. The top group of models are from Anthropic.
WATCH: Amazon set to kick off AI conference next week: Maxim’s Tom Forte on what to expect
-
Sports2 years agoStory injured on diving stop, exits Red Sox game
-
Sports3 years ago‘Storybook stuff’: Inside the night Bryce Harper sent the Phillies to the World Series
-
Sports2 years agoGame 1 of WS least-watched in recorded history
-
Sports3 years agoButton battles heat exhaustion in NASCAR debut
-
Sports3 years agoMLB Rank 2023: Ranking baseball’s top 100 players
-
Sports4 years ago
Team Europe easily wins 4th straight Laver Cup
-
Environment3 years agoJapan and South Korea have a lot at stake in a free and open South China Sea
-
Environment1 year agoHere are the best electric bikes you can buy at every price level in October 2024