(Bloomberg) -- Alibaba Group Holding Ltd. hasn’t been able to completely fulfill demand for AI training from clients because of global supply constraints, its top executive said, suggesting a shortage of critical components such as artificial intelligence chips is weighing on Chinese efforts to ramp up in the cutting-edge technology.

Most Read from Bloomberg

“In the past quarter, we have received strong demand for model training and related services on cloud infrastructure, which were only partially fulfilled due to the near-term supply chain constraints globally,” Chairman and Chief Executive Officer Daniel Zhang, who steps down in September, told analysts on a conference call. He will focus on Alibaba’s cloud business full-time after ceding his dual roles to Alibaba co-founders Joseph Tsai and Eddie Wu.

A shortage of high-powered semiconductors is undermining Chinese efforts to keep pace with the US in AI. Washington has banned Chinese firms from buying the most advanced chips made by Nvidia Corp., impeding attempts to build rivals to OpenAI’s ChatGPT. Nvidia has since created an inferior version of its most potent A100 chips for the country, and major Chinese tech firms including Alibaba have reportedly placed billions of dollars’ worth of orders.

The Biden administration on Wednesday announced new rules to partially limit American firms’ investments in quantum computing and AI sectors in China, which could further disadvantage the Asian country in its ambition to advance its technological capabilities.

Read more: Alibaba Takes a Step Toward Comeback As Growth Finally Returns

Most Read from Bloomberg Businessweek

©2023 Bloomberg L.P.

Jidoka Technologies, among the pioneers in the field of automated cognitive inspection for manufacturing is happy to announce the launch of its innovative self-training software designed to revolutionize the AI-based object detection techniques.

In the era of rapidly evolving technologies, visual quality in manufacturing is not always binary – OK or NOK, and is highly subjective. It relies on people’s experience and expertise for decisions that need to be taken within seconds despite the presence of rules and SOP.

The criteria for accepting or rejecting defects in components can fluctuate due to factors like environmental conditions, the influence of external elements such as oil and dust on the part, and changes in standards based on the part’s functionality and its intended user. Although these variations are in the acceptable range, automated machine systems that rely on rule-based approaches would be rejecting the parts leading to a high percentage of false positives.

For over three years, Jidoka Technologies has been deploying its AI-driven software to address complex visual inspection tasks, with a track record of over 40 success stories spanning across automotive, consumer goods, print, and packaging industries. Customers have been asking for self-training for two usecases, where there is a change in environment or variations for existing lines or where there is a high mix-low volume eg: PCB industry where the requirements are dynamically changing. By building on this rich experience, the organization is now introducing its self-training software, empowering customers with the ability to independently train and deploy AI models on the go.

Speaking during the launch, Sekar Udayamurthy, CEO and Co-founder, of Jidoka Technologies, said, “We are happy to introduce our revolutionary self-training software for AI-based object detection in the visual inspection domain.  This technology is a game changer enabling the manufacturers to achieve quality inspection independence and ensure their own teams perform quality control of their ever-changing production lines quickly and effortlessly.” He also added. “With this new technology, manufacturers can attain an unmatched level of accuracy, efficiency, and adaptability in their quality inspection processes and reduce the dependency on Jidoka for changes”

This state-of-the-art software leverages the power of advanced AI technologies with continuous learning from different datasets and adapting to variations to enhance the accuracy and efficiency of visual inspections to unprecedented levels either to automated means or “human in the loop”.  The power to train AI models by the end user allows manufacturers to upskill their teams, manage the defects and work on timely upgrades based on their process and criteria changes.  Furthermore, this software significantly reduces the implementation time needed to take cutting-edge AI deployments live in production environments.  The user interface guides users with no machine vision expertise, step by step in the training of new defects and new products to AI, making the process extremely intuitive.

Also Read: Cloud Computing Trends to Look for in 2024

Jidoka works with manufacturers across automotive, FMCG, pharma, general manufacturing, electronics, textiles, and printing industry domains currently and will expand to other verticals, in the future. Its key list of customers includes Mudhra Fine Blanc Private Ltd, Nexteer Automotive, IP Rings, ZF Rane, Sansera Engineering, IM Gears in the automotive sector, Shriji Polymers in the Pharma sector, and ITC Ltd. and partners of Mondelez International, Marico in the FMCG sector.

Check Out The New Enterprisetalk Podcast. For more such updates follow us on Google News Enterprisetalk News.

Manufacturers, of course, know about this range anxiety and have every incentive to comfort drivers with the knowledge that the range of their current and future EVs will be more than sufficient to meet the needs of the vast majority of drivers. This urge to comfort drivers about EV range, however, could create downstream complications for manufacturers because of past legal cases involving the advertising of fuel economy and the range of vehicles equipped with internal combustion engines. These potential issues have their roots in the statutory origins of the Environmental Protection Agency's (EPA) fuel economy estimates, as well as the Federal Trade Commission's (FTC) regulations governing the disclosure of fuel economy estimates and their use in advertising.

Regulatory Background

The testing and disclosure of estimated fuel economy for new vehicles sold in the United States are governed by a comprehensive federal regulatory scheme – developed by Congress through the Energy Policy and Conservation Act of 1975 (EPCA) – administered by the EPA, and the FTC. Under that regulatory scheme, every new vehicle below 8,500 GVWR sold in the United States must be equipped with a "Monroney" label – or window sticker – that sets forth the fuel economy estimates derived, and required to be disclosed, under EPA regulations. The EPA refers to the figures on a Monroney label as fuel economy "estimates" because they are, and are intended to be just that: approximate figures, generated for the purpose of enabling comparisons between different vehicles based on a common certification process.¹

Methods for calculating city and highway fuel economy have been in place since the 1970s, and EPA estimates have appeared on the window stickers of new automobiles sold in the U.S. since the latter part of that decade.² Although the methods for estimating fuel economy for internal combustion engines and electric vehicles are distinct, the purpose of the estimates is the same. EPA fuel economy estimates are not, and have never been, guarantees of real-world fuel economy performance, whether for internal combustion vehicles or electric vehicles. As the EPA itself has stressed, its fuel economy "ratings are a useful tool for comparing the fuel economies of different vehicles but may not accurately predict the average [miles per gallon] you will get." Indeed, a vehicle's fuel economy will vary.³ For this reason, when designing the Monroney label, regulators required that it contain a "statement . . . informing the buyer that the values on the label are not guaranteed[.]"⁴ And as the EPA has long acknowledged, its required fuel economy estimates are not – and can never be – "perfect" figures that can predict the performance of each vehicle for each driver under all conditions:

It is important to emphasize that fuel economy varies from driver to driver for a wide variety of reasons, such as different driving styles, climates, traffic patterns, use of accessories, loads, weather, and vehicle maintenance. Even different drivers of the same vehicle will experience different fuel economy as these and other factors vary. Therefore, it is impossible to design a "perfect" fuel economy test that will provide accurate, real-world fuel economy estimates for every consumer. With any estimate, there will always be consumers that get better or worse actual fuel economy. The EPA estimates are meant to be a general guideline for consumers, particularly to compare the relative fuel economy of one vehicle to another.⁵

While acknowledging no single test can produce a prediction of fuel economy for all drivers, the EPA has still recognized a need for consumers to have some quantifiable information about fuel economy, so that comparisons can be made between vehicles.⁶ This is why the EPA requires manufacturers to use fuel economy estimates derived from the complex testing regime described above. At the same time, the EPA has openly cautioned that using different procedures, such as on-road testing, will often yield different results than those obtained under the EPA's testing process.⁷ Even in the lab, the EPA recognizes different test equipment can lead to different results.⁸

In addition to requiring that manufacturers determine fuel economy estimates pursuant to its detailed testing procedures and calculations, the EPA also requires that manufacturers post those estimates on Monroney labels. The precise form and content of that label are fixed in exacting detail by federal law and EPA regulation.⁹ These labels clearly state, in federally mandated language, that "[a]ctual results will vary for many reasons, including driving conditions and how you drive and maintain your vehicle." 40 C.F.R. 600.302-12(b)(4). That language was carefully selected by the EPA following an extensive rulemaking process that included public comment and focus group testing of alternative language.¹⁰ During that process, the EPA acknowledged "[a]ll factors that impact fuel economy cannot be listed on the fuel economy label because they are too numerous."¹¹

The fuel economy disclosures on the Monroney label are slightly different between internal combustion and electric vehicles. Significantly, the internal combustion label references combined city and highway fuel economy values, derived from the mandatory EPA testing procedure – but the internal combustion label does not include any explicit statement about vehicle range:¹²

The label for an electric vehicle is similar but modified in several significant respects owing to the difference in how the vehicles are powered. The full electric EV label¹³ includes "MPGe" values for combined, city, and highway driving. MPGe is "miles per gasoline gallon equivalent," and is used to allow for performance comparisons between internal combustion vehicles and electric vehicles applying a metric that is familiar to consumers. Importantly for purposes of navigating potential legal issues associated with referencing electric vehicle range, the Monroney label for electric vehicles does include an explicit driving range disclosure, stating "when fully fueled, the vehicle can travel about . . ." X miles:

In furtherance of the federal government's objective to provide consistent fuel economy information to consumers, the FTC regulates the advertising of fuel economy estimates by manufacturers and sellers. The FTC states "[i]t is deceptive to misrepresent, directly or by implication, the fuel economy or driving range of an automobile."¹⁴ "Because it is highly unlikely that advertisers can substantiate all reasonable interpretations of these [fuel economy claims], advertisers making general fuel economy claims should disclose the advertised vehicle's EPA fuel economy estimate in the form of the EPA MPG rating."¹⁵ Accordingly, any advertisement that references "EPA estimates" or "equivalent language that informs consumers that they will not necessarily achieve the stated MPG rating or driving range" is sufficient to satisfy FTC requirements.¹⁶ To that end, the FTC also requires the estimates be "clear and prominent . . . in close proximity to [a] qualified claim."¹⁷ The FTC warns that "[f]ailure to comply with [its] guides may result in corrective action by the Commission under applicable statutory provisions."¹⁸

History of Lawsuits Based on Fuel Economy Claims

There have been several lawsuits filed through the years relating to fuel economy claims. Those lawsuits have generally been unsuccessful to the extent that the claims directly challenged the accuracy of EPA label values. Historically the leading case is Paduano v. American Honda Motor Co., Inc.,¹⁹ where the consumer complained that his 2004 Civic Hybrid's fuel economy performance was half his EPA estimated values. The trial court found that claim to be preempted by federal law, and the appellate court agreed that the plaintiff "may not directly challenge the accuracy of the EPA estimates by way of state law causes of action."²⁰ The appellate court further held that a statement about fuel economy based on the EPA estimates, whether on the Monroney label or in an advertisement, "does not constitute an independent warranty that [plaintiff's] vehicle would achieve the EPA fuel economy estimates or a similar level of fuel economy."²¹

But Paduano also held that consumer protection claims that challenged fuel economy could proceed to the extent the claim went "beyond the label" and included suggestions by the manufacturer or seller that the EPA estimates were achievable in the ordinary use of the vehicle.²² The court said such claims were actionable where the manufacturer "has voluntarily made additional assertions, beyond the disclosure of the mileage estimates, that are untrue or misleading, and the federal law does not require, or even address, these additional assertions."²³

Many other cases followed the guidance of Paduano, dismissing claims where they directly challenged the EPA label values,²⁴ and permitting claims to proceed where the claims involved what the courts characterized as "beyond the label" representations or statements.²⁵

EV Vehicle Range Claims: Pitfalls and Protections

Based on the foregoing regulatory and legal background, advertising and marketing intended to reassure consumers about the range of EVs needs to be carefully crafted. Representations about real-world performance are highly likely to be viewed by courts as beyond the protection of the FTC guidance, and thus potentially actionable. When vehicle range is discussed, it is important to make clear to the reader that the range addressed in the advertisement or marketing statement is the EPA-estimated range, and to include the disclaimer that actual mileage and performance will vary.

EVs do have one major advantage over internal combustion vehicles in terms of discussion of vehicle range. EVs have vehicle range incorporated as an explicit element of their mandatory disclosures, whereas internal combustion vehicles do not. Thus, EV manufacturers and sellers are better positioned to argue that references to EV vehicle range are "on the label" disclosures that are entitled to the same protection as the use of EPA-estimated miles per gallon statements under Paduano, Gray, and the many other cases finding repeating of EPA estimates as non-actionable. Superimposing range estimates – even EPA-approved range estimates – on maps or other similar graphic representations of distance should probably be discouraged so as not to be construed as an implied representation of achievability in the real world.

EVs represent a significant and growing segment of the automotive market and addressing range anxiety is an important element of speeding their adoption by consumers. But prudent manufacturers who wish to avoid later claims from disappointed consumer expectations of vehicle range should be diligent in supporting their advertising and public-facing statements of vehicle range with references to EPA estimates. Staying "on the label" will help those manufacturers and sellers reduce the risk of individual or consumer class actions claims on fuel economy or range representations.

¹ See 71 Fed. Reg. 77872, 77874 (Dec. 27, 2006) ("We believe the new fuel economy estimates will provide car buyers with useful information when comparing the fuel economy of different vehicles."); see also 76 Fed. Reg. 39478, 39505 (Jul. 6, 2011) (adopting a redesigned fuel economy label but continuing a tradition of having a statement on the label informing the buyer that the values on the label are not guaranteed).

² See 71 Fed. Reg. at 77873–74.

³ Id.

⁴ 76 Fed. Reg. at 39505.

⁵ 71 Fed. Reg. at 77874; see also 76 Fed. Reg. at 39505 (emphasizing "tradition" of ensuring consumers know estimates do not reflect real-world economy).

⁶ See 71 Fed. Reg. at 77874.

⁷ Id. at 77874, 77879.

⁸ See 79 Fed. Reg. 23537(Apr. 28, 2014) (fuel economy variability between two-and four-wheel dynamometer for certification testing).

⁹ See 49 U.S.C. § 32908(a); 40 C.F.R. § 600.302-12.

¹⁰ See 76 Fed. Reg. at 39482; see also 71 Fed. Reg. at 77903.

¹¹ 71 Fed. Reg. at 77903.

¹² While the internal combustion Monroney label does not include a range statement, it does cross-reference fueleconomy.gov – the EPA's consumer-facing website for vehicle fuel economy information – that does provide a vehicle range via simple multiplication of the vehicle tank size by combined EPA mileage rating.

¹³ There are different labels for plug-in hybrid vehicles, that utilize both electric motors and gasoline motors.

¹⁴ 16 C.F.R. §259.4(a) (emphasis added).

¹⁵ Id. at §259.4(b) (emphasis added).

¹⁶ Id. at §259.4(d), (e) (emphasis added).

¹⁷ Id.

¹⁸ Id. at §1.5

¹⁹ 169 Cal. App. 4th 1453 (Cal. App. 2009).

²⁰ Id. at 1468 n.9.

²¹ Id. at 1467.

²² Id. at 1477.

²³ Id.

²⁴ See Gray v. Toyota Motor Sales, U.S.A, Inc., 2012 WL 313703, at * 5 (C.D. Cal. Jan. 23, 2012), aff'd, 554 F. App'x 608 (9th Cir. 2014) ("[T]he claims must fail as they rely solely on advertisements that merely repeat the approved EPA mileage estimates, without any additional representations as to, for example, a consumer's ability to achieve those figures under normal driving conditions."); Jarvis v. BMW of North America, 2015 WL 2201690 (M.D. Fla. May 11, 2015) (dismissing complaint challenging advertisements that repeated EPA label values); In re Ford Fusion & C-MAX Fuel Economy Litigation, 2015 WL 7018369, at *21, 30 (S.D.N.Y. Nov. 12, 2015) (rejection direct challenges to EPA label values as preempted or within primary jurisdiction of EPA); In re Ford Motor Co. F-150 & Ranger Truck Fuel Economy Marketing & Sales Prac. Litig., 65 F.4th 851 (6th Cir. 2023) (finding direct challenges to Monroney sticker and EPA values impliedly preempted); Espinosa v. Hyundai Motor America, 2012 U.S. Dist. LEXIS 191088, at * 5 (C.D. Cal. Apr. 23, 2012) ("To the extent that Plaintiff's claims rest on Defendant's mere use of the EPA estimates and all of the federally-mandated disclosures in their advertising and marketing materials, such claims are preempted.").

²⁵ See In re Ford Fusion & C-MAX Fuel Economy Litig., 2015 WL 7018369, at *33 (allowing "beyond the label" claims "to the extent that Plaintiffs have referenced specific ads that made specific promises as to the real-world performance of the Vehicles") (emphasis in original); Espinosa v. Hyundai Motor America, 2012 U.S. Dist. LEXIS 191088, at * 6–7 (C.D. Cal. Apr. 23, 2012) ("[T]o the extent that Plaintiff's claims rest on allegations that Hyundai voluntarily made additional assertions, beyond the disclosure of mileage estimates, that are untrue or misleading, and that federal law does not require, or even address, these additional assertions, Plaintiff's claims are not preempted.") (internal quotations omitted); Kim v. General Motors, LLC, 99 F. Supp. 3d 1096 (C.D. Cal.2015) (court permitting claims "because [the advertisement] implies that a consumer will be able to actually achieve the EPA fuel economy figures when driving in the real world," and noting the use of "a real world map to emphasize the point" as part of the basis for its ruling).

Summary

For many industrial manufacturers, time is running out to digitally transform and update or migrate their legacy distributed control systems (DCSs). The shelf life on these systems is expiring and taking a wait-and-see approach is no longer a viable option. These systems are overwhelmingly fragmented and fraught with challenges. Manufacturers risk losing operational control of their processes.

Compound these issues with the lack of available resources to maintain or repair older equipment and the potential for cybersecurity attacks, safety and environmental risks increase. Manufacturers must change to keep pace with ever-evolving technology and consumer demands; otherwise, they risk becoming uncompetitive in their markets.

Opportunity abounds for those who want improved asset utilization, access to real-time data, improved control and enhanced connectivity. Manufacturers need to reevaluate existing operations and leverage innovative technologies that deliver the promise of greater interconnectivity and system visibility across the enterprise. Consider the following features and functions where a new and improved DCS can make a competitive impact now and in the future.
 

Open communication

For many manufacturers, the underlying control strategies in the DCS haven’t really changed much. A modern DCS, however, ensures open communication to smart field devices, subsystems and higher-level enterprise resource planning (ERP) and manufacturing execution systems (MESs), making real-time data easily accessible across the enterprise as it comes directly from the system controlling the facility.

Diagnostic information about–and calibration of–the facility’s instruments, for example, is often now directly available from a DCS workstation without the need for third-party asset management systems or communicators (e.g., HART). This feature can result in a large cost savings for manufacturers in implementation and maintenance. The ease of integration into MES and ERP systems elevates the DCS from being a system that not only operates the facility but also can be used as a key component in managing the business at the corporate level.

Electrical Substation Control System Integration

Today’s DCSs are increasingly able to communicate over the IEC 61850 protocol – the common standard used for network communications in electrical substations. This means that personnel can access and operate a facility’s process controls and the electrical substation controls from a single DCS portal, rather than having separate points of entry for process and electrical controls. This also means there can be tighter integration between process and electrical controls, such as load shedding based on process upsets, without the need for sometimes unreliable communications between distinct electrical and process control systems.
 

Faster response to abnormal situations

The modern DCS has much more capabilities when it comes to human-machine interfaces (HMIs) and graphics techniques where only the most critical information is provided. Operators can more effectively facilitate, identify and respond to abnormal situations. Rather than having to understand and tediously navigate multiple menus required in legacy systems, modern high-performance HMIs call attention to a problem before it escalates, allowing operators to easily jump to where they need to be as process problems arise. Some DCSs have tools to automatically generate visualization of logic, providing operators with more troubleshooting tools to resolve problems and safely continue production.
 

Alarm management/analysis

Many of today’s DCSs have built-in or add-on alarm management and analysis packages. These systems help suppress alarm floods or nuisance alarms and let personnel measure the health of an alarm management system in terms of identifying top bad actor alarms, frequency of alarms, etc. Where this capability did exist in the past, it required integrating the DCS with a third-party alarm management package. A modern system requires little setup beyond simply activating the feature.
 

Advanced controls

Many DCSs today have various forms of multivariable, advanced control built directly into them. This allows manufacturers to do small-scale advanced process control with the DCS they already have rather than requiring a separate and expensive model predictive control platform.
 

Virtualization/simulation

Nearly all DCSs today can be run on virtual servers, resulting in better reliability, portability and disaster recovery. Live migration of virtual machines will move the facility’s servers to a new physical host in the case of catastrophic hardware failure without any noticeable impact to operations.

Simulation tools and software help personnel gain hands-on, practical training in a controlled environment, mitigating safety risks. Virtualization technology makes it easier and cheaper to maintain a stand-alone virtual DCS used for simulated operations training and program development.
 

The next-generation engineer

As many legacy DCSs were built on proprietary hardware and software platforms, up-and-coming engineers will have little familiarity or experience with them. These engineers will better understand the Windows-based platforms that are the backbone of nearly all modern DCSs.

This presents a challenge for manufacturers trying to recruit and retain new engineers that won’t want to work on an aging infrastructure. These engineers will better relate to a more integrated, modern environment where they can learn cutting-edge software and applications (e.g., simulation tools, cloud and edge computing, smart manufacturing tools and so on) to help them move forward in their career.
 

The DCS road ahead

The road to a more modern DCS may not be easy but the features and functionality gained far outweigh any issues. In today’s competitive world, it’s time for manufacturers to rethink what capabilities they want and expect from a modern DCS (e.g., Rockwell Automation’s PlantPAx):

• Enables plant-wide control and optimization for lower total cost of ownership (TCO)
• Is reliable, smart and provides built-in scalability and modularity for engineering flexibility and faster time to market (e.g., can automatically diagnose and modify systems when changes or upgrades are required)
• Is an open, interoperable platform with remote security and support capabilities that can continuously optimize, maintain and sustain system operations
• Enables new technologies that Strengthen production operation and integration enterprise-wide
• Provides predictive maintenance capabilities (e.g., has built-in software that continually diagnoses and troubleshoots potential issues)
• Provides the ability to use simulation as a training platform for potential safety risks, which helps lower operations and maintenance training costs

The migration process requires an organized and planned approach leveraging a proven methodology (e.g., DCSNext) and best practices. When resource bandwidth is an issue, a third-party partner with industry knowledge in a wide variety of platforms can help overcome potential challenges. Manufacturers must maximize the benefits of a modern DCS to stay ahead of the competition and keep their facilities up and running.

(Bloomberg) -- Alibaba Group Holding Ltd. hasn’t been able to completely fulfill demand for AI training from clients because of global supply constraints, its top executive said, suggesting a shortage of critical components such as artificial intelligence chips is weighing on Chinese efforts to ramp up in the cutting-edge technology.

Most Read from Bloomberg

“In the past quarter, we have received strong demand for model training and related services on cloud infrastructure, which were only partially fulfilled due to the near-term supply chain constraints globally,” Chairman and Chief Executive Officer Daniel Zhang, who steps down in September, told analysts on a conference call. He will focus on Alibaba’s cloud business full-time after ceding his dual roles to Alibaba co-founders Joseph Tsai and Eddie Wu.

A shortage of high-powered semiconductors is undermining Chinese efforts to keep pace with the US in AI. Washington has banned Chinese firms from buying the most advanced chips made by Nvidia Corp., impeding attempts to build rivals to OpenAI’s ChatGPT. Nvidia has since created an inferior version of its most potent A100 chips for the country, and major Chinese tech firms including Alibaba have reportedly placed billions of dollars’ worth of orders.

The Biden administration on Wednesday announced new rules to partially limit American firms’ investments in quantum computing and AI sectors in China, which could further disadvantage the Asian country in its ambition to advance its technological capabilities.

Read more: Alibaba Takes a Step Toward Comeback As Growth Finally Returns

Most Read from Bloomberg Businessweek

©2023 Bloomberg L.P.

(Bloomberg) -- Alibaba Group Holding Ltd. hasn’t been able to completely fulfill demand for AI training from clients because of global supply constraints, its top executive said, suggesting a shortage of critical components such as artificial intelligence chips is weighing on Chinese efforts to ramp up in the cutting-edge technology.

Most Read from Bloomberg

“In the past quarter, we have received strong demand for model training and related services on cloud infrastructure, which were only partially fulfilled due to the near-term supply chain constraints globally,” Chairman and Chief Executive Officer Daniel Zhang, who steps down in September, told analysts on a conference call. He will focus on Alibaba’s cloud business full-time after ceding his dual roles to Alibaba co-founders Joseph Tsai and Eddie Wu.

A shortage of high-powered semiconductors is undermining Chinese efforts to keep pace with the US in AI. Washington has banned Chinese firms from buying the most advanced chips made by Nvidia Corp., impeding attempts to build rivals to OpenAI’s ChatGPT. Nvidia has since created an inferior version of its most potent A100 chips for the country, and major Chinese tech firms including Alibaba have reportedly placed billions of dollars’ worth of orders.

The Biden administration on Wednesday announced new rules to partially limit American firms’ investments in quantum computing and AI sectors in China, which could further disadvantage the Asian country in its ambition to advance its technological capabilities.

Read more: Alibaba Takes a Step Toward Comeback As Growth Finally Returns

Most Read from Bloomberg Businessweek

©2023 Bloomberg L.P.

(Bloomberg) -- Alibaba Group Holding Ltd. hasn’t been able to completely fulfill demand for AI training from clients because of global supply constraints, its top executive said, suggesting a shortage of critical components such as artificial intelligence chips is weighing on Chinese efforts to ramp up in the cutting-edge technology.

Most Read from Bloomberg

“In the past quarter, we have received strong demand for model training and related services on cloud infrastructure, which were only partially fulfilled due to the near-term supply chain constraints globally,” Chairman and Chief Executive Officer Daniel Zhang, who steps down in September, told analysts on a conference call. He will focus on Alibaba’s cloud business full-time after ceding his dual roles to Alibaba co-founders Joseph Tsai and Eddie Wu.

A shortage of high-powered semiconductors is undermining Chinese efforts to keep pace with the US in AI. Washington has banned Chinese firms from buying the most advanced chips made by Nvidia Corp., impeding attempts to build rivals to OpenAI’s ChatGPT. Nvidia has since created an inferior version of its most potent A100 chips for the country, and major Chinese tech firms including Alibaba have reportedly placed billions of dollars’ worth of orders.

The Biden administration on Wednesday announced new rules to partially limit American firms’ investments in quantum computing and AI sectors in China, which could further disadvantage the Asian country in its ambition to advance its technological capabilities.

Read more: Alibaba Takes a Step Toward Comeback As Growth Finally Returns

Most Read from Bloomberg Businessweek

©2023 Bloomberg L.P.

(Bloomberg) -- Alibaba Group Holding Ltd. hasn’t been able to completely fulfill demand for AI training from clients because of global supply constraints, its top executive said, suggesting a shortage of critical components such as artificial intelligence chips is weighing on Chinese efforts to ramp up in the cutting-edge technology.

Most Read from Bloomberg

“In the past quarter, we have received strong demand for model training and related services on cloud infrastructure, which were only partially fulfilled due to the near-term supply chain constraints globally,” Chairman and Chief Executive Officer Daniel Zhang, who steps down in September, told analysts on a conference call. He will focus on Alibaba’s cloud business full-time after ceding his dual roles to Alibaba co-founders Joseph Tsai and Eddie Wu.

A shortage of high-powered semiconductors is undermining Chinese efforts to keep pace with the US in AI. Washington has banned Chinese firms from buying the most advanced chips made by Nvidia Corp., impeding attempts to build rivals to OpenAI’s ChatGPT. Nvidia has since created an inferior version of its most potent A100 chips for the country, and major Chinese tech firms including Alibaba have reportedly placed billions of dollars’ worth of orders.

The Biden administration on Wednesday announced new rules to partially limit American firms’ investments in quantum computing and AI sectors in China, which could further disadvantage the Asian country in its ambition to advance its technological capabilities.

Read more: Alibaba Takes a Step Toward Comeback As Growth Finally Returns

Most Read from Bloomberg Businessweek

©2023 Bloomberg L.P.