Semiconductors

Nvidia introduces high-performance AI machines for the future

At GTC 2025, Nvidia CEO Jensen Huang unveiled a new generation of AI-focused personal supercomputers designed to redefine computing in the era of AI. The two new machines, DGX Spark and DGX Station, are powered by Nvidia’s Grace Blackwell chip platform and promise to deliver unprecedented AI computing power at the edge.

DGX Spark, available immediately, features the GB10 Grace Blackwell Superchip, capable of up to 1,000 trillion operations per second. Meanwhile, the DGX Station, set for release later this year, is built with the GB300 Grace Blackwell Ultra Desktop Superchip and 784GB of memory. According to Nvidia, these supercomputers will allow users to prototype, fine-tune, and deploy AI models with greater efficiency.

Huang described the devices as the future of computing, highlighting their role in supporting AI applications across enterprises. Nvidia has partnered with major manufacturers, including Asus, Dell, HP, and Lenovo, to bring these machines to market. As AI adoption continues to surge, these systems could become essential tools for developers and businesses looking to stay ahead in an increasingly AI-driven world.

For more information on these topics, visit diplomacy.edu.

Nvidia holds back on optical technology for GPUs

Nvidia’s CEO, Jensen Huang, has stated that a promising new chip technology, co-packaged optics, is not yet reliable enough for use in the company’s flagship GPUs.

The technology, which uses laser beams to transfer data via fiber optic cables instead of traditional copper, is more energy-efficient and faster.

However, Huang emphasized that copper connections remain ‘orders of magnitude’ more reliable than today’s optical alternatives, making them the preferred choice for now.

Speaking at Nvidia’s annual developer conference in San Jose, Huang announced that the company will use co-packaged optics in two upcoming networking chips designed for server switches, increasing their energy efficiency by three and a half times.

These switch chips will be released later this year and into 2026, marking a gradual technological step forward. However, Huang clarified that Nvidia currently has no plans to implement optical connections between GPUs, as reliability remains a key priority for its AI-focused customers like OpenAI and Oracle.

Silicon Valley startups such as Ayar Labs, Lightmatter, and Celestial AI have invested heavily in co-packaged optics, seeing it as essential for building more powerful AI systems. Nvidia itself has backed some of these ventures, despite Huang’s cautious approach.

While optical connections could eventually help AI models process complex tasks more efficiently, Nvidia is prioritizing proven technology for its near-term roadmap, ensuring stability in an industry preparing to invest hundreds of billions in AI infrastructure.

For more information on these topics, visit diplomacy.edu.

Google reduces costs with MediaTek in AI chip development

Google is set to collaborate with Taiwan’s MediaTek on the next generation of its Tensor Processing Unit (TPU) chips, which are expected to be produced next year.

The partnership is partly driven by cost considerations, as MediaTek offers Google a lower price per chip than its long-time partner Broadcom. MediaTek’s close ties with Taiwan Semiconductor Manufacturing Company (TSMC) also played a role in Google’s decision.

Despite the new partnership, Google has not severed ties with Broadcom, which has exclusively worked on its AI chips for several years.

Broadcom remains involved in the project, and an employee at the company confirmed that the relationship with Google is still intact. Google has been developing its own AI server chips, allowing it to reduce reliance on Nvidia, whose processors dominate the industry.

Google introduced its sixth-generation TPU last year to provide itself and its cloud customers with an alternative to Nvidia’s highly sought-after chips. The company reportedly spent between $6 billion and $9 billion on TPUs in 2023, based on revenue targets from Broadcom.

By bringing MediaTek into the fold, Google aims to strengthen its AI chip strategy while managing production costs more efficiently.

For more information on these topics, visit diplomacy.edu.

Intel’s new CEO plans major changes to revive the company

Intel’s incoming CEO Lip-Bu Tan is considering major changes to the company’s chip manufacturing and AI strategies to revive the struggling tech giant.

Sources revealed that Tan aims to restructure Intel’s approach to AI and implement staff cuts to streamline operations, focusing on addressing slow-moving middle management.

One of Tan’s core priorities is revamping Intel’s manufacturing operations, which have expanded to include producing semiconductors for external clients like Nvidia.

The changes come as Intel looks to regain its competitive edge after a decade of missed opportunities in smartphone chips and AI processors, allowing competitors such as Arm Holdings and Nvidia to dominate.

At a recent town hall, Tan told employees that the company would need to make ‘tough decisions’ to improve performance. Intel’s shares rose over 8% following his appointment, as investors await further details on his plans.

Tan’s immediate focus includes bolstering Intel Foundry’s performance and attracting new customers in sectors such as AI and robotics.

For more information on these topics, visit diplomacy.edu.

Samsung faces tough shareholder meeting over AI struggles

Samsung Electronics faces a challenging annual general meeting as shareholders express frustration over its failure to capitalise on the AI boom.

Despite being South Korea’s most valuable company, Samsung’s stock tumbled nearly a third last year, making it one of the worst-performing tech firms.

Executives, including Co-CEO Han Jong-hee, will address concerns over lagging innovation, competition in semiconductor technology, and strategies to counter US tariffs.

Internal discussions at Samsung have revealed concerns about losing its technological edge, particularly in high bandwidth memory (HBM) chips, where it trails rival SK Hynix.

Chairman Jay Y. Lee reportedly criticised the company for focusing on maintaining the status quo rather than driving major innovation.

A stagnation like this has contributed to Samsung losing market share to competitors like TSMC in chip manufacturing and Apple in smartphones.

Adding to its challenges, Samsung has warned of sluggish AI chip sales due to US export restrictions to China, its biggest market. This puts the company at greater risk from potential US tariffs on Chinese trade.

In an attempt to regain investor confidence, Samsung launched a $7.2 billion share buyback plan in November, which has helped its stock recover slightly. However, shareholders remain sceptical about its future growth strategy.

For more information on these topics, visit diplomacy.edu.

Europe’s tech giants push for sovereign fund

More than 90 European technology companies and lobby groups, including Airbus and Dassault Systèmes, have called on European Commission President Ursula von der Leyen to establish a sovereign infrastructure fund.

In an open letter dated 14 March, they emphasised the urgent need for Europe to strengthen its strategic autonomy in critical digital infrastructure, from AI frameworks to semiconductor manufacturing.

The letter warns that Europe’s reliance on foreign technology creates security risks and weakens economic growth. It highlights the importance of public investment, particularly in capital-intensive sectors like quantum computing and microchips. The signatories also suggest a ‘buy European’ policy in government procurement to boost demand and encourage local businesses to invest.

Prominent supporters of the initiative include French cloud provider OVH Cloud, the European Software Institute, and the German AI Association. The appeal also reached EU tech chief Henna Virkkunen, as Europe faces increasing pressure to compete with major US and Asian technology powers.

For more information on these topics, visit diplomacy.edu.

Intel appoints new CEO to compete in AI chip market

Intel has appointed tech industry veteran Lip-Bu Tan as its chief executive, aiming to revitalise the struggling chipmaker as it falls behind in the AI race.

Tan, set to take over next week, told employees that overcoming Intel’s challenges would not be easy but reaffirmed his commitment to an engineering-first approach.

Following the announcement, Intel’s shares surged by more than 10 per cent in after-market trading.

Once a dominant force in the semiconductor industry, Intel has been outpaced by Taiwan Semiconductor Manufacturing Co (TSMC) and Samsung Electronics, which lead in made-to-order chip production.

It also lags behind Nvidia, which has emerged as the top AI chip provider. Tan replaces Pat Gelsinger, who was ousted last year after the board lost confidence in his turnaround efforts, which included cutting 15,000 jobs and delaying chipmaking projects.

Tan, previously head of Cadence Design Systems, pledged to restore Intel’s reputation by taking calculated risks to outmanoeuvre competitors.

He intends to continue the company’s plan to manufacture chips for other firms, directly challenging TSMC. However, analysts remain cautious, questioning whether Intel will split its foundry and chip design businesses or prove its ability to deliver cutting-edge technology.

Intel also faces a growing battle in AI, where Nvidia dominates the data centre chip market. Analysts warn that without a compelling AI strategy, Intel could struggle to regain investor confidence.

Tan, however, remains optimistic, vowing to transform Intel into a world-class chipmaker while ensuring customer satisfaction.

For more information on these topics, visit diplomacy.edu.

Singapore fraud case involves $390 million in transactions

Singapore prosecutors revealed on Thursday that a fraud case involving local firms accused of illegally supplying US servers to Malaysia involves transactions worth $390 million.

Three men—Singaporeans Aaron Woon and Alan Wei, along with Chinese national Li Ming—have been charged with deceiving tech giants Dell and Super Micro by misrepresenting the servers’ final destination.

The case has been linked to Chinese AI firm DeepSeek, which is under US scrutiny over the potential use of banned Nvidia chips.

While Singapore authorities confirmed the servers may have contained Nvidia components, they did not specify whether these were the restricted high-end semiconductors subject to US export controls.

Singapore’s Law and Home Affairs Minister K Shanmugam declined to comment on the alleged connection.

Prosecutors claim Wei paid himself tens of millions in dividends, while Woon received a multimillion-dollar bonus. Singaporean authorities are investigating a wider network of 22 individuals and companies suspected of similar fraudulent practices, with six additional arrests made.

The accused are set to reappear in court on May 2, while Malaysian authorities are also probing potential legal violations.

For more information on these topics, visit diplomacy.edu.

Celestial AI aims to rival Nvidia with innovative photonic technology

Celestial AI has raised an additional $250 million in venture capital, bringing its total funding to $515 million. The Silicon Valley startup is developing photonics-based technology to improve the speed and efficiency of AI computing.

By using light instead of electrical signals to connect AI chips with memory, the company aims to address the growing demand for higher memory bandwidth, a crucial factor in AI development.

Nvidia currently dominates this space with its NVLink and NVSwitch technologies, prompting a race among startups to develop alternative solutions.

Celestial AI’s ‘photonic fabric’ technology is designed to act as a high-speed bridge between chips, offering improved energy efficiency and lower latency. Backed by AMD’s venture arm, the company is positioning itself as a viable alternative to Nvidia’s proprietary systems.

The latest funding round was led by Fidelity Management & Research and included major investors such as BlackRock, Maverick Capital, and Tiger Global. Other participants included Temasek, Porsche Automobil Holding, and The Engine Ventures.

As AI hardware innovation accelerates, Celestial AI is among a growing group of startups seeking to reshape the industry with new approaches to chip design.

For more information on these topics, visit diplomacy.edu.

Meta has developed an AI chip to cut reliance on Nvidia, Reuters reports

Meta, the owner of Facebook, Instagram, and WhatsApp, is testing its first in-house chip designed for training AI systems, sources told Reuters.

The social media giant has started a limited rollout of the chip, planning to scale up production if testing delivers positive results. The move represents a crucial step in Meta’s strategy to lessen dependence on external suppliers like Nvidia and lower substantial infrastructure costs.

The company has projected expenses between $114 billion and $119 billion for 2025, with up to $65 billion dedicated to AI infrastructure.

The chip, part of Meta’s Meta Training and Inference Accelerator (MTIA) series, is a dedicated AI accelerator, meaning it is specifically designed for AI tasks rather than general processing. This could make it more power-efficient than traditional GPUs.

Meta is collaborating with Taiwan-based chip manufacturer TSMC to produce the new hardware. The test phase follows Meta’s first ‘tape-out’ of the chip, a crucial milestone in silicon development where an initial design is sent to a chip factory.

However, this process is costly and time-consuming, with no guarantee of success, and any failure would require repeating the tape-out step.

Meta has previously faced setbacks in its custom chip development, including scrapping an earlier version of an inference chip after poor test results. However, the company has since used another MTIA chip for AI-powered recommendations on Facebook and Instagram.

The new training chip aims to first enhance recommendation systems before expanding to generative AI applications like the chatbot Meta AI.

Meta executives hope to implement their own chips for AI training by 2026, although the company continues to be one of Nvidia’s biggest customers, investing heavily in GPUs for its AI operations.

The development comes as AI researchers increasingly question whether scaling up large language models by adding more computing power will continue to drive progress. The recent emergence of more efficient AI models, such as those from Chinese startup DeepSeek, has intensified these debates.

While Nvidia remains a dominant force in AI hardware, fluctuating investor confidence and broader market concerns have caused turbulence in the company’s stock value.

For more information on these topics, visit diplomacy.edu.