TSMC and Qualcomm highlight new directions for AI chips

• TSMC is using AI to design more efficient chips.
• Qualcomm unveiled new chips with remote management for PCs.

The spotlight this week has been on AI chips — and the growing push to make them more efficient.

AI processors, the engines behind today’s artificial intelligence systems, are notorious for consuming huge amounts of electricity. The challenge has become harder as AI models scale up, demanding more data and more compute power. On Wednesday, Taiwan Semiconductor Manufacturing Co. (TSMC), the world’s largest contract chipmaker and supplier to Nvidia, outlined its latest effort to rein in that power use. The company is turning to AI itself to design chips that run more efficiently, aiming for improvements of up to tenfold.

At a Silicon Valley conference, TSMC shared how it is combining new hardware approaches with AI-driven design. One method is packaging multiple “chiplets” together, rather than relying on a single large processor. Each chiplet can use a different technology, and when combined, they form a more flexible computing package. Nvidia’s current flagship AI servers can consume as much as 1,200 watts during heavy workloads. If those servers ran continuously, the power draw would equal what 1,000 US homes use. Cutting down that kind of energy demand is becoming critical as more AI data centres come online.

To make the most of new designs, chipmakers are also leaning on software that can speed up or even outperform human engineers. Companies like Cadence Design Systems and Synopsys introduced new AI-powered tools this week, developed closely with TSMC. These programs can handle complex parts of the design process in minutes, compared to days of work by skilled engineers.

“That helps to max out TSMC technology’s capability, and we find this is very useful,” said Jim Chang, deputy director of TSMC’s 3DIC Methodology Group. “This thing runs five minutes while our designer needs to work for two days.”

But progress on efficiency faces barriers that go beyond design. As chips grow more advanced, physical limits are creating bottlenecks. Moving data on and off chips using electrical connections, for instance, is running into speed and energy constraints. One solution could be optical links between chips, which promise faster, lower-power transfers. Making those systems reliable enough for large-scale data centres remains a work in progress. “Really, this is not an engineering problem,” said Kaushik Veeraraghavan, an engineer at Meta. “It’s a fundamental physical problem.”

Qualcomm broadens its chip strategy

While TSMC focused on design breakthroughs, Qualcomm used the same event to spotlight new processors for both phones and PCs. The San Diego-based company is best known for smartphone chips, which connect billions of devices to wireless networks. In recent years, however, Qualcomm has been pushing into the PC market, competing against Apple’s custom silicon and Intel’s long-running dominance.

One of the centrepieces of Qualcomm’s announcements was the Snapdragon X2 Elite, a laptop chip set to ship next year. Beyond raw performance, it introduces a new security feature aimed at corporate buyers called Guardian.

Kedar Kondap, Qualcomm’s senior vice president of gaming and compute, said Guardian allows IT departments to securely connect to company machines to install updates or provide technical support, even if the computers are turned off. For businesses managing thousands of laptops across a distributed workforce, that kind of access could save time and reduce downtime.

Intel, which still controls most of the corporate PC market, has provided remote management features for years. But Qualcomm plans to combine Guardian with its modem chips, making it possible to manage machines over 5G and other mobile networks. That means companies could reach their devices almost anywhere in the world with cell coverage, not just when they are on a company’s internal network.

“Nobody else can offer something like that,” said Ben Bajarin, chief executive of consultancy Creative Strategies. “I can actually see that being attractive for a portion of the workforce and something that will get stronger interest in Qualcomm for enterprise fleets.”

The Snapdragon X2 Elite is part of Qualcomm’s broader push to expand into high-performance PCs, a market where it has trailed rivals. By combining energy-efficient design with features like Guardian, the company is betting it can carve out a place not only among consumers but also in enterprise fleets that prize security and manageability. Devices powered by the new chip are expected to ship in 2026.

Chipmakers are all looking for ways to meet the demands of AI while keeping energy use in check. For TSMC, the path forward leans on AI-driven design and new packaging methods. For Qualcomm, the focus is on bringing more capable processors into PCs and pairing them with features that appeal to businesses. Both approaches point to the same reality: as AI spreads further into daily life and work, the chips behind it will need to be faster, smarterand far more efficient.

Want to experience the full spectrum of enterprise technology innovation? Join TechEx in Amsterdam, California, and London. Covering AI, Big Data, Cyber Security, IoT, Digital Transformation, Intelligent Automation, Edge Computing, and Data Centres, TechEx brings together global leaders to share real-world use cases and in-depth insights. Click here for more information.

TechHQ is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.